Article

Fully Automatic Segmentations of Liver and Hepatic Tumors From 3-D Computed Tomography Abdominal Images: Comparative Evaluation of Two Automatic Methods

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

An adaptive initialization method was developed to produce fully automatic processing frameworks based on graph-cut and gradient flow active contour algorithms. This method was applied to abdominal Computed Tomography (CT) images for segmentation of liver tissue and hepatic tumors. Twenty-five anonymized datasets were randomly collected from several radiology centres without specific request on acquisition parameter settings nor patient clinical situation as inclusion criteria. Resulting automatic segmentations of liver tissue and tumors were compared to their reference standard delineations manually performed by a specialist. Segmentation accuracy has been assessed through the following evaluation framework: dice similarity coefficient (DSC), false negative ratio (FNR), false positive ratio (FPR) and processing time. Regarding liver surfaces, graph-cuts achieved a DSC of 95.49% ( FPR=2.35% and FNR=5.10%), while active contours reached a DSC of 96.17% (FPR=3.35% and FNR=3.87%). The analyzed datasets presented 52 tumors: graph-cut algorithm detected 48 tumors with a DSC of 88.65%, while active contour algorithm detected only 44 tumors with a DSC of 87.10%. In addition, in terms of time performances, less time was requested for graph-cut algorithm with respect to active contour one. The implemented initialization method allows fully automatic segmentation leading to superior overall performances of graph-cut algorithm in terms of accuracy and processing time. The initialization method here presented resulted suitable and reliable for two different segmentation techniques and could be further extended.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... Based on the accuracy values of both the DSC and TC, we can conclude that the proposed automatic segmentation algorithm is robust and very accurate, particularly in comparison to the results of other automatic segmentation algorithms. For instance, in [76] the DSC measurement was applied to the results produced by two automated segmentation algorithms, namely, an active contour algorithm and a graph-cut algorithm, that were used to segment liver volume, and it was found that these algorithms achieved a segmentation accuracy of 96.17% and 95.49%, respectively. In another previous work [11], according to the DSC, accuracy results of 96.9% ± 0.08 were achieved for automatic lung field segmentation outputs. ...
... Thus, according to our knowledge, our proposed automatic segmentation algorithm outperforms other comparable methods in the literature according to both DSC and TC measurements of segmentation accuracy. Table 7 shows the accuracy results for our proposed automatic segmentation algorithm and the above-mentioned methods in [76], [11], [77], and [78], respectively. [76] 95.83% -Hybrid scheme: Rule-based segmentation pixel classification [11] 96.90% -Hierarchical atlas registration and weighting scheme [77] 91.51% 84.99% Convolutional neural network based on the leave-2-out scheme [78] 92% - ...
... Table 7 shows the accuracy results for our proposed automatic segmentation algorithm and the above-mentioned methods in [76], [11], [77], and [78], respectively. [76] 95.83% -Hybrid scheme: Rule-based segmentation pixel classification [11] 96.90% -Hierarchical atlas registration and weighting scheme [77] 91.51% 84.99% Convolutional neural network based on the leave-2-out scheme [78] 92% - ...
Article
Full-text available
The rendering of digitally reconstructed radiograph (DRR) images involves creating a digital reconstruction of an image made by a three-dimensional (3D) imaging system, such as a computed tomography (CT) scan, to produce a new two-dimensional (2D) image that emulates a medical X-ray scan. Acceleration of DRR generation has been the focus of much research in recent years, especially for medical applications that are designed to work on different imaging systems such as 2D/3D medical image registration. In this study, we succeeded in reducing the DRR generation time by using an automatic body segmentation approach on the CT volume. The proposed automatic body segmentation methodology consists of several steps that are carried out sequentially, starting with the use of fractional order Darwinian particle swarm optimization to generate a binary mask for all CT volume slices and ending with the use of a watershed transform to obtain the parameters required to identify the human body or part thereof as a confirmed region of interest (ROI). In our experiment, nearly half of the CT volume was identified as a non-ROI by the proposed approach. The approach also achieved an acceleration in the DRR generation process because the time complexity rate is reduced from O(N3) to O(M3), where M≈12N. That is to say, the DRR generation process was accelerated from 9.3 s when using the original (512 × 512 × 350) CT volume to 4.9 s when using the automated segmented CT volume, which involved the applications of the traditional ray-casting projection technique. The proposed methodology was applied by using a computer with an Intel® Core™ i7-7500 U processor, 8 GB RAM and 2.70 GHz processing speed. We used the two most commonly employed measurements, i.e., the dice similarity coefficient (DSC) and the Tanimoto coefficient (TC) to evaluate the segmentation results for the tested datasets. The results were also compared with ground truth images drawn by a specialist at the King Hussein Cancer Center in Jordan. The comparison of the automated segmented volume and its ground truth showed that the proposed methodology obtained 99.50% accuracy based on the DSC and 99.03% based on TC.
... An adaptive method was employed to perform the automatic segmentation on the abdominal CT image in the research of Casciaro et al (2012). The proposed method based on Graph-Cut and gradient flow active contour algorithms. ...
... From 3-D Computed Tomography Abdominal Images: Comparative Evaluation of Two Automatic Methods (Casciaro et al., 2012) Graph-Cut and gradient flow active contour algorithms. ...
... Figure 6.3 shows the segmentation accuracy values produced by the DSC and TC segmentation measurements when used on the automated segmented CT volume produced by our proposed segmentation algorithm. For instance, when Casciaro et al (2012) applied the DSC measurement to their automated segmentation results for the segmentation of liver volume using an active contour and a graphcut algorithm they got a segmentation accuracy of 96.17% and 95.49%, respectively. In another previous work, Ginneken and Romeny (2000) got accuracy results of 96.9% ± 0.08 when they evaluated their automatic lung field segmentation outputs. ...
Thesis
Full-text available
The Digitally Reconstructed Radiograph (DRR) process is considered to be an important procedure in different clinical applications that are depend on the medical images. The main applications of generated DRR images are in radiography and radiosurgery. The generation of DRR images involves creating a digital reconstruction of an image created by a Three-Dimensional (3D) imaging system such as a Computed Tomography (CT) scan or Magnetic Resonance Imaging (MRI) to produce a new Two-Dimensional (2D) image that emulates a medical X-ray scan. The DRR generation process assists medical specialists such as oncologists to determine the exact pathology for patients and helps them to decide what radiation or surgical intervention is required. Therefore, acceleration of DRR generation has been the focus of researchers in recent years, especially for medical applications that are designed to work on different imaging systems such as 2D/3D medical image registration. xv In this thesis, we have succeeded in reducing the DRR generation time by using an automatic body segmentation approach on CT volume. This acceleration in the processing time has been achieved by reducing the CT volume voxels that are involved in the DRR generation process. We have been able to obtain an acceleration rate that halves the time taken to generate a DRR image from automated segmented CT volume as compared to the time taken when using the original CT volume using traditional ray casting projection technique. That is to say, we have accelerated the DRR generation process from 9.3 seconds using the original CT volume to 4.9 seconds using the automated segmented CT volume. Regarding the quality of generated DRR images, it measured by calculating the peak signal-to-noise ratio (PSNR) between the DRR image of automated segmented CT volumes and ground truth DRR images. As average of PSNR values for al tested CT volumes, we get a value of 54.75 dB and it indicates to high image quality of generated DRR images based on automated segmented CT volume. The proposed automatic body segmentation methodology consists of five steps that are carried out sequentially: First, load CT volume as slices. Second, a binary mask is generated for all CT volume slices based on Otsu’s thresholding and mathematical morphology. Third, the Euclidean Distance Transform (EDT) is computed for all binary slices of CT volume with normalized results. Fourth, apply watershed transform based on marker-controlled segmentation between the extracted marker images and their original CT slices. Fifth and finally, the human organ will be segmented from CT slices by getting the final results of the required parameters for identifying the human organ as a confirmed Region of Interest (ROI). The automatic segmentation results produced by the proposed algorithm were found to have a high degree of accuracy. We used the most two common measurements to evaluate the segmentation results for the tested datasets, namely the Dice Similarity Coefficient (DSC) and the Tanimoto Coefficient (TC). Ground truth images were drawn by a specialist at the King Hussein Cancer Center. We obtained 99.50% accuracy when applying the DSC and 99.03 when applying the TC between the automated segmented volume and its ground truth. In our experiment, nearly half of the CT volume was identified as a non-ROI and neglected. The main reason for the acceleration of the DRR generation process is the reduction in the time complexity rate from O(N3) to O(M3) , where M3 ≈12N3 was proved by an experiment.
... Meanwhile, others were fully automatic and the user only had a verification role. In [9], Casciaro et al. developed an adaptive method to produce fully automatic processing frameworks based on graph-cut and gradient flow active contour algorithms. Their method was applied to abdominal computed tomography images for the segmentation of liver tissue and hepatic tumours. ...
... The key feature, of the method described in this paper, is that automatic segmentation does not require a priori knowledge of body region of interest; as opposed to previous works in the field, to segment lungs, pelvis, kidneys, etc. [3][4] [9]. This method only requires an automatic identification of body RoI and the background of non-RoI. ...
... Another advantage of this method is that it can be implemented as a priori step for other segmentation modalities; as this method is able to precisely eliminate the background from CT images. Consequently, this work would speed-up and refine segmentation results for different human organs, as in [3][4][5][6][7][8][9], or speed-up the image registration process, as in [25]. One limitation of this method is that it cannot be directly implemented in CT images where the intensity of part of the image is close to the background intensity values. ...
Conference Paper
Full-text available
Medical imaging segmentation provides vital information for surgical diagnosis, and usually demands an accurate segmentation. A fully automated computed tomography image segmentation method is proposed. This method is unsupervised and automatic estimation of the required parameters for identifying the human body as a region of interest. The proposed methodology consists of four steps: First, a body region of interest is masked by a method based on thresholding and basic morphological operations. Second, a body region of interest is identified using chain codes and a method for collecting adjacent contours. Next, the identification of background non-regions of interest is performed using an entropy algorithm. Finally, the human body segment is identified using a GrabCut algorithm. According to the visual evaluation results, segmentation of the human body, from the Computed Tomography images, was seen to be precise and accurate. The analysis provided evidence that the human body segmentation method could be applied to segmenting other organs, registering different image modalities or speeding-up the generation of digitally reconstructed radiographs.
... As discussed earlier, there is also a different categorization that is based on the extent of human intervention; some of them are automatic [41][42][43][44][45][46][47][48] that do not require human intervention for the generation of segmentation masks and some are semi-automatic requiring human assistance, say for seed selection or segmentation mask refinement [49][50][51][52][53]. Linguraru et al. [42] suggest an automatic segmentation approach for liver segmentation based on an affine invariant shape formulation. The paper makes a point-to-point comparison of various 3D surface features in the affine parameter space. ...
... Wang et al. [60] suggest a model-based algorithm to detect and segment bile duct carcinoma. Similar model-based approaches have been proposed for detecting and segmenting liver malignancies (e.g., HCC) [42,43]. ...
Article
Full-text available
Clinical imaging (e.g., magnetic resonance imaging and computed tomography) is a crucial adjunct for clinicians, aiding in the diagnosis of diseases and planning of appropriate interventions. This is especially true in malignant conditions such as hepatocellular carcinoma (HCC), where image segmentation (such as accurate delineation of liver and tumor) is the preliminary step taken by the clinicians to optimize diagnosis, staging, and treatment planning and intervention (e.g., transplantation, surgical resection, radiotherapy, PVE, embolization, etc). Thus, segmentation methods could potentially impact the diagnosis and treatment outcomes. This paper comprehensively reviews the literature (during the year 2012–2021) for relevant segmentation methods and proposes a broad categorization based on their clinical utility (i.e., surgical and radiological interventions) in HCC. The categorization is based on the parameters such as precision, accuracy, and automation.
... There has been also a different classification of the segmentation techniques that is based on the extent of human intervention; some of them are automatic [35], [36], [37], [38], [39], [40], [41], [42] that do not require human intervention for the generation of segmentation masks and some are semi-automatic requiring human assistance, say for seed selection or segmentation mask refinement [43], [44], [45], [46], [47]. Linguraru et al. [36] suggest an automatic segmentation approach for the liver segmentation based on an affine invariant shape formulation. ...
... Wang et al. [54] suggest a model-based algorithm to detect and segment bile duct carcinoma. Similar model-based approaches have been proposed for detecting and segmenting liver malignancies (e.g., HCC) [36], [37]. ...
Preprint
Full-text available
Medical images (e.g., magnetic resonance imaging (MRI) and computed tomography (CT)) provide critical information to the clinicians in order to diagnose pathology and plan interventions. Image segmentation is the first and foremost step taken by the clinicians while optimizing analytic diagnosis and treatment planning for interventions (e.g., transplantation and complete resection) and therapeutic procedures (e.g., radiotherapy, PVE, and embolization approaches), especially in hepatocellular carcinoma. Thus, segmentation techniques certainly impact the diagnosis and treatment outcomes. This paper studies the literature during the year 2012 until 2021 and reviews the segmentation methods classifying them into three categories based on their clinical utility (i.e., surgical and radiological interventions). The classification is based on the parameters such as precision, accuracy, location, liver condition, and other clinical considerations.
... All correlation coefficients of five data sets are around 0.985 after registration. Table 2 shows the comparison of the Dice similarity coefficient of the proposed method and the above-mentioned methods in [20,21,22,23]. ...
... The proposed method 98.92% Graph-cut and gradient flow active contour algorithms [20] 95.83% ...
Conference Paper
Full-text available
The radiological examination of congenital clubfoot is used not only in primary differential diagnostics and documentation, but frequently in follow-up as well. To eliminate the radiation risk, ultrasound is a relatively simple, and it is easily available, also, an inexpensive procedure for objective documentation of clubfoot correction. Image registration has a vast range of applications, and it is essential in order to study the evolution of pathology of a patient or to take full advantage of the complementary information coming from multimodal imagery. The purpose of the registration is to determine a transformation source image onto target image . Ideally, one wants to determine a displacement field to improve the evaluation of the anatomy and the mobility of talus and calcaneus in clubfeet. The deformable registration calculates the displacement field, by minimizing energy functional, consisting of a regularization term and sum of square residuals of corresponding two sets of N points, between the two images. The proposed method is successfully applied to estimate displacement vector fields to improve the evaluation of the anatomy and the mobility of talus and calcaneus in clubfeet using a registration multimodal medical method based on Partial Differential Equation (PDE). Keywords— Clubfoot equinovarus, partial differential equations, image registration, displacement field.
... Thus, an advanced filter may be required to reduce image noise without deterioration in edges of anatomical structures. Further, in [7], liver's regions were described by a statistical distribution model of gray-intensities. Nonetheless, this method excludes gradient information that possibly assists in the segmentation. ...
... This approach statistically describes liver regions by measuring mean µ and standard deviation σ of its gray intensities. These measures were mentioned in [7] [9] and we called them the statistical thresholding (ST) technique. According to those papers, this method starts from using a mean-shift filter [10] to reduce image noise. ...
Article
Full-text available
This paper presents an efficient liver-segmentation system developed by combining three ideas under the operations of a level-set method and consequent processes. First, an effective initial process creates mask and seed regions. The mask regions assist in prevention of leakage regions due to an overlap of gray-intensities between liver and another soft-tissue around ribs and verte-brae. The seed regions are allocated inside the liver to measure statistical values of its gray-intensities. Second, we introduce liver-corrective images to represent statistical regions of the liver and preserve edge information. These images help a geodesic active contour (GAC) to move without obstruction from high level of image noises. Lastly, the computation time in a level-set based on reaction-diffusion evolution and the GAC method is reduced by using a concept of multi-resolution. We applied the proposed system to 40 sets of 3D CT-liver data, which were acquired from four patients (10 different sets per patient) by a 4D-CT imaging system. The segmentation results showed 86.38% ± 4.26% (DSC: 91.38% ± 2.99%) of similarities to outlines of manual delineation provided by a radiologist. Meanwhile, the results of liver segmentation only using edge images presented 79.17% ± 5.15% or statistical regions showed 74.04% ± 9.77% of similarities.
... This approach however is time-consuming and segmentation results may not be as accurate as needed [4]. To save time, fully [5][6][7][8][9][10][11] or semi-automatic methods [12][13][14][15][16][17][18][19][20][21][22] have been investigated to segment the liver volume from abdomen CT images. From these two major types of algorithms, the automatic liver segmentation still possesses weaknesses and drawbacks. ...
... A comparison study between ten automatic and six interactive methods for liver segmentation from contrast-enhanced CT images showed that in general, the interactive methods reached higher average scores than the automatic approaches and featured a better consistency of segmentation quality [23]. Algorithms used for liver segmentation include grey level evaluation [6,11,[24][25][26][27], clustering [17,[28][29][30], region-based method [31,32], Snakesbased method [33], grow-cut [34], graph cuts [15,[35][36][37], level set [16,[38][39][40][41], combinations of different approaches as for example Snakes and grow-cut [33], or graph cut and gradient flow active contour [5], or morphological operations and graph cuts [9,42], grey level and a priori knowledge like CT numbers and location [25], hidden Markov measure field model [18], multi-class smoothed Bayesian classification [20,21], and edge based methods [43]. The use of segmentation algorithm based on priority knowledge about appearance, shape and size of the liver [10,23,[44][45][46][47][48][49][50][51][52] and methods based on neural networks [53,54] have been also proposed. ...
... Level set techniques are another popular approach that is used for liver lesion segmentation due to their ability to handle image noise, intensity heterogeneities, and discontinuous object boundaries (Casciaro et al., 2012;Li et al., 2013;Li et al., 2011). The semi-automatic method presented by Smeets et al. (2010) is based on a level set fitted on a fuzzy classification of the image data. ...
Preprint
Full-text available
We propose a novel method, the adaptive local window, for improving level set segmentation technique. The window is estimated separately for each contour point, over iterations of the segmentation process, and for each individual object. Our method considers the object scale, the spatial texture, and changes of the energy functional over iterations. Global and local statistics are considered by calculating several gray level co-occurrence matrices. We demonstrate the capabilities of the method in the domain of medical imaging for segmenting 233 images with liver lesions. To illustrate the strength of our method, those images were obtained by either Computed Tomography or Magnetic Resonance Imaging. Moreover, we analyzed images using three different energy models. We compare our method to a global level set segmentation and to local framework that uses predefined fixed-size square windows. The results indicate that our proposed method outperforms the other methods in terms of agreement with the manual marking and dependence on contour initialization or the energy model used. In case of complex lesions, such as low contrast lesions, heterogeneous lesions, or lesions with a noisy background, our method shows significantly better segmentation with an improvement of 0.25+- 0.13 in Dice similarity coefficient, compared with state of the art fixed-size local windows (Wilcoxon, p < 0.001).
... Looking at the error introduced by the operator's intervention, Casciaro et al. (2011) propose a graph-cut and 3-D initialization method for gradient vector flow (GVF) active contour approach for segmentation. The average intensity of the liver's statistical model distribution and its standard deviation serve as the foundation for this approach. ...
Article
Full-text available
The clinicians usually desire to know the shape of the liver during treatment planning to minimize the damage to the surrounding healthy tissues and hepatic vessels, thus, building the geometric model of the liver becomes paramount. There have been several liver image segmentation methods to build the model over the years. Considering the advantages of conventional image segmentation methods, this paper reviews them that spans over last 2 decades. The review examines about twenty-five automated and eleven semi-automatic approaches that include Probabilistic atlas, K-means, Model and knowledge-based (such as active appearance model, live wire), Graph cut, Region growing, Active contour-based, Expectation Maximization-based, Level sets, Laplacian network optimization, etc. The main contribution of this paper is to highlight their clinical suitability by providing their advantages and possible limitations. It is nearly impossible to assess the methodologies on a single scale because a common patient database is usually not used, rather, diverse datasets such as MICCAI 2007 Grand Challenge (Sliver), 3DIRCADb, Zhu Jiang Hospital of Southern Medical University (China) and others have been used. As a result, this study depends on the popular metrics such as FPR, FNR, AER, JCS, ASSD, DSC, VOE, and RMSD. offering a sense of efficacy of each approach. It is found that while automatic segmentation methods perform better technically, they are usually less preferred by the clinicians. Since the objective of this paper is to provide a holistic view of all the conventional methods from clinicians’ stand point, we have suggested a conventional framework based on the findings in this paper. We have also included a few research challenges that the readers could find them interesting.
... Contouring performance was evaluated by two quantitative metrics, including volumetric Dice [30,38,39] for volume overlapping evaluation and surface Dice [30,39,40] for border overlapping evaluation with a distance tolerance of 1 mm. To evaluate the subjective acceptability of OAR DLCS , an objective volume-based and a surface-based oncologist satisfaction rate (VOSR and SOSR) were defined by the OAR DLCS and the corresponding OAR p-DLCS to objectively measure the volume and surface deviation of DLCS contouring to oncologists' desired targets. ...
Article
Full-text available
Background: To comprehensively investigate the behaviors of oncologists with different working experiences and institute group styles in deep learning-based organs-at-risk (OAR) contouring. Methods: A deep learning-based contouring system (DLCS) was modeled from 188 CT datasets of patients with nasopharyngeal carcinoma (NPC) in institute A. Three institute oncology groups, A, B, and C, were included; each contained a beginner and an expert. For each of the 28 OARs, two trials were performed with manual contouring first and post-DLCS edition later, for ten test cases. Contouring performance and group consistency were quantified by volumetric and surface Dice coefficients. A volume-based and a surface-based oncologist satisfaction rate (VOSR and SOSR) were defined to evaluate the oncologists' acceptance of DLCS. Results: Based on DLCS, experience inconsistency was eliminated. Intra-institute consistency was eliminated for group C but still existed for group A and group B. Group C benefits most from DLCS with the highest number of improved OARs (8 for volumetric Dice and 10 for surface Dice), followed by group B. Beginners obtained more numbers of improved OARs than experts (7 v.s. 4 in volumetric Dice and 5 v.s. 4 in surface Dice). VOSR and SOSR varied for institute groups, but the rates of beginners were all significantly higher than those of experts for OARs with experience group significance. A remarkable positive linear relationship was found between VOSR and post-DLCS edition volumetric Dice with a coefficient of 0.78. Conclusions: The DLCS was effective for various institutes and the beginners benefited more than the experts.
... Other limitation is that in some cases the boundary between liver and neighboring organs disappears and it becomes hard to determine the difference among different organs and this leads to erroneous segmentation. Casciaro et al. gives an adaptive initialization method to produce fully automatic processing frameworks based on graph-cut and gradient flow active contour algorithms [9]. They used mean shift filter for pre-processing because mean shift filter preserve edges and does not blur edges. ...
Article
Full-text available
Segmentation of liver is the initial and fundamental step for the diagnosis of liver disease, 3-D volume construction and volume measurement. However, segmentation of liver is a challenging task due to it's inter and intra intensity and texture similarities among other organs present in CT abdominal images. A semiautomatic method has been proposed to segment the liver portion from CT abdominal images and their three dimensional volume construction by (i)Noise removal using median filter,(ii)Segmentation of liver portion based on active contour method using sparse field method and(iii)liver volume construction using marching cube method. Evaluation of proposed method is carried out on clinically acquired CT images and effectiveness of algorithm is evaluated by comparing manually segmented liver portion marked by radiologist with proposed method.
... Typical values for an image are between 30dB and 50dB, when the PSNR is greater than dB. Dice Similarity metric is always between 0 and 1 with higher values returning a better match between automatic and manual segmentation (Casciaro et al., 2012). ...
Article
Full-text available
Hepatocellular carcinoma (HCC) is the most common primary malignant neoplasm of the liver representing the fifth most common malignancy worldwide. This tumor is more common in men than women, with a ratio of 2.7:1. Unlike HCC, Dysplasia is the precancerous nature of liver nodules and is characterized by cellular and nuclear enlargement, nuclear pleomorphism, and multinucleation. Area based Adaptive Expectation Maximization (EM) uses texture, layout, and context features of cells, and grows clusters to obtain texton maps of nucleus. A discriminative model of nucleus and cytoplastic changes of tumor is built by incorporating texture, layout, and context information efficiently. A bootsrap regression model of nuclei and cytoplastic changes are built by incorporating the aforementioned features efficiently. Mean squared error, Peak Signal to Noise ratio and Dice similarity values are used to evaluate the method's classification performance. The proposed method provides high classification and segmentation accuracy of nucleus and extra nuclear content in HCC and dysplasia, which are exceedingly textured in histopathology images, when compared to Adaptive K means, EM method and the state-of-the-art method, Convolutional Neural Networks (CNN). As texton detection reduces the cluttered background of nuclei, the proposed method would be a convenient mechanism for the classification of nuclei and non-nuclear features. In conclusion, this system can detect more eligible cells of precancerous nature as well as malignant cells even in a cluttered background of nuclei.
... Similarly, Li et al. [22] used active contour and semi-automatic separation in CT images, but they found little discrepancy in input CT images of the sample liver. Casciara et al. [23] performed automatic separation of liver tissue and hepaticinduced tumors through a three-dimensional model of CT images. In this study, two solutions were presented and the results were compared. ...
Article
Full-text available
Liver cancer emerges as a mass in the right upper of the abdomen with general symptoms such as jaundice and weakness. In recent years, liver cancer has been responsible for increasing the rate of deaths. Due to some discrepancies in the analytical results of CT images and the disagreement among specialists about different parts of the liver, accurate diagnosis of possible conditions requires skill, experience, and precision. In this paper, a new integrative model based on image processing techniques and machine learning is provided, which is used for segmentation of damages caused by the liver disease on CT images. The implementation process consists of three steps: (1) using discrete wavelet transform to remove noise and separate the region of interest (ROI) in the image; (2) creating the recognition pattern based on feature extraction by Gray-Level Co-occurrence matrix and hierarchical visual HMAX model; reducing the feature dimensions is also optimized by principal component analysis and support vector machine (SVM) classification, and finally (3) evaluating the algorithm performance by using K-fold method. The results of implementation were satisfactory both in performance evaluation and use of features selection. The mean recognition accuracy on test images was 91.7%. The implementation was in the presence of both descriptors irrespective of feature dimension reduction; with unique HMAX model and feature dimension reduction and application of both descriptors and reduction of feature dimensions and their effect on recognition were measured.
... That is still interesting because of the formation of imperfect droplets of functionalizing materials around nanoparticle core. The proposed segmentation has many applications in the field of medicine [18] [19]. Moreover special filtering is necessary by paying attention on edge detection and circular hough transform. ...
... Note that the groundtruth segmentation result is manually created, which meets overall integrity and consistency of detail to some extent. Here, we adopt "false-positive ratio (FPR)," "false-negative ratio (FNR)" and "error ratio (ER)" as objective evaluation criteria (Casciaro et al. 2012;Davanzo et al. 2011), and the concrete definition is as follows: ...
Article
Full-text available
In this paper, an unsupervised image segmentation technique is proposed. Firstly, for obtaining a multiresolution representation of the original image, the probability model of the nonsubsampled contourlet coefficients of the image is established. A region-based active contour model is then applied to the multiresolution representation for segmenting the image. The proposed technique has been conducted on challenging images to illustrate the robust and accurate segmentations. At last, an in-depth study of the behaviors of the above techniques in response to the proposed model is given, and the segmentation results are compared with several state-of-the-art methods.
... Validation of the medical images is an important task to check the accuracy, robustness, consistency, etc. Some of the validation techniques used in several papers are Dice similarity coefficient [7][8][9], Jaccard index [10][11][12], relative volume difference [13], etc. ...
... The level set algorithm is one of the efficient automatic segmentation method, and has been widely used and popularized in the recent ten years. Level set algorithm is one of the popular approaches that is used for medical ultrasound segmentations due to their performance to handle image noise, intensity heterogeneities, and discontinuous object boundaries (Casciaro et al., 2012;Li et al., 2013). The semi-automatic method is a level set method using a fuzzy classification for the image data, reported by Smeets et al. in 2010. ...
Article
Automatic image segmentation for phased array ultrasonic testing image is a developing area for testing the traditional and new segmentation methods. The precision of the segmentation is the most significant topics. This research selected C-V model utilizing in segmenting ultrasonic image and analyses the key technologies that would affect the segmentation accuracy: parameters updating, iterative process and iterative times selection. The method to control the parameters during updating procedure was determined, through the Neumann boundary condition, the Heaviside function, the Dirac function and the curvature function. The iterative process was implemented successfully and the iteration times were efficient. And we proposed a precision test experiment based on the defects centers. The method was comparing each defect center coordinate calculated by automatic level set segmentation with its manual coordinate, then calculated the relative error. The experiment results show that the algorithm used in this article performs with satisfied accuracy and high repetition precision for phased array ultrasonic testing image with high noise and fuzzy edges.
... There is no preassumption about shape and intensity range the limitation of the method was leakage in cases where liver contour is not clear due to inhomogeneity. Casciaro et al. developed an adaptive initialization method to produce fully automatic processing frameworks based on graph-cut and GVF active contour algorithms [7]. Mean shift filter with 64 squared regions are employed for initialization. ...
... The proposed method has a PSNR value of 72.16 dB, which is considerably higher than that of the existing methods. When DSC metric [8] is applied between the automatic and the manual segmentation, the values obtained are always between 0 to 1 where, the higher values are considered as the better match which shows that the accuracy has improved. ...
Article
Full-text available
The differentiation of a cluster of nuclei and multi-nucleation is a critical issue in automated diagnosis systems. Due to the similarities between said clusters and malignant nuclei, misclassification of these regions can affect the automated systems’ final decision. In this paper, a method for differentiating clusters from multi nucleated cells in histopathological images is proposed. Hepatocellular Carcinoma(HCC) and Dysplasia are characterized by cellular and nuclear enlargement, nuclear pleomorphism and multinucleation, which possess prominent threat Data was obtained from Global Hospital and Research Center from patients diagnosed with Hepatocellular Carcinoma and Dysplasia. This paper introduces a hybrid diagnosis method that uses texture, layout and context features of nuclei and cytoplastic cells in order to enhance the poor diagnosis of liver tumors in Infra Red (IR) images. We propose a Area based Adaptive Expectation Maximization(EM) that grows the clusters, which avoids the need for initial cluster selection in order to obtain texton maps of nuclei and cytoplasm. A linear regression model of nuclei and cytoplastic changes were built by incorporating the aforementioned features efficiently. The proposed method provides better classification and segmentation accuracy of nuclei and extra nuclear content in HCC and dysplasia, compared to the state-of-the-art methods like convolutional networks and classical methods like Adaptive K means and EM method in constant time. In conclusion, this system detects the malignant cells and the highly eligible precancerous cells which is cost effective and reproducible.
... Implemented methods therefore use this algorithm in combination with a robust initialization method that can provide an as close as possible solution so that the contours converge in few iterations. In the context of liver segmentation, many solutions were proposed to this well known initialization problem, such as k-means clustering [15], adaptive thresholds [16], [17], Markov random fields [18] and texture classification [19]. However, even then the active contours become highly dependent on the initialization method used and will tend to diverge in the absence of image information, or will not be able to reach complex details if the rigidity is too strong. ...
Article
Objective: The purpose of this paper is to describe a semiautomated segmentation method for the liver and evaluate its performance on CT-scan and MR images. Methods: First, an approximate 3-D model of the liver is initialized from a few user-generated contours to globally outline the liver shape. The model is then automatically deformed by a Laplacian mesh optimization scheme until it precisely delineates the patient's liver. A correction tool was implemented to allow the user to improve the segmentation until satisfaction. Results: The proposed method was tested against 30 CT-scans from the SLIVER07 challenge repository and 20 MR studies from the Montreal University Hospital Center, covering a wide spectrum of liver morphologies and pathologies. The average volumetric overlap error was 5.1% for CT and 7.6% for MRI and the average segmentation time was 6 min. Conclusion: The obtained results show that the proposed method is efficient, reliable, and could effectively be used routinely in the clinical setting. Significance: The proposed approach can alleviate the cumbersome and tedious process of slice-wise segmentation required for precise hepatic volumetry, virtual surgery, and treatment planning.
... In this paper we have presented a useful approach of CT and HRCT -based biomedical images of people suffering from lung diseases. The approach is economically cheaper than those using other techniques using nanoparticles [38] and segmentations [39] for better targeting the region of interest. These anomalies occur in the lung as bulges of trophic vessels and so they are characterized by circular shape with a gradation of gray closer to white. ...
Article
Full-text available
Diagnosis by imaging is one of the most important findings in biomedical imaging because it allows not only the diagnosing of a specific pathology but to perform online and offline surgical operations using imaging as it is noticed in interventional radiology. This paper illustrates the use Hough transform in identifying pathological structures included in CT (Computer Tomography) and HRCT (High Resolution Computer Tomography) images related to patients suffering from lung disease. These abnormal areas appear as bulges of the trophic vessels and they are similar to circular structures with level of lighter gray near to white. Circular Hough transform (CHT) identifies regions with a circular shape. However, a metrics is defined in order to understand if the pointed out area has a pathological morphology. CHT is used here for helping to detect possible events of indolent tumors or undetermined significance pathologies for lung apparatus. For this aim, we use entropy approach with CHT because it measures the scatter of the directional elements in an image. In fact a high entropy value is related to areas with a strong contrast in grayscale, and abnormalities in the image are present as a set of points with more lighter than the dark background. The results have shown, by means of an accuracy true table, rendering a comparison between clinicians'diagnosis and CHT detection, it is possible to indicate, with a better accuracy, potential areas of undetermined significance pathologies. Finally, a receiver operational curve (ROC) is used as an accuracy index for evaluating the positive impact of entropy on diagnosis.
... The development of innovative systems for computer-aided medical diagnoses is involving a progressively increasing number of clinical applications [1]. In fact, last-generation algorithms for fully automatic processing of biomedical images are being continuously developed and implemented, in order to extract quantitative and objective information regarding, for instance, contrast agent presence [2], tumour features [3], vessel morphology [4], real-time position of endovascular devices [5], childbirth labour progression [6]. ...
Article
Aim of this study was to perform a detailed clinical validation of a novel fully automatic method for vertebral morphometry. About 80 spine lateral radiographs were evaluated both automatically, by the proposed algorithm, and manually, by an experienced radiologist. The following metrics were used for algorithm performance assessment: Sensitivity and specificity in vertebra detection; errors in the localisation of characteristic points of vertebral border; errors in the measurement of six diagnostic parameters; level of agreement and correlation between manual and automatic morphometric measurements; overall accuracy of automatic diagnoses with respect to manual ones. Obtained results showed a very good performance in vertebra detection (sensitivity = 89.1% and specificity = 100.0%). Average errors in the localisation of vertebral characteristic points were always smaller than 3 mm (range 0.85-2.79 mm), causing relative errors in diagnostic parameter values ranging from -5.01 to +6.10%. Bland-Altman analysis documented a mean error in automatic measurements of diagnostic ratios of 0.01 ± 0.18 (bias ± 2 SDs), while Pearson's correlation coefficient resulted r = 0.71 (p < 0.001). Finally, an optimal diagnostic coincidence (92.8%) was found between automatic and manual diagnoses. Therefore, the adopted method has a potential for an effective employment in clinical routine for reliable diagnosis of vertebral fractures.
... The development of innovative systems for computer-aided medical diagnoses is involving a progressively increasing number of clinical applications [1]. In fact, last-generation algorithms for fully automatic processing of biomedical images are being continuously developed and implemented, in order to extract quantitative and objective information regarding, for instance, contrast agent presence [2], tumour features [3], vessel morphology [4], real-time position of endovascular devices [5], childbirth labour progression [6]. ...
Article
Full-text available
Vertebral morphometry is a common clinically-used method for vertebral fracture detection and classification, based on height measurements of vertebral bodies in radiographic images. This method is quantitative and does not require specific operator skills, but its actual accuracy is affected by errors made during the timeconsuming manual or semi-automatic measurements. In this paper, we propose an innovative fully automatic approach to vertebral morphometry. A novel algorithm, based on a local phase symmetry measure and an "Active Shape Model", was implemented and tested on lateral X-ray radiographs of 50 patients. Thoracic and lumbar vertebral bodies in each image were independently segmented and measured by both the automatic algorithm and an experienced radiologist, whose manually-obtained results were assumed as the ground truth. The algorithm showed reasonably low error rates regarding both vertebral localization and morphometric measurements with a sensitivity of 86.5% and a perfect specificity of 100%, because no false positive were present. Furthermore, its performance did not appreciably worsen on poor quality images, emphasizing a significant potential for a prompt translation into clinical routine. Copyright © (2014) by the International Measurement Confederation (IMEKO) All rights reserved.
... Further, to improve the effectiveness of denoising on MRI, the concept of self-similarity [30] was utilized by Buades et al. [31] to develop a Non-Local Means (NLM) approach for natural images corrupted with Gaussian noise. NLM was further improved by Manjon et al. by optimizing its tuning parameters [32], making it adaptive to spatially varying noise levels [33] and improving its computational complexity [34]. In this paper, a Non-Local Averaging approach has been used to suppress noise in Gaussian corrupted MRI, where, the restored values of pixels have been calculated on the basis of self-similar structures. ...
Chapter
Full-text available
Magnetic Resonance Images (MRI) are known to be corrupted by the additive Gaussian noise during the acquisition process. The presence of this noise affects the diagnosis as it tends to alter image details and pixel intensities. Conventional iterative denoising approaches fail to preserve the details and structures during MRI restoration. This paper proposes a Non-Local Averaging based MRI denoising algorithm to facilitate preservation of the finer structures. The proposed algorithm computes the weighted average of the similar pixels of the image within the local window. Method noise has been used as a measure for detail preservation which corresponds to the difference between original and the restored image. Simulation trials are performed on the image at differing levels of Gaussian noise which are then justified by method noise analysis and performance evaluation factors such as Peak Signal-Noise Ratio (PSNR) and Structural Similarity (SSIM). The proposed algorithm has demonstrated good performance, both in terms of visual quality as well as values of performance parameter.
... For two input images A, B and the fused image F, fusion factor is given by Eq. (2). FF = I AF + I BF (2) where: I AF and I BF are mutual information between source images and the fused image. Higher values of fusion factor indicate better fusion results. ...
Conference Paper
Full-text available
Medical image fusion facilitates the retrieval of complementary information from medical images for diagnostic purposes. This paper presents a combination of Principal Component Analysis (PCA) and Dual Tree Complex Wavelet (DTCWT) as an improved fusion approach for MR and CT-scan images. Unlike real valued discrete wavelet transforms, DTCWT provides shift invariance and improved directionality along with preservation of spectral content. The decomposed images are then processed using PCA a based fusion rule to improve upon the resolution and reduce the redundancy. Simulation results demonstrate an improvement in visual quality of the fused image supported by higher values of fusion metrics; this further justifies the effectiveness of the proposed approach in comparison to other approaches.
... The main advantage is related to its linearity and its relation to the heat PDE, about which much is known from physics and mathematics. The big disadvantage of Gaussian scale-space approach is the fact that linear smoothers blur and shift important image features, for examples edges [3] [4]. ...
Conference Paper
Full-text available
Human body has its basal temperature that can be exploited for different uses. Obviously, thermal properties of human tissues allow to retrieve specific characteristics if special attention is paid. Therefore, thermal imaging is a suitable tool for many applications. It could be used in medical issues for checking temperature variations displayed by a volume under investigation during surgery operations related to humans. It plays a double role: imaging and temperature measurements. This paper presents a wide and joint experimental research for determining the decreasing temperature encompassed in a volume during an urological intervention in order to established, in a realtime, which tissues and adherences must be taken in consideration for continuing and optimizing the surgery process. A further processing is performed using Hough transform that exhibits encouraging results for this specific approach. This is very important for some other applications like contour extraction, contour matching and surface spline fitting.
... It provides support to the subsequent marshaling and de-marshaling of all the defined argument types and object dispatch addresses as per the specifications for the underlying network. The concept of integration, interoperability, TEDS calibration, synchronized measurements and enhanced post-processing at the STIM can be also extended to medical measurements diagnosis via other modalities like: computer-tomography [48], ultrasound [49], [50], molecular imaging [51], [52], EEG [53], [54], ECG [55] and diffusion tensor imaging [56] and physiological monitoring related to diagnosis of pulmonary edema [57]- [61]. ...
Article
Full-text available
Mobile mammography programs provide a vast access to breast cancer screening and also ensure participation from remote areas which are generally deprived of fixed centers. However, there exists, limitations to provide on the spot analysis and diagnosis by experienced radiologists. The rapid development of ISO/IEC/IEEE 21451-x Standards has lead to the emergence of smart bio-medical devices for screening and detection of life threatening diseases. IEEE 21451-1 provides for networking of these smart transducers, yielding economical and reliable solutions for diverse measurements, control and clinical applications. The present work proposes the technical advancements to Mobile mammography by integrating Smart Transducer Interface (STIM) and Network Capable Application Processor (NCAP) Information Modules with digital mammography (X-ray) units. The digital data extracted by the hardware interface is processed using Type-0 and Type-I Non-Linear Polynomial Filters (NPF) for improving the contrast of the Region-of-Interest (ROI) as well as signal-to-noise ratios. NPF serves as a post-processing module, providing enhancement of mammograms for the purpose of precise analysis and diagnosis by radiologists. The enhanced mammograms are then transmitted to a remote monitoring station via NCAP through an optical fiber or satellite communication link. The proposed model yields a network as well as vendor independent transducer interfaces for remote screening/diagnosis of breast cancer patients.
Conference Paper
Liver segmentation is detrimental. The toughest segment is the liver. Segment CAD abdominal CT scans in order to look for liver cancers. The tumor’s size, shape, location, and other objects with comparable intensities in the CT scans make automatic tumor segmentation difficult. Therefore, precise tumor segmentation was originally made possible by liver segmentation. Since MRI is essential in medicine, we are concentrating on domain management from MRI to CT volumes, utilizing 3D and 2D liver segmentation. As a result, we must provide automated liver segmentation in CT pictures. Utilizing cuckoo optimization, fuzzy c means, and the random walker’s method, clinical data from patients was segmented. The suggested method was validated using a clinical liver dataset with one of the highest numbers of tumors for liver tumor segmentation. Zones impacted by liver illness are separated using fuzzy clustering. Liver boundary data is segmented using fuzzy C means, fuzzy clustering, and SVM classification. The user may choose a location of interest and do the contour operations again to increase accuracy with spatial liver boundary limitations.
Article
Full-text available
Image processing has become an essential component in many fields of biomedical research such as tumor detection, automatically determining the volume of a heart chamber, screening lung scans for possible diseases. Different techniques for automatic detection of liver tumor involve various steps: image acquisition, segmentation, classification using neural network and optimization, and identification of tumor type. Most common segmentation approaches are: Region based, Threshold based, Level set, Clustering based and Edge detection. Liver tumor segmentation is done with region based approach in this research work. Region based methods partition an image into regions that are similar according to a set of predefined criteria where as other segmentation approaches like edge detection methods partition an image according to rapid changes in intensity near the edges. In this research work Particle Swarm Optimization (PSO) and Seeker Optimization algorithm (SOA) have been compared for classification of tumor using CT scan images. The main focus of this work is to detect liver tumor and compare results of PSO and SOA in term of detection and classification accuracy and elapsed time. Region based segmentation approach has been used for segmentation of liver and liver tumor from CT scan images. PSO and SOA are used for classification and PSO optimization gives better results in term of detection and classification accuracy and elapsed time. For liver tumor classification, PSO results with as 93.3% detection and classification accuracy where as SOA results in 60% detection and classification accuracy.
Article
The process of functionalization in the field of nanotechnology is an important area of scientific and technical activities. Functionalization involves different expertise insofar as its complexity impacts physics, chemistry, electronics, measurement and instrumentation, and materials science. The research presented in this paper proposes criteria for verifying silica nanoparticles in terms of shape and dimensions based on processing transmission electron microscopy (TEM) images related to the nanoparticles using two techniques: monadic vision, mostly employed in robotics, and particle image velocimetry (PIV) used in fluid motion characterization through imaging. The first technique is connected to the characterization of the dimensions of objects by a robot having vision. The second, used in fluid velocimetry, is related to motion. Here, we do not deal with motion in the traditional meaning but during functionalization droplets of materials are generally seen. So droplets can be considered as quasi-motion from the main body to the outside due to centrifugal forces. In any case, both algorithms yield to shape and dimension determination for assessing the quality of functionalization.
Article
Objectives To develop and evaluate a deep learning–based model capable of detecting primary hepatic malignancies in multiphase CT images of patients at high risk for hepatocellular carcinoma (HCC).MethodsA total of 1350 multiphase CT scans of 1280 hepatic malignancies (1202 HCCs and 78 non-HCCs) in 1320 patients at high risk for HCC were retrospectively analyzed. Following the delineation of the focal hepatic lesions according to reference standards, the CT scans were categorized randomly into the training (568 scans), tuning (193 scans), and test (589 scans) sets. Multiphase CT information was subjected to multichannel integration, and livers were automatically segmented before model development. A deep learning–based model capable of detecting malignancies was developed using a mask region–based convolutional neural network. The thresholds of the prediction score and the intersection over union were determined on the tuning set corresponding to the highest sensitivity with < 5 false-positive cases per CT scan. The sensitivity and the number of false-positives of the proposed model on the test set were calculated. Potential causes of false-negatives and false-positives on the test set were analyzed.ResultsThis model exhibited a sensitivity of 84.8% with 4.80 false-positives per CT scan on the test set. The most frequent potential causes of false-negatives and false-positives were determined to be atypical enhancement patterns for HCC (71.7%) and registration/segmentation errors (42.7%), respectively.Conclusions The proposed deep learning–based model developed to automatically detect primary hepatic malignancies exhibited an 84.8% of sensitivity with 4.80 false-positives per CT scan in the test set.Key Points • Image processing, including multichannel integration of multiphase CT and automatic liver segmentation, enabled the application of a deep learning–based model to detect primary hepatic malignancy. • Our model exhibited a sensitivity of 84.8% with a false-positive rate of 4.80 per CT scan.
Article
Background and purpose: Manual delineation of clinical target volumes (CTVs) and organs at risk (OARs) is time-consuming, and automatic contouring tools lack clinical validation. We aimed to construct and validate the use of convolutional neural networks (CNNs) to set better contouring standards for rectal cancer radiotherapy. Materials and methods: We retrospectively collected and evaluated computed tomography (CT) scans of 199 rectal cancer patients treated at our hospital from February 2018 to April 2019. Two CNNs-DeepLabv3+ for extracting high-level semantic information and ResUNet for extracting low-level visual features-were used for the CTV and small intestine contouring, and bladder and femoral head contouring, respectively. Contouring quality was compared using the paired t test. Five-point objective grading was performed independently by two experienced radiation oncologists and verified by a third. The CNN manual correction time was recorded. Results: CTVs calculated using DeepLabv3+ (CTVDeepLabv3+) had significant quantitative parameter advantages over CTVResUNet (volumetric Dice coefficient, 0.88 vs 0.87, P = 0.0005; surface Dice coefficient, 0.79 vs 0.78, P = 0.008). Among 315 graded cases, DeepLabv3+ obtained the highest scores with 284 cases, consistent with the objective criteria, whereas CTVResUNet had the minimum mean manual correction time (7.29 min). DeepLabv3+ performed better than ResUNet for small intestine contouring and ResUNet performed better for bladder and femoral head contouring. The manual correction time for OARs was <4 min for both models. Conclusion: CNNs at various feature resolution levels well delineate rectal cancer CTVs and OARs, displaying high quality and requiring shorter computation and manual correction time.
Article
With the evolution of technology, it is now possible to build more flexible surgical instruments with multiple degrees of freedom (DOFs) that can be deployed in Natural Orifice Translumenal Endoscopic Surgery (NOTES), further reducing tissue damage, blood loss and recovery time. However, the limits of flexible robots are mainly represented by the lack of accurate position feedback during the surgical procedure. The most critical problems of shape sensors are size and quantity limits, as well as sterilizability. This paper presents a novel helical configuration of single optic fiber with multiple fiber Bragg grating (FBG) sensors to measure the shape of flexible robots. Detailed design of helix parameters, simulation and verification experiments were performed. We have shown the feasibility of helical configuration to reduce the effect of temperature, and detect the torsion and bending of the flexible shaft. The experimental results demonstrated the potential to provide a shape detection for flexible robot.
Article
Full-text available
Computed tomography (CT) imaging remains the most utilized modality for liver-related cancer screening and treatment monitoring purposes. Liver, liver tumor and liver vasculature segmentation from CT data is a prerequisite for treatment planning and computer assisted detection/diagnosis systems. In this paper, we present a survey on liver, liver tumor and liver vasculature segmentation methods that are using CT images, recent methods presented in the literature are viewed and discussed along with positives, negatives and statistical performance of these methods. Liver computer assisted detection/diagnosis systems will also be discussed along with their limitations and possible ways of improvement. In this paper, we concluded that although there is still room for improvement, automatic liver segmentation methods have become comparable to human segmentation. However, the performance of liver tumor segmentation methods can be considered lower than expected in both automatic and semi-automatic methods. Furthermore, it can be seen that most computer assisted detection/diagnosis systems require manual segmentation of liver and liver tumors, limiting clinical applicability of these systems. Liver, liver tumor and liver vasculature segmentation is still an open problem since various weaknesses and drawbacks of these methods can still be addressed and improved especially in tumor and vasculature segmentation along with computer assisted detection/diagnosis systems.
Article
Purpose: In 2005, an application for surgical planning called AYRA[Formula: see text] was designed and validated by different surgeons and engineers at the Virgen del Rocío University Hospital, Seville (Spain). However, the segmentation methods included in AYRA and in other surgical planning applications are not able to segment accurately tumors that appear in soft tissue. The aims of this paper are to offer an exhaustive validation of an accurate semiautomatic segmentation tool to delimitate retroperitoneal tumors from CT images and to aid physicians in planning both radiotherapy doses and surgery. Methods: A panel of 6 experts manually segmented 11 cases of tumors, and the segmentation results were compared exhaustively with: the results provided by a surgical planning tool (AYRA), the segmentations obtained using a radiotherapy treatment planning system (Pinnacle[Formula: see text]), the segmentation results obtained by a group of experts in the delimitation of retroperitoneal tumors and the segmentation results using the algorithm under validation. Results: 11 cases of retroperitoneal tumors were tested. The proposed algorithm provided accurate results regarding the segmentation of the tumor. Moreover, the algorithm requires minimal computational time-an average of 90.5% less than that required when manually contouring the same tumor. Conclusion: A method developed for the semiautomatic selection of retroperitoneal tumor has been validated in depth. AYRA, as well as other surgical and radiotherapy planning tools, could be greatly improved by including this algorithm.
Conference Paper
We propose a novel external force for active contours by normalizing normal and tangent components of the gradient vector flow (GVF) along a generalized edge within the iteration process. The normal and tangent components normalization based GVF (NTCN-GVF) is inspired by the CN-GGVF which normalizes the x- and y-components of the generalized GVF (GGVF) to strengthen the smaller downward component of the external force within the long and thin indentations (LTIs). However, the strengthening effect is sensitive to the orientation of LTIs and excessive in homogeneous areas. NTCN-GVF behaves like CN-GGVF along the edge and conventional vector-based normalized GVF in homogeneous areas. Consequently, the NTCN-GVF snake can capture differently orientated LTIs and preserve weak edges while maintaining other desirable properties of enlarged capture range and noise robustness. Finally, experimental results are presented to verify the effectiveness of the method.
Article
We propose a novel method, the adaptive local window, for improving level set segmentation technique. The window is estimated separately for each contour point, over iterations of the segmentation process, and for each individual object. Our method considers the object scale, the spatial texture, and changes of the energy functional over iterations. Global and local statistics are considered by calculating several gray level co-occurrence matrices. We demonstrate the capabilities of the method in the domain of medical imaging for segmenting 233 images with liver lesions. To illustrate the strength of our method, those images were obtained by either Computed Tomography or Magnetic Resonance Imaging. Moreover, we analyzed images using three different energy models. We compare our method to a global level set segmentation and to local framework that uses predefined fixed-size square windows. The results indicate that our proposed method outperforms the other methods in terms of agreement with the manual marking and dependence on contour initialization or the energy model used. In case of complex lesions, such as low contrast lesions, heterogeneous lesions, or lesions with a noisy background, our method shows significantly better segmentation with an improvement of 0.25+- 0.13 in Dice similarity coefficient, compared with state of the art fixed-size local windows (Wilcoxon, p < 0.001).
Article
Liver segmentation in CT images is an important step for liver volumetry and vascular evaluation in liver pre-surgical planning. In this paper, a segmentation method based on distance regularized level set evolution(DRLSE) model was proposed, which incorporated a distance regularization term into the conventional Chan-Vese (C-V) model. In addition, the region growing method was utilized to generate the initial liver mask for each slice, which could decrease the computation time for level-set propagation. The experimental results show that the method can dramatically decrease the evolving time and keep the accuracy of segmentation. The new method is averagely 15 times faster than the method based on conventional C-V model in segmenting a slice. © 2012 by Editorial Board of Journal of Donghua University, Shanghai, China.
Article
Nowadays, different methods are being published for the segmentation of the liver but, in general, most of them are not suitable for clinical practice due to several inconveniences as high computational cost, excessive user dependence or low accuracy. The purpose of this paper is to present the performance and validation of a liver segmentation method in computed tomography images (contrast venous phase) where automation, easy user interaction, and low computational cost (besides the required accuracy for clinical purposes) have been taken into account. Firstly, an adaptive filter based on intrinsic parameters of the liver is applied to reduce noise but preserving external liver gradients. In a second step, from a seed or a group of them, voxels with similar intensities are included in an initial 3D mask. Finally, thanks to the combination of morphological operators in different orientations, several non-liver structures (cava vein, ribs, stomach or heart) are removed and the final 3D liver mask is obtained. Thirty public datasets have been used to estimate the accuracy of the proposed algorithm, twenty for training the method and ten for testing it. An average Jaccard index of 0.91 (+/- 0.03), a Hausdorff distance of 26.68 (+/- 10.42) mm, and a runtime of 0.25 seconds per slice, state a promising efficiency and efficacy in the test datasets. To our knowledge, liver segmentation methods in the state of the art are achieving high accuracy at the expense of requiring an exhaustive training stage and so much clinician interaction time in different steps of the process. In this paper, a method based on intensity properties is carried out with a high grade of automatism, an easy user interaction and a low computational cost. The results obtained for different patients state a low variance and a good accuracy in most images, thus the robustness of the method is demonstrated.
Article
Full-text available
To compare differences between volumetric interpolated breath-hold examination (VIBE) using two-point Dixon fat-water separation (Dixon-VIBE) and chemically selective fat saturation (FS-VIBE) with magnetic resonance imaging examination. Forty-nine patients were included, who were scanned with two VIBE sequences (Dixon-VIBE and FS-VIBE) in hepatobiliary phase after gadoxetic acid administration. Subjective evaluations including sharpness of tumor, sharpness of vessels, strength and homogeneity of fat suppression, and artifacts that were scored using a 4-point scale. The liver-to-lesion contrast was also calculated and compared. Dixon-VIBE with water reconstruction had significantly higher subjective scores than FS-VIBE in strength and homogeneity of fat suppression (< 0.0001) but lower scores in sharpness of tumor (P < 0.0001), sharpness of vessels (P = 0.0001), and artifacts (P = 0.034). The liver-to-lesion contrast on Dixon-VIBE images was significantly lower than that on FS-VIBE (16.6% ± 9.4% vs 23.9% ± 12.1%, P = 0.0001). Dixon-VIBE provides stronger and more homogenous fat suppression than FS-VIBE, while has lower clarity of focal liver lesions in hepatobiliary phase after gadoxetic acid administration.
Article
An intensity statistics based graph cut segmentation algorithm is proposed in this paper to improve the accuracy and adaptive capacity of liver segmentation. The proposed segmentation method consists of four steps as follows: First, combined with the Otsu algorithm and associated with a cropped liver image, we defined a gray interval as the livers intensity range. Second, the fuzzy c-means clustering algorithm was applied to compute the average intensity and the variance. Third, we establish the cost function with the statistic results. Finally, we employed the improved graph cut model to extract the liver parenchyma from a large cross-section liver image. Experimental results show that the proposed segmentation method is feasible for different liver images of different intensity statistics.
Conference Paper
Full-text available
Fusion of various images aids the rejuvenation of complementary attributes of the images. Similarly, medical image fusion constructs a composite image comprehending significant traits from multimodal source images. Current work exhibits medical image fusion utilizing Laplacian Pyramid (LP) employing DCT. LP decomposes the source medical images as different low pass filtered images, resembling a pyramidal structure. As the pyramidal level of decomposition increases, the quality of the fused image also increases. The proposed technique provides a fused image with better edges and information content from human visual system (HVS) point of view. Qualitative and quantitative analysis of the proposed technique is found to be superior than that of Daubechies complex wavelet transform (DCxWT).
Conference Paper
Aim of the present work was to evaluate the performance of a novel fully automatic algorithm for 3D segmentation and volumetric reconstruction of liver vessel network from contrast-enhanced computed tomography (CECT) datasets acquired during routine clinical activity. Three anonymized CECT datasets were randomly collected and were automatically analyzed by the new vessel segmentation algorithm, whose parameter configuration had been previously optimized on a phantom model. The same datasets were also manually segmented by an experienced operator that was blind with respect to algorithm outcome. Automatic segmentation accuracy was quantitatively assessed for both single 2D slices and 3D reconstruction of the vessel network, accounting manual segmentation results as the reference “ground truth”. Adopted evaluation framework included the following two groups of calculations: 1) for 3D vessel network, sensitivity in vessel detection was quantified as a function of both vessel diameter and vessel order; 2) for vessel images on 2D slices, dice similarity coefficient (DSC), false positive ratio (FPR), false negative ratio (FNR), Bland-Altman plots and Pearson correlation coefficients were used to judge the correctness of single pixel classifications. Automatic segmentation resulted in a 3D vessel detection sensitivity of 100% for vessels larger than 1 mm in diameter, 64.6% for vessels in the range 0.5-1.0 mm and 27.8% for smaller vessels. An average area overlap of 99.1% was obtained between automatically and manually segmented vessel sections, with an average difference of 0.53 mm2. The corresponding average values of FPR and FNR were 1.8% and 1.6%, respectively. Therefore, the tested method showed significant robustness and accuracy in automatic extraction of the liver vessel tree from CECT datasets. Although further verification studies on larger patient populations are required, the described algorithm has an exciting potential for supportin- liver surgery planning and intraoperative resection guidance.
Conference Paper
Snake algorithm is widely used for edge detection. Snake is a parametric curve based on energy minimization principle which has internal forces which retains its characteristics and external forces which is responsible for pulling it towards lines, edges. The initialization of contours is a critical factor in active contour model. Poorly initialized contour fails to capture edges. Various methods have been developed for automatic initialization of contours. Satellite images generally cover large area and they are large in size. So their computational complexity is high. In this paper, we are estimating external energy from static force using Poisson’s equations and determines most likely initial contour. We have used vector field convolution force for deformation. We have extended Bing Li and Scott T. Acton’s [4] algorithm for satellite images. Comparison is done with Center of divergent method, Force field segmentation method and Canny edge detection method. Poisson inverse gradient method requires minimum number of iteration and time required for snake deformation.
Article
Full-text available
Combinatorial graph cut algorithms have been successfully applied to a wide range of problems in vision and graphics. This paper focusses on possibly the simplest application of graph-cuts: segmentation of objects in image data. Despite its simplicity, this application epitomizes the best features of combinatorial graph cuts methods in vision: global optima, practical efficiency, numerical robustness, ability to fuse a wide range of visual cues and constraints, unrestricted topological properties of segments, and applicability to N-D problems. Graph cuts based approaches to object extraction have also been shown to have interesting connections with earlier segmentation methods such as snakes, geodesic active contours, and level-sets. The segmentation energies optimized by graph cuts combine boundary regularization with region-based properties in the same fashion as Mumford-Shah style functionals. We present motivation and detailed technical description of the basic combinatorial optimization framework for image segmentation via s/t graph cuts. After the general concept of using binary graph cut algorithms for object segmentation was first proposed and tested in Boykov and Jolly (2001), this idea was widely studied in computer vision and graphics communities. We provide links to a large number of known extensions based on iterative parameter re-estimation and learning, multi-scale or hierarchical approaches, narrow bands, and other techniques for demanding photo, video, and medical applications.
Conference Paper
Full-text available
We introduce a new approach to modelling gradient flows of contours and surfaces. While standard variational methods (e.g. level sets) compute local interface motion in a differential fashion by estimating local contour velocity via energy derivatives, we propose to solve surface evolution PDEs by explicitly estimating integral motion of the whole surface. We formulate an optimization problem directly based on an integral characterization of gradient flow as an infinitesimal move of the (whole) surface giving the largest energy decrease among all moves of equal size. We show that this problem can be efficiently solved using recent advances in algorithms for global hypersurface optimization [4,2,11]. In particular, we employ the geo-cuts method [4] that uses ideas from integral geometry to represent continuous surfaces as cuts on discrete graphs. The resulting interface evolution algorithm is validated on some 2D and 3D examples similar to typical demonstrations of level-set methods. Our method can compute gradient flows of hypersurfaces with respect to a fairly general class of continuous functionals and it is flexible with respect to distance metrics on the space of contours/surfaces. Preliminary tests for standard L 2 distance metric demonstrate numerical stability, topological changes and an absence of any oscillatory motion.
Conference Paper
Full-text available
Automatic liver segmentation from abdominal computed tomography (CT) images is one of the most important steps for computer-aided diagnosis (CAD) for liver CT. However, the liver must be separated manually or semi-automatically since surface features of the liver and partial-volume effects make automatic discrimination from other adjacent organs or tissues very difficult. In this paper, we present an unsupervised liver segmentation algorithm with three steps. In the preprocessing, we simplify the input CT image by estimating the liver position using a prior knowledge about the location of the liver and by performing multilevel threshold on the estimated liver position. The proposed scheme utilizes the multiscale morphological filter recursively with region-labeling and clustering to detect the search range for deformable contouring. Most of the liver contours are positioned within the search range. In order to perform an accurate segmentation, we produce the gradient-label map, which represents the gradient magnitude in the search range. The proposed algorithm performed deformable contouring on the gradient-label map by using regular patterns of the liver boundary. Experimental results are comparable to those of manual tracing by radiological doctors and shown to be efficient.
Article
Full-text available
The objective of this work is to evaluate a new concept of intraoperative three-dimensional (3D) visualization system to support hepatectomy. The Resection Map aims to provide accurate cartography for surgeons, who can therefore anticipate risks, increase their confidence and achieve safer liver resection. In an experimental prospective cohort study, ten consecutive patients admitted for hepatectomy to three European hospitals were selected. Liver structures (portal veins, hepatic veins, tumours and parenchyma) were segmented from a recent computed tomography (CT) study of each patient. The surgeon planned the resection preoperatively and read the Resection Map as reference guidance during the procedure. Objective (amount of bleeding, tumour resection margin and operating time) and subjective parameters were retrieved after each case. Three different surgeons operated on seven patients with the navigation aid of the Resection Map. Veins displayed in the Resection Map were identified during the surgical procedure in 70.1% of cases, depending mainly on size. Surgeons were able to track resection progress and experienced improved orientation and increased confidence during the procedure. The Resection Map is a pragmatic solution to enhance the orientation and confidence of the surgeon. Further studies are needed to demonstrate improvement in patient safety.
Article
Full-text available
Liver tumors segmentation is an important prerequisite for planning of surgical interventions. For clinical applicability, the segmentation approach must be able to cope with the high variation in shape and gray-value appearance of the liver. We present a fully automatic 3D segmentation method for the liver tumors from contrast-enhanced CT data. The method consists of two main stages. First an initial histogram and statistical distribution functions are created, and from them a new image is created where, in each voxel, a weighted function is attached in accordance with the probability of the voxel grey level. Next, we use the active contour method on the new image, where the active contour evolution is based upon the minimization of variances between the liver tumor and its closest neighborhood.
Article
Full-text available
Benign hepatic and biliary tumors can present a difficult diagnostic challenge. Spiral computed tomography (CT) and magnetic resonance (MR) imaging are useful in the detection and characterization of these tumors. Imaging characteristics of lesions such as hepatic cyst, hemangioma, focal nodular hyperplasia (FNH), and hepatic adenoma are well known. Hepatic cysts demonstrate water attenuation at CT, are isointense relative to water at MR imaging, and do not enhance after intravenous administration of contrast material. Hemangiomas demonstrate characteristic nodular peripheral enhancement on early-phase images with subsequent fill-in centrally at both modalities. FNH classically demonstrates intense early enhancement with washout on delayed images. Although hepatic adenoma can also demonstrate intense early enhancement, it has a tendency to bleed and thus often appears more heterogeneous than FNH due to hemorrhage. Benign hepatic tumors that are less well described in the imaging literature include hepatic lipoma or angiomyolipoma, infantile hemangioendothelioma, and mesenchymal hamartoma. Hepatic lipoma has fat attenuation at CT, is isointense relative to fat at MR imaging, and does not enhance after intravenous administration of contrast material. Hepatic angiomyolipomas contain a variable amount of soft tissue in addition to fat and may therefore demonstrate enhancement at both modalities. The CT and MR imaging appearances of infantile hemangioma are similar to those of adult hemangioma. Infantile hemangioendothelioma occurs in infants under 6 months of age and is typically a larger lesion. Mesenchymal hamartoma also occurs in children, and its imaging appearance depends on the presence of stromal elements and the protein content of the cyst fluid. Familiarity with these imaging features can help distinguish particular disease entities.
Article
Full-text available
This work presents first quantitative results of a method for automatic liver segmentation from CT data. It is based on a 3D deformable model approach using a-priori statistical information about the shape of the liver gained from a training set. The model is adapted to the data in an iterative process by analysis of the grey value profiles along its surface normals after nonlinear diffusion filtering. Leave-one-out experiments over 26 CT data sets reveal an accuracy of 2.4 mm with respect to the manual segmentation.
Article
Full-text available
The purpose of this paper is to describe a framework for evaluating image segmentation algorithms. Image segmentation consists of object recognition and delineation. For evaluating segmentation methods, three factors-precision (reliability), accuracy (validity), and efficiency (viability)-need to be considered for both recognition and delineation. To assess precision, we need to choose a figure of merit, repeat segmentation considering all sources of variation, and determine variations in figure of merit via statistical analysis. It is impossible usually to establish true segmentation. Hence, to assess accuracy, we need to choose a surrogate of true segmentation and proceed as for precision. In determining accuracy, it may be important to consider different 'landmark' areas of the structure to be segmented depending on the application. To assess efficiency, both the computational and the user time required for algorithm training and for algorithm execution should be measured and analyzed. Precision, accuracy, and efficiency factors have an influence on one another. It is difficult to improve one factor without affecting others. Segmentation methods must be compared based on all three factors, as illustrated in an example wherein two methods are compared in a particular application domain. The weight given to each factor depends on application.
Article
Full-text available
Accurate knowledge of the liver structure, including liver surface and lesion localization, is usually required in treatments such as liver tumor ablations and/or radiotherapy. This paper presents a new method and corresponding algorithm for fast segmentation of the liver and its internal lesions from CT scans. No interaction between the user and analysis system is required for initialization since the algorithm is fully automatic. A statistical model-based approach was created to distinguish hepatic tissue from other abdominal organs. It was combined to an active contour technique using gradient vector flow in order to obtain a smoother and more natural liver surface segmentation. Thereafter, automatic classification was performed to isolate hepatic lesions from liver parenchyma. Twenty-one datasets, presenting different anatomical and pathological situations, have been processed and analyzed. Special focus has been driven to the resulting processing time together with quality assessment. Our method allowed robust and efficient liver and lesion segmentations very close to the ground truth, in a relatively short processing time (average of 11.4 s for a 512 x 512-pixel slice). A volume overlap of 94.2% and an accuracy of 3.7 mm were achieved for liver surface segmentation. Sensitivity and specificity for tumor lesion detection were 82.6% and 87.5%, respectively.
Article
Full-text available
Noninvasive methods are strongly needed to detect and quantify not only tumor growth in murine tumor models but also the development of vascularization and necrosis within tumors. This study investigates the use of a new imaging technique, flat-panel detector volume computed tomography (fpVCT), to monitor in vivo tumor progression and structural changes within tumors of two murine carcinoma models. After tumor cell inoculation, single fpVCT scans of the entire mice were performed at different time points. The acquired isotropic, high-resolution volume data sets enable an accurate real-time assessment and precise measurements of tumor volumes. Spreading of contrast agent-containing blood vessels around and within the tumors was clearly visible over time. Furthermore, fpVCT permits the identification of differences in the uptake of contrast media within tumors, thus delineating necrosis, tumor tissues, and blood vessels. Classification of tumor tissues based on the decomposition of the underlying mixture distribution of tissue-related Hounsfield units allowed the quantitative acquisition of necrotic tissues at each time point. Morphologic alterations of the tumor depicted by fpVCT were confirmed by histopathologic examination. Concluding, our data show that fpVCT may be highly suitable for the noninvasive evaluation of tumor responses to anticancer therapies during the course of the disease.
Article
Full-text available
Snakes, or active contours, are used extensively in computer vision and image processing applications, particularly to locate object boundaries. Problems associated with initialization and poor convergence to boundary concavities, however, have limited their utility. This paper presents a new external force for active contours, largely solving both problems. This external force, which we call gradient vector flow (GVF), is computed as a diffusion of the gradient vectors of a gray-level or binary edge map derived from the image. It differs fundamentally from traditional snake external forces in that it cannot be written as the negative gradient of a potential function, and the corresponding snake is formulated directly from a force balance condition rather than a variational formulation. Using several two-dimensional (2-D) examples and one three-dimensional (3-D) example, we show that GVF has a large capture range and is able to move snakes into boundary concavities
Article
Full-text available
The aim of this project is computer-based planning of a partial organ resection for three dimensional visualisation of anatomical structures and volumetric estimations. Using quantitative parameters for the estimation of the operation risk, the selection of patients can be further objectified. Another important effect of this project is support for surgical interventions by predefinition of different operation scenarios. All preoperative analysis results are based on tomographic images acquired during the clinical routine. In the end, the integration of planning results into the operation theater as well as into the clinical workflow is part of this project. Surgical strategy is determined by the exact location of the tumor, the respective donor liver part, and the proximity of supporting and draining vessels. Depending on the acquired images, the portal, hepatic, arterial and biliary systems may be analyzed. Therefore, fundamentals for the qualitative and quantitative analysis of images have been developed and implemented since the beginning of the project. The qualitative aspects are, e.g., the three dimensional visualisation of vessel structures, tumors, the security margin and the organ itself. These were evaluated for use in the clinical routine. Preoperatively, the operation strategy is planned by segmenting the important anatomical structures and defining the different vessel systems using a special operation planning software. The visualisation and volumetric results can be visualised and adapted during surgical intervention with the help of a beamer and a touchscreen monitor installed above the patient. In living donor liver transplantation (LDLT) the most important factor is to ensure that the anatomy of the donor patient allows surgical splitting of the li...
Article
The aim of this project is computer-based planning of a partial organ resection for three dimensional visualisation of anatomical structures and volumetric estimations. Using quantitative parameters for the estimation of the operation risk, the selection of patients can be further objectified. Another important effect of this project is support for surgical interventions by predefinition of different operation scenarios. All preoperative analysis results are based on tomographic images acquired during the clinical routine. In the end, the integration of planning results into the operation theater as well as into the clinical workflow is part of this project. Surgical strategy is determined by the exact location of the tumor, the respective donor liver part, and the proximity of supporting and draining vessels. Depending on the acquired images, the portal, hepatic, arterial and biliary systems may be analyzed. Therefore, fundamentals for the qualitative and quantitative analysis of images have been developed and implemented since the beginning of the project. The qualitative aspects are, e.g., the three dimensional visualisation of vessel structures, tumors, the security margin and the organ itself. These were evaluated for use in the clinical routine. Preoperatively, the operation strategy is planned by segmenting the important anatomical structures and defining the different vessel systems using a special operation planning software. The visualisation and volumetric results can be visualised and adapted during surgical intervention with the help of a beamer and a touchscreen monitor installed above the patient. In living donor liver transplantation (LDLT) the most important factor is to ensure that the anatomy of the donor patient allows surgical splitting of the liver. Also, a volumetric analysis of the donor liver is of interest to achieving less complicated surgery. Actually, all LDLTs performed in our Surgical Clinic are supported by this operation planning system.
Article
Computer-aided surgery requires the accurate registration of tomographic pre-operative images with intra-operative surface points acquired with 3D spatial localizers. Surface-based registration of these tomographic images with the surface points does, in turn, require a precise and accurate surface representation of the structures to be registered. This paper presents a level set technique for the automatic segmentation of the liver in abdominal CT scans. The main difficulty with level set methods is the design of appropriate speed functions for particular applications. Here we propose a novel speed function that is designed to (1) stop the propagating front at organ boundaries with weak edges, and (2) incorporate a-priori information on the relative position of the liver and other structures. We show that the new speed function we have designed to stop the front a weak edges is superior to other approaches proposed in the literature both on simulated images and on CT images. Results obtained with our approach for the segmentation of the liver in several CT scans are compared to contours obtained manually.
Article
Noninvasive methods are strongly needed to detect and quantify not only tumor growth in murine tumor models but also the development of vascularization and necrosis within tumors. This study investigates the use of a new imaging technique, flat-panel detector volume computed tomography (fpVCT), to monitor in vivo tumor progres- sion and structural changes within tumors of two murine carcinoma models. After tumor cell inoculation, single fpVCT scans of the entire mice were performed at different time points. The acquired isotropic, high-resolution volume data sets enable an accurate real-time assessment and precise measurements of tumor volumes. Spread- ing of contrast agent-containing blood vessels around and within the tumors was clearly visible over time. Further- more, fpVCT permits the identification of differences in the uptake of contrast media within tumors, thus delineating necrosis, tumor tissues, and blood vessels. Classification of tumor tissues based on the decomposition of the underlying mixture distribution of tissue-related Hounsfield units allowed the quantitative acquisition of ne- crotic tissues at each time point. Morphologic alterations of the tumor depicted by fpVCT were confirmed by his- topathologic examination. Concluding, our data show that fpVCT may be highly suitable for the noninvasive evaluation of tumor responses to anticancer therapies during the course of the disease.
Article
Given the high concentration of blood vessels in liver, and the speckle noise inherent in ultrasound imagery, segmenting liver tumors in ultrasound images is a hard problem. Snakes, or active contours, have had limited success in such noisy and complex images. By coupling the smoothness of the edge map to the initial size of the snake, enhancing the tumor boundaries, tuning the regulating parameters of the snake, and an improved numerical approximation of a vector diffusion equation of the gradient vector field, we have obtained good results on 12 ultrasound images of liver tumors of various sizes.
Article
Objective: To improve the planning of hepatic surgery, we have developed a fully automatic anatomical, pathological, and functional segmentation of the liver derived from a spiral CT scan. Materials and Methods: From a 2 mm-thick enhanced spiral CT scan, the first stage automatically delineates skin, bones, lungs, kidneys, and spleen by combining the use of thresholding, mathematical morphology, and distance maps. Next, a reference 3D model is immersed in the image and automatically deformed to the liver contours. Then an automatic Gaussian fitting on the imaging histogram estimates the intensities of parenchyma, vessels, and lesions. This first result is next improved through an original topological and geometrical analysis, providing an automatic delineation of lesions and veins. Finally, a topological and geometrical analysis based on medical knowledge provides hepatic functional information that is invisible in medical imaging: portal vein labeling and hepatic anatomical segmentation according to the Couinaud classification. Results: Clinical validation performed on more than 30 patients shows that delineation of anatomical structures by this method is often more sensitive and more specific than manual delineation by a radiologist. Conclusion: This study describes the methodology used to create the automatic segmentation of the liver with delineation of important anatomical, pathological, and functional structures from a routine CT scan. Using the methods proposed in this study, we have confirmed the accuracy and utility of the creation of a 3D liver model compared with the conventional reading of the CT scan by a radiologist. This work may allow improved preoperative planning of hepatic surgery by more precisely delineating liver pathology and its relationship to normal hepatic structures. In the future, this data may be integrated with computer-assisted surgery and thus represents a first step towards the development of an augmented-reality surgical system.
Conference Paper
Liver segmentation on computed tomography (CT) images is a challenging task due to the anatomic complexity and the imaging system noises. In this paper a complex algorithm based on active contour is proposed to automatically extract the liver region in abdominal CT images. Combined with threshold segmentation, morphology image processing and active contour models, we can automatically extract the initial contour and segment the liver slice by slice, Experimental results show that the proposed method gives automatic and accurate liver structure segmentation.
Conference Paper
As a mean for liver investigation, abdominal CT images have been widely studied in the recent years. Processing CT image includes the automatic diagnosis of liver pathologies, such as detecting lesions and following vessels ramification, and its 3D volume rendering. The first step in all these studies is the automatic liver segmentation. This paper presents a fully automatic method to segment the liver from abdominal CT data with no interaction from user. A statistical model-based approach is used to distinguish roughly liver tissue from other abdominal organs. It is followed by applying force-driven optimized active contour (snake) in order to obtain a smoother and finer liver contour. The new segmentation technique has been evaluated on fifteen datasets, by comparing the automatically detected liver contour to the liver boundaries manually traced by an expert. Tests are reported on 15 datasets and promising result shows that sensitivity and specificity for automatic liver segmentation are 95% and 99% respectively.
Article
Computed tomography (CT) images have been widely used for diagnosis of liver disease and volume measurement for liver surgery or transplantation. Automatic liver segmentation and volume measurement based on the segmentation are the most essential parts in computer-aided diagnosis for liver CT as well as computer-aided surgery. However, liver segmentation, in general, has been performed by outlining the medical image manually or segmenting CT images semi-automatically because surface features of the liver and partial-volume effects make automatic discrimination from other adjacent organs or tissues very difficult. Accordingly, in this paper, we propose a new approach to automatic segmentation of the liver for volume measurement in sequential CT images. Our method analyzes the intensity distribution of several abdominal CT samples and exploits a priori knowledge, such as CT numbers and location of the liver to identify coherent regions that correspond to the liver. The proposed scheme utilizes recursively morphological filter with region-labeling and clustering to detect the search range and to generate the initial liver contour. In this search range, we deform liver contour using the labeling-based search algorithm following pattern features of the liver contour. Lastly, volume measurement is automatically performed on the segmented liver regions. The experimental measurement of area and volume is compared with those using manual tracing method as a gold standard by the radiological doctors, and demonstrates that this algorithm is effective for automatic segmentation and volume measurement method of the liver.
Article
SUMMARY With the rapid progress of CT (Computed Tomogra- phy) scanners, four-phase CT images with resolutions as high as 1 mm have started to be used for diagnosing liver diseases. The first-, second-, third-, and fourth-phase CT images correspond to before dye injection, the early stage, the full stage, and the wash-out stage of the injected dye. Such CT data offer useful information for diagnosing he- patic cancer. This paper describes an automatic method for segmenting the liver region using the blood vessel stream in the first- and third-phase CTs by tracing the portal vein and then the hepatic vein. First, the tracing of the veins is carried out by performing a 3D labeling operation in the region extracted with a threshold that separates blood ves- sels from liver soft tissue. The adjoining stomach and spleen regions are separated by applying erosion and dilation, which are 3D morphological operations. Then, an operation for determining the accurate liver region is carried out to derive an approximate liver region, by enlarging the liver blood vessel region by morphological dilation and then extracting the liver region with an optimal threshold. This method was applied to eight CT data. The resulting regions agree well with the manually detected regions. © 2004
Article
An improved and general approach to connected-component labeling of images is presented. The algorithm presented in this paper processes images in predetermined order, which means that the processing order depends only on the image representation scheme and not on specific properties of the image. The algorithm handles a wide variety of image representation schemes (rasters, run lengths, quadrees, bintrees, etc.). How to adapt the standard UNION-FIND algorithm to permit reuse of temporary labels is shown. This is done using a technique called age balancing, in which, when two labels are merged, the older label becomes the father of the younger label. This technique can be made to coexist with the more conventional rule of weight balancing, in which the label with more descendants becomes the father of the label with fewer descendants. Various image scanning orders are examined and classified. It is also shown that when the algorithm is specialized to a pixel array scanned in raster order, the total processing time is linear in the number of pixels. The linear-time processing time follows from a special property of the UNION-FIND algorithm, which may be of independent interest. This property states that under certain restrictions on the input, UNION-FIND runs in time linear in the number of FIND and UNION operations. Under these restrictions, linear-time performance can be achieved without resorting to the more complicated Gabow-Tarjan algorithm for disjoint set union.
Article
The determination of sample size is a common task for many organizational researchers. Inappropriate, inadequate, or excessive sample sizes continue to influence the quality and accuracy of research. The procedures for determining sample size for continuous and categorical variables using Cochran's (1977) formulas are described. A discussion and illustration of sample size formulas, including the formula for adjusting the sample size for smaller populations, is included
Article
Liver cancer is one of the major death factors in the world. Transplantation and tumor removal are two main therapies in common clinical practice. Both tasks need image assisted planning and quantitative evaluations. Automatic liver segmentation is required for corresponding quantitative evaluations. Conventional approaches in liver segmentation consist of finding the initial liver border followed by tuning the border to the final mask. Finding the liver initial border is of great importance as the latter step largely depends on the initial step. In the previous works, the liver initial border was determined by applying thresholding and morphological filters. In order to estimate the liver initial boundary, we have proposed a technique based on anatomical knowledge of liver, its surrounding tissues as well as the approach that a clinician follows in screening liver in a CT dataset. Based on the above reasoning, we developed a multi-step heuristic technique to segment liver from other tissues in multi-slice CT images. The proposed technique can deal with various shapes, locations, and liver sizes. The method was evaluated in the presence of 50 actual liver data sets and the results were encouraging.
Article
This paper presents a comparison study between 10 automatic and six interactive methods for liver segmentation from contrast-enhanced CT images. It is based on results from the "MICCAI 2007 Grand Challenge" workshop, where 16 teams evaluated their algorithms on a common database. A collection of 20 clinical images with reference segmentations was provided to train and tune algorithms in advance. Participants were also allowed to use additional proprietary training data for that purpose. All teams then had to apply their methods to 10 test datasets and submit the obtained results. Employed algorithms include statistical shape models, atlas registration, level-sets, graph-cuts and rule-based systems. All results were compared to reference segmentations five error measures that highlight different aspects of segmentation accuracy. All measures were combined according to a specific scoring system relating the obtained values to human expert variability. In general, interactive methods reached higher average scores than automatic approaches and featured a better consistency of segmentation quality. However, the best automatic methods (mainly based on statistical shape models with some additional free deformation) could compete well on the majority of test images. The study provides an insight in performance of different segmentation approaches under real-world conditions and highlights achievements and limitations of current image analysis techniques.
Article
Tumor volume is an important parameter for clinical decision making. At present, semiautomatic image segmentation is not a standard for tumor volumetry. The aim of this work was to investigate the usability of semiautomatic algorithms for tumor volume determination. Semiautomatic region- and volume-growing, isocontour, snakes, hierarchical, and histogram-based segmentation algorithms were tested for accuracy, contour variability, and time performance. The test were performed on a newly developed organic phantom for the simulation of a human liver and liver metastases. The real tumor volumes were measured by water displacement. These measured volumes were used as the gold standard for determining the accuracy of the algorithms. Variability of the segmented volumes ranging from 3.9 +/- 3.2% (isocontour algorithm) to 11.5 +/- 13.9% (hierarchical segmentation) was observed. The segmentation time per slice varied between 32 (volume-growing) and 72 seconds (snakes) on an IBM/RS6000 workstation. Only the region-growing and isocontour algorithms have the potential to be used for tumor volumetry. However, further improvements of these algorithms are necessary before they can be placed into clinical use.
Article
To compare the accuracy and repeatability of a semiautomatic segmentation algorithm with those of manual segmentation for determining liver volume in living liver transplant donors at magnetic resonance (MR) imaging. The institutional review board approved this retrospective study and waived the requirement for informed consent. The semiautomatic segmentation algorithm is based on geometric deformable models and the level-set technique. It entails (a) placing initialization circle(s) on each image section, (b) running the algorithm, (c) inspecting and possibly manually modifying the contours obtained with the segmentation algorithm, and (d) placing lines to separate the liver segments. For 18 living donors (eight men and 10 women; mean age, 34 years; age range, 25-46 years), two observers each performed two semiautomatic and two manual segmentations on contrast material-enhanced T1-weighted MR images. Each measurement was timed. Actual graft weight was measured during surgery. The time needed for manual and that needed for semiautomatic segmentation were compared. Accuracy and repeatability were evaluated with the Bland-Altman method. Mean interaction time was reduced from 25 minutes with manual segmentation to 5 minutes with semiautomatic segmentation. The mean total time for the semiautomatic process was 7 minutes 20 seconds. Differences between the actual volume and the estimated volume ranged from -223 to +123 mL for manual segmentation and from -214 to +86 mL for semiautomatic segmentation. The 95% limits of agreement for the ratio of actual graft volume to estimated graft volume were 0.686 and 1.601 for semiautomatic segmentation and 0.651 and 1.957 for manual segmentation. Semiautomatic segmentation improved estimation in 15 of 18 cases. Inter- and intraobserver repeatability was higher with semiautomatic segmentation. Use of the semiautomatic segmentation algorithm substantially reduces the time needed for volumetric measurement of liver segments while improving both accuracy and repeatability.
Article
After [15], [31], [19], [8], [25], [5], minimum cut/maximum flow algorithms on graphs emerged as an increasingly useful tool for exact or approximate energy minimization in low-level vision. The combinatorial optimization literature provides many min-cut/max-flow algorithms with different polynomial time complexity. Their practical efficiency, however, has to date been studied mainly outside the scope of computer vision. The goal of this paper is to provide an experimental comparison of the efficiency of min-cut/max flow algorithms for applications in vision. We compare the running times of several standard algorithms, as well as a new algorithm that we have recently developed. The algorithms we study include both Goldberg-Tarjan style "push-relabel" methods and algorithms based on Ford-Fulkerson style "augmenting paths." We benchmark these algorithms on a number of typical graphs in the contexts of image restoration, stereo, and segmentation. In many cases, our new algorithm works several times faster than any of the other methods, making near real-time performance possible. An implementation of our max-flow/min-cut algorithm is available upon request for research purposes.
Article
Evaluation of the use of planning CTs with slow revolution time (4s/slice; 'slow CTs') in the planning procedure of radiotherapy for lung cancer patients in comparison to commonly used 'fast' planning CTs; impact on margin assessment of planning target volume (PTV) design. Eighteen lung cancer patients (six upper, lower lobe and central tumors, respectively) have been scanned each by three series of slow CTs and one fast CT (spiral CT). Patients have been freely breathing. The largest transversal tumor diameters in slow and fast CTs have been measured. Tumor edge positions have been determined for six spatial directions in all slices of the three slow CT series. Slow CTs show larger dimensions of the visible tumor than fast CTs. The median difference of the diameter for all tumors is 2mm (range 0-4mm). Slow CTs deliver constant depictions of lung tumors within a range of 1.6mm in all directions. This margin is considered to be sufficient to compensate for tumor movements by respiration and cardiovascular motions (internal margin). A margin of 7 mm added to the GTV of a single slow CT series to draw the PTV is proposed. Slow planning CTs show larger, but highly constant depictions of lung tumors in comparison to conventional fast CT scanning, yielding an integral delineation of almost all positions of the moving tumors. Thus the use of slow planning CTs enables the drawing of tighter margins in external beam treatment planning of lung cancer.
Article
This study was performed to design an automatic liver region extraction system to facilitate clinical liver size estimation and further serve as a prestage for liver reconstruction and volume estimation. We present a modification of the well-known snakes algorithm for extracting liver regions in noisy CT images. Our modification addresses the issues of selection of the control points on an estimate of the contour and the determination of the weighting coefficients. The weighting coefficients are determined dynamically on the basis of the distance between the control points and the local curvature of the contour. The proposed method was used in extracting liver regions from 98 cross-sectional abdominal images. The overall performance was estimated by comparisons with original liver regions. The deformable model method enables an efficient and effective automatic liver region extraction in noisy environments. This approach eliminates human-in-the loop, which is the common practice for the majority of current methods.
Article
Accurate liver segmentation on computed tomography (CT) images is a challenging task especially at sites where surrounding tissues (e.g., stomach, kidney) have densities similar to that of the liver and lesions reside at the liver edges. We have developed a method for semiautomatic delineation of the liver contours on contrast-enhanced CT images. The method utilizes a snake algorithm with a gradient vector flow (GVF) field as its external force. To improve the performance of the GVF snake in the segmentation of the liver contour, an edge map was obtained with a Canny edge detector, followed by modifications using a liver template and a concavity removal algorithm. With the modified edge map, for which unwanted edges inside the liver were eliminated, the GVF field was computed and an initial liver contour was formed. The snake algorithm was then applied to obtain the actual liver contour. This algorithm was extended to segment the liver volume in a slice-by-slice fashion, where the result of the preceding slice constrained the segmentation of the adjacent slice. 551 two-dimensional liver images from 20 volumetric images with colorectal metastases spreading throughout the livers were delineated using this method, and also manually by a radiologist for evaluation. The difference ratio, which is defined as the percentage ratio of mismatching volume between the computer and the radiologist's results, ranged from 2.9% to 7.6% with a median value of 5.3%.
Article
Active contour segmentation and its robust implementation using level set methods are well-established theoretical approaches that have been studied thoroughly in the image analysis literature. Despite the existence of these powerful segmentation methods, the needs of clinical research continue to be fulfilled, to a large extent, using slice-by-slice manual tracing. To bridge the gap between methodological advances and clinical routine, we developed an open source application called ITK-SNAP, which is intended to make level set segmentation easily accessible to a wide range of users, including those with little or no mathematical expertise. This paper describes the methods and software engineering philosophy behind this new tool and provides the results of validation experiments performed in the context of an ongoing child autism neuroimaging study. The validation establishes SNAP intrarater and interrater reliability and overlap error statistics for the caudate nucleus and finds that SNAP is a highly reliable and efficient alternative to manual tracing. Analogous results for lateral ventricle segmentation are provided.
Article
Progress in radiotherapy is guided by the need to realize improved dose distributions, i.e. the ability to reduce the treatment volume toward the target volume and still ensuring coverage of that target volume in all dimensions. Poor ability to control the tumour's location limits the accuracy with which radiation can be delivered to tumour-bearing tissue. Image-guided radiation therapy (IGRT) aims at in-room imaging guiding the radiation delivery based on instant knowledge of the target location and changes in tumour volume during treatment. Advancements are usually not to be attributed to a single event, but rather a combination of many small improvements that together enable a superior result. Image-guidance is an important link in the treatment chain and as such a major factor in this synergetic process. A historic review shows that many of the so-called new developments are not so new at all, but did not make it into mainstream radiotherapy practice at that time. Recent developments in improved IT infrastructures, novel irradiation techniques, and better knowledge of functional and morphologic information may have created the need and optimal environment to revive the interest in IGRT.
Conference Paper
Geodesic active contours and graph cuts are two standard image segmentation techniques. We introduce a new segmentation method combining some of their benefits. Our main intuition is that any cut on a graph embedded in some continuous space can be interpreted as a contour (in 2D) or a surface (in 3D). We show how to build a grid graph and set its edge weights so that the cost of cuts is arbitrarily close to the length (area) of the corresponding contours (surfaces) for any anisotropic Riemannian metric. There are two interesting consequences of this technical result. First, graph cut algorithms can be used to find globally minimum geodesic contours (minimal surfaces in 3D) under arbitrary Riemannian metric for a given set of boundary conditions. Second, we show how to minimize metrication artifacts in existing graph-cut based methods in vision. Theoretically speaking, our work provides an interesting link between several branches of mathematics -differential geometry, integral geometry, and combinatorial optimization. The main technical problem is solved using Cauchy-Crofton formula from integral geometry.
Article
A general non-parametric technique is proposed for the analysis of a complex multimodal feature space and to delineate arbitrarily shaped clusters in it. The basic computational module of the technique is an old pattern recognition procedure: the mean shift. For discrete data, we prove the convergence of a recursive mean shift procedure to the nearest stationary point of the underlying density function and, thus, its utility in detecting the modes of the density. The relation of the mean shift procedure to the Nadaraya-Watson estimator from kernel regression and the robust M-estimators; of location is also established. Algorithms for two low-level vision tasks discontinuity-preserving smoothing and image segmentation - are described as applications. In these algorithms, the only user-set parameter is the resolution of the analysis, and either gray-level or color images are accepted as input. Extensive experimental results illustrate their excellent performance
Article
Geodesic active contours and graph cuts are two standard image segmentation techniques. We introduce a new segmentation method combining some of their benefits. Our main intuition is that any cut on a graph embedded in some continuous space can be interpreted as a contour (in 2D) or a surface (in 3D). We show how to build a grid graph and set its edge weights so that the cost of cuts is arbitrarily close to the length (area) of the corresponding contours (surfaces) for any anisotropic Riemannian metric.