Senthil Periaswamy

National Institutes of Health, Bethesda, MD, United States

Are you Senthil Periaswamy?

Claim your profile

Publications (31)25.76 Total impact

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we propose a new registration method for prone and supine computed tomographic colonography scans using graph matching. We formulate 3-D colon registration as a graph matching problem and propose a new graph matching algorithm based on mean field theory. In the proposed algorithm, we solve the matching problem in an iterative way. In each step, we use mean field theory to find the matched pair of nodes with highest probability. During iterative optimization, one-to-one matching constraints are added to the system in a step-by-step approach. Prominent matching pairs found in previous iterations are used to guide subsequent mean field calculations. The proposed method was found to have the best performance with smallest standard deviation compared with two other baseline algorithms called the normalized distance along the colon centerline (NDACC) ( p = 0.17) with manual colon centerline correction and spectral matching ( p < 1e-5). A major advantage of the proposed method is that it is fully automatic and does not require defining a colon centerline for registration. For the latter NDACC method, user interaction is almost always needed for identifying the colon centerlines.
    IEEE transactions on information technology in biomedicine: a publication of the IEEE Engineering in Medicine and Biology Society 04/2012; 16(4):676-82. · 1.69 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes a registration method for supine and prone CTC scans. The method matches graphs built using the teniae coli, three muscles that run the length of the colon. The teniae are visible on CTC and were detected using fully-automatically software. Then key points of the teniae were obtained by non-uniformed sampling of the teniae. Graphs were built using these key points. The colon registration was formulated as a graph matching problem. Mean field theory was applied to match the graphs. The proposed method was tested on 10 pairs of supine and prone CTC scans. The average registration error was 2.5cm (±0.7 cm, 95% C.I. [2.1 2.9]), significantly improving the baseline graph matching method for CTC registration.
    Proceedings / IEEE International Symposium on Biomedical Imaging: from nano to macro. IEEE International Symposium on Biomedical Imaging 01/2012;
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we propose a new registration method for supine and prone computed tomographic colonography scans based on graph matching. We first formulated 3D colon registration as a graph matching problem and utilized a graph matching algorithm based on mean field theory. During the iterative optimization process, one-to-one matching constraints were added to the system step-by-step. Prominent matching pairs found in previous iterations are used to guide subsequent mean field calculations. The advantage of the proposed method is that it does not require a colon centerline for registration. We tested the algorithm on a CTC dataset of 19 patients with 19 polyps. The average registration error of the proposed method was 4.0cm (std. 2.1cm). The 95% confidence intervals were [3.0cm, 5.0mm]. There was no significant difference between the proposed method and our previous method based on the normalized distance along the colon centerline (p=0.1).
    Proc SPIE 03/2011;
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we propose a new graph matching algorithm based on mean field theory. We first convert the original graph matching problem which is a quadratic integer programming problem to a spin model with quadratic interaction by dropping the matching constraints. Then the matching constraints are added to the system iteratively after each round of mean field calculation. Prominent matching pairs found in previous iterations will guide the mean field calculation in the next round. Experiments on the CMU house dataset and a CTC dataset show promising matching results.
    Image Processing (ICIP), 2010 17th IEEE International Conference on; 10/2010
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To assess the effect of using computer-aided detection (CAD) in second-read mode on readers' accuracy in interpreting computed tomographic (CT) colonographic images. The contributing institutions performed the examinations under approval of their local institutional review board, with waiver of informed consent, for this HIPAA-compliant study. A cohort of 100 colonoscopy-proved cases was used: In 52 patients with findings positive for polyps, 74 polyps of 6 mm or larger were observed in 65 colonic segments; in 48 patients with findings negative for polyps, no polyps were found. Nineteen blinded readers interpreted each case at two different times, with and without the assistance of a commercial CAD system. The effect of CAD was assessed in segment-level and patient-level receiver operating characteristic (ROC) curve analyses. Thirteen (68%) of 19 readers demonstrated higher accuracy with CAD, as measured with the segment-level area under the ROC curve (AUC). The readers' average segment-level AUC with CAD (0.758) was significantly greater (P = .015) than the average AUC in the unassisted read (0.737). Readers' per-segment, per-patient, and per-polyp sensitivity for all polyps of 6 mm or larger was higher (P < .011, .007, .005, respectively) for readings with CAD compared with unassisted readings (0.517 versus 0.465, 0.521 versus 0.466, and 0.477 versus 0.422, respectively). Sensitivity for patients with at least one large polyp of 10 mm or larger was also higher (P < .047) with CAD than without (0.777 versus 0.743). Average reader sensitivity also improved with CAD by more than 0.08 for small adenomas. Use of CAD reduced specificity of readers by 0.025 (P = .05). Use of CAD resulted in a significant improvement in overall reader performance. CAD improves reader sensitivity when measured per segment, per patient, and per polyp for small polyps and adenomas and also reduces specificity by a small amount.
    Radiology 09/2010; 256(3):827-35. · 6.21 Impact Factor
  • RSNA; 01/2010
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In computed tomographic colonography (CTC), a patient will be scanned twice-Once supine and once prone-to improve the sensitivity for polyp detection. To assist radiologists in CTC reading, in this paper we propose an automated method for colon registration from supine and prone CTC scans. We propose a new colon centerline registration method for prone and supine CTC scans using correlation optimized warping (COW) and canonical correlation analysis (CCA) based on the anatomical structure of the colon. Four anatomical salient points on the colon are first automatically distinguished. Then correlation optimized warping is applied to the segments defined by the anatomical landmarks to improve the global registration based on local correlation of segments. The COW method was modified by embedding canonical correlation analysis to allow multiple features along the colon centerline to be used in our implementation. We tested the COW algorithm on a CTC data set of 39 patients with 39 polyps (19 training and 20 test cases) to verify the effectiveness of the proposed COW registration method. Experimental results on the test set show that the COW method significantly reduces the average estimation error in a polyp location between supine and prone scans by 67.6%, from 46.27 +/- 52.97 to 14.98 mm +/- 11.41 mm, compared to the normalized distance along the colon centerline algorithm (p < 0.01). The proposed COW algorithm is more accurate for the colon centerline registration compared to the normalized distance along the colon centerline method and the dynamic time warping method. Comparison results showed that the feature combination of z-coordinate and curvature achieved lowest registration error compared to the other feature combinations used by COW. The proposed method is tolerant to centerline errors because anatomical landmarks help prevent the propagation of errors across the entire colon centerline.
    Medical Physics 12/2009; 36(12):5595-603. · 3.01 Impact Factor
  • Source
    Kazunori Okada, S. Periaswamy, Jinbo Bi
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes a stratified regularity measure: a novel entropic measure to describe data regularity as a function of data domain stratification. Jensen-Shannon divergence is used to compute a set-similarity of intensity distributions derived from stratified data. We prove that derived regularity measures form a continuum as a function of the stratificationpsilas granularity and also upper-bounded by the Shannon entropy. This enables to interpret it as a generalized Shannon entropy with an intuitive spatial parameterization. This measure is applied as a novel feature extraction method for a real-world medical image analysis problem. The proposed measure is employed to describe ground-glass lung nodules whose shape and intensity distribution tend to be more irregular than typical lung nodules. Derived descriptors are then incorporated into a machine learning-based computer-aided detection system. Our ROC experiment resulted in 83% success rate with 5 false positives per patient, demonstrating an advantage of our approach toward solving this clinically significant problem.
    Computer Vision and Pattern Recognition Workshops, 2008. CVPR Workshops 2008. IEEE Computer Society Conference on; 07/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To assess the performance of automatic lung nodule detection algorithm (research prototype) on the data acquired by Imaging Database Resources Initiative of the Lung Imaging Database Consortium (LIDC/IDRI) public/private partnership of the U.S. National Cancer Institute. METHOD AND MATERIALS The data set included 57 thin-slice (1-2mm) IDRI MDCT scans: 16 Philips (Brilliance 40 and 16P, B, C and D convolution kernels(CK)), 41 GE (Lightspeed Ultra, 16 and Power, 'STANDARD' CK) and 31 mixed (25 2mm and six 3mm slice thickness) Siemens cases (Sensation 16 and 64, B30f and B45f CK). The cases were marked per the LIDC process: four radiologists, outlined nodules >3mm and marked locations of the smaller ones. Reads were consolidated after an un-blinded read. Detections were consolidated if the center of the contour provided by one radiologist was located inside the contour marked by another radiologist. Detections <3mm were consolidated based on locations proximity. A confidence rating 1-4 was assigned based on number of radiologists detecting the nodule. 21 of IDRI GE cases were added to a proprietary algorithm development image set that consisted of 223 thin-slice images from sources outside of IDRI. RESULTS The algorithm trained on the development set was tested on the remaining 67 IDRI cases. The sensitivity was estimated for nodules from 3 to 30mm. Sensitivity for IDRI Philips cases was 74.2% (23/31 nodules) with 2.9 false positives per case (FP) at confidence level (CL) =1, 80.8% (21/26) with 3.1 FP at CL=2, 83.33% (20/24) with 3.1 FP at CL=3 and 90% (18/20) with 3.3 FP at CL=4. The sensitivity for IDRI GE cases was 87.1% (27/31) with 3.0 FP at CL=1, 88.89% (24/27) with 3.1 FP at CL=2, 86.36% (19/22) with 3.6 FP at CL=3 and 93.33% (14/15) with 3.8 FP at CL=4. The sensitivity for IDRI Siemens cases was relatively lower while the specificity was higher, due to low CK and thicker slices: 71.1% (32/45) with 1.1 FP at CL=1, 73.81% (31/42) with 1.2 at CL=1.2, 76.92% (30/39) with 1.2 FP at CL=3 and 80.65% (25/31) with 1.4 FP at CL=4. CONCLUSION The sensitivity of CAD algorithm was directly correlated with CL (number of radiologists in the cohort that detected a nodule). CLINICAL RELEVANCE/APPLICATION CAD for lung nodule detection in MDCT
    Radiological Society of North America 2007 Scientific Assembly and Annual Meeting; 11/2007
  • Source
    Ming Yang, S. Periaswamy, Ying Wu
    [Show abstract] [Hide abstract]
    ABSTRACT: Lung nodule detection, especially ground glass opacity (GGO) detection, in helical computed tomography (CT) images is a challenging computer-aided detection (CAD) task due to the enormous variances in nodules' volumes, shapes, appearances, and the structures nearby. Most of the detection algorithms employ some efficient candidate generation (CG) algorithms to spot the suspicious volumes with high sensitivity at the cost of low specificity, e.g. tens even hundreds of false positives per volume. This paper proposes a learning based method to reduce the number of false positives given by CG based on a new general 3D volume shape descriptor. The 3D volume shape descriptor is constructed by concatenating spatial histograms of gradient orientations, which is robust to large variabilities in intensity levels, shapes, and appearances. The proposed method achieves promising performance on a difficult mixture lung nodule dataset with average 81% detection rate and 4.3 false positives per volume.
    Acoustics, Speech and Signal Processing, 2007. ICASSP 2007. IEEE International Conference on; 05/2007
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE T2 W STIR MRI is becoming a popular choice for metastasis screening. Robust segmentation of spinal cord as a reliable reference object is essential for automatic vertebrae segmentation, metastasis detection, multi-modal registration, and curved MPR visualization. The aim of this study was to develop a fully automatic algorithm for spinal cord detection in T2 W STIR images. METHOD AND MATERIALS The preprocessing steps include image intensity inhomogeneity correction, scaling and thresholding. The spinal cord centerline is modeled as a global 4th order polynomial in 3D. The model parameters are estimated using random sample consensus (RANSAC) algorithm with subsequent least squares based fitting refinement. Spinal cord is segmented as a curved 3D cylinder around detected centerline. The detection results in 35 patients (metastasis screening population) coronal WB MRI (1.5 T, MAGNETOM AVANTO, Siemens) images were evaluated against manually segmented ground truth (GT). Data was randomly split into training (20 patients) and test (15) sets. RESULTS The average segmentation accuracy, estimated as ratio of overlapping automaticaly detected (AD) and GT spinal cord volumes to the GT volume, was 91% (on all images) and 87% (test set) with standard deviations(STD) of 14 and 16% respectively. The accuracy of centerline position evaluated as average distance from all GT spinal cord voxels to the AD centerline was 4.4mm (all images) and 4.5mm (test set) with STD of 1.9 and 2.5mm respectively. The presence of collapsed vertebrae and edema (in 1 patient) and multiple vertebrae metastasis (12 patients) did not affect the segmentation accuracy in all cases but one (from the test set), where all vertebrae had severe metastatic changes of similar image intensity as spinal cord with no visible boundary. As a result spinal cord centerline was shifted toward the center of the vertebral body. This had a slight negative effect on the test set statistics. CONCLUSION 3D-polynomial modeling provides a fully automatic rapid and reliable estimate of centerline position and segmentation of the spinal cord. CLINICAL RELEVANCE/APPLICATION Spinal cord examination, metastasis screening, multi-modal registration, curved MPR visualization, automatic vertebrae segmentation
    Radiological Society of North America 2006 Scientific Assembly and Annual Meeting; 11/2006
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To evaluate the effectiveness of a MC algorithm in improving readers’ performance when diagnosing breast lesions on contrast enhanced MR based on morphological and kinetic features. METHOD AND MATERIALS Retrospective analysis was performed on 40 breast MRIs with 75 lesions (n= 28 malignant) with and without MC. These exams included proven lesions, with pathology and/or radiology follow up and varying degrees of motion artifacts. Bilateral dynamic exam was performed; precontrast T1 and T2 images, post contrast, and subtraction images were available for review. Two readout sessions were performed. In each session, all exams were randomly presented to each reader (in either its corrected (MC) or uncorrected (UC) versions). Only one of these versions was shown in each session. Three readers reviewed all lesions as outlined by the BIRADS MR Lexicon. For each lesion, its size, an image quality score, a BIRADS assessment and a confidence score on this assessment were recorded. Statistical analysis was performed and considered to be significant when having a p value < 0.05. RESULTS The MC algorithm corrected a subset of the lesions while for all other lesions it did not impact readers’ performance. Following MC, the readers could more easily assess the shape of a mass as spiculated, 84.4% vs. 68.9%, p=0.014 and determine whether lesions less than 10mm were benign or malignant. This was observed regardless of whether the lesion was mass-like or non-mass-like, 80% vs. 66%, p=0.05. The corrected images also increased the diagnostic accuracy in assessing the morphologic features of a lesion when a Type 2 plateau curve was observed, 81.2% vs. 60.3%, p=0.014. Six of 29 malignant cancers were DCIS. All these lesions presented as non-mass like enhancement (7–20 mm) and were more frequently identified as suspicious on the MC images. CONCLUSION MC images corrected a subset of problematic cases. These cases include sub-centimeter lesions and lesions where the kinetic evaluation is indeterminate. This MC algorithm may be helpful in the subset of MRI studies where both the mass and mass like lesions are difficult to evaluate. CLINICAL RELEVANCE/APPLICATION MC may improve our ability to interpret problematic MR lesions.
    Radiological Society of North America 2006 Scientific Assembly and Annual Meeting; 11/2006
  • Source
    Senthil Periaswamy, Hany Farid
    [Show abstract] [Hide abstract]
    ABSTRACT: We have developed a general-purpose registration algorithm for medical images and volumes. The transformation between images is modeled as locally affine but globally smooth, and explicitly accounts for local and global variations in image intensities. An explicit model of missing data is also incorporated, allowing us to simultaneously segment and register images with partial or missing data. The algorithm is built upon a differential multiscale framework and incorporates the expectation maximization algorithm. We show that this approach is highly effective in registering a range of synthetic and clinical medical images.
    Medical Image Analysis 07/2006; 10(3):452-64. · 3.68 Impact Factor
  • Source
    Senthil Periaswamy, Yangming Ou, Anna K. Jerebko
    [Show abstract] [Hide abstract]
    ABSTRACT: A method of extracting a spinal cord from a digitized medical image includes providing a digitized medical image, selecting a set of points from said image as candidates for belonging to the spine, initializing a probability for each candidate point to belong to said spine, minimizing a weighted sum of square differences of image intensities of said candidate points and intensities determined by a mathematical model of said spine to estimate parameter values for said model, calculating a residual error for each point from the differences at each point between an estimated image intensity calculated from said estimated model parameters and an actual image intensity, updating the candidate point probabilities from said residual errors, and eliminating candidate points whose probability falls below a predetermined value.
    Year: 07/2006
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: A novel approach for generating a set of features derived from properties of patterns of curvature is introduced as a part of a computer aided colonic polyp detection system. The resulting sensitivity was 84% with 4.8 false positives per volume on an independent test set of 72 patients (56 polyps). When used in conjunction with other features, it allowed the detection system to reach an overall sensitivity of 94% with a false positive rate of 4.3 per volume.
    Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention. 02/2006; 9(Pt 2):169-76.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper describes a novel classification method for com- puter aided detection (CAD) that identifies structures of in- terest from medical images. CAD problems are challenging largely due to the following three characteristics. Typical CAD training data sets are large and extremely unbalanced between positive and negative classes. When searching for descriptive features, researchers often deploy a large set of experimental features, which consequently introduces irrel- evant and redundant features. Finally, a CAD system has to satisfy stringent real-time requirements. This work is distinguished by three key contributions. The first is a cascade classification approach which is able to tackle all the above difficulties in a unified framework by employing an asymmetric cascade of sparse classifiers each trained to achieve high detection sensitivity and satisfac- tory false positive rates. The second is the incorporation of feature computational costs in a linear program formu- lation that allows the feature selection process to take into account different evaluation costs of various features. The third is a boosting algorithm derived from column genera- tion optimization to effectively solve the proposed cascade linear programs. We apply the proposed approach to the problem of detect- ing lung nodules from helical multi-slice CT images. Our approach demonstrates superior performance in comparison against support vector machines, linear discriminant analy- sis and cascade AdaBoost. Especially, the resulting detec- tion system is significantly sped up with our approach.
    Proceedings of the Twelfth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Philadelphia, PA, USA, August 20-23, 2006; 01/2006
  • [Show abstract] [Hide abstract]
    ABSTRACT: Early detection of lung nodules is an important clinical indication for obtaining routine CT studies of the thorax. To date, research has mainly focused on the sensitivity of CAD compared with expert chest radiologists using data obtained from single or multi-detector CT scanners. But, beside sensitivity and specificity it is also import to know how well the CAD system does perform on datasets from different sites with different slice thicknesses and differences in the dosage. The present study focuses on the generalization ability of a prototype CAD system that is not yet commercially available. It describes the architecture of a recent CAD system and assesses the performance on a heterogeneous dataset collected from multiple geographically diverse sites.
    International Congress Series 05/2005; 1281:1104-1108.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Colorectal cancer is the third most common cancer in both men and women. It is estimated that in 2004, nearly 147,000 cases of colon and rectal cancer will be diagnosed in the USA, and approximately 57,000 people would die from the disease; however, only 44% of the eligible population undergoes any type of colorectal cancer screening. Many reasons have been identified for non-compliance, with key ones being patient comfort, bowel preparation and cost. Virtual colonoscopy derived from computed tomography (CT) images is gaining broader acceptance as a screening method for colorectal neoplasia. Our research suggests that computer-aided detection (CAD) as a second reader has great potential in improving polyp detection. The ColonCAD prototype presented in this paper was developed and tested on cases representative of the variability and quality in true clinical practice. Results of this study with 150 patients demonstrate that: the developed algorithm generalises well: the sensitivity for polyps > or = 6 mm is on average 90%; and the median false positive rate is a manageable 3 per volume.
    British Journal of Radiology 02/2005; 78 Spec No 1:S57-62. · 1.53 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Colorectal cancer is the third most common cancer in both men and women and it was estimated that in 2003 nearly 150,000 cases would be diagnosed and 57,000 people would die. Screening has been accepted as a means for early detection of the disease, yet only a portion of the eligible population undergoes colorectal cancer screening, partially due to comfort issues of undergoing a full colonoscopic examination. Virtual colonoscopy (VC) has been demonstrated to be an effective means of performing screening and given the number of people who are candidate for screening better tools, such as Computer Aided Detection (CAD), are required to fulfill this increasing need. The developed CAD system, presented in this paper, has focused on the detection of polyps of sizes up to and including 20 mm. The results have demonstrated that: the developed algorithm generalizes, the sensitivity and specificity for middle- to large-sized polyps is on the average 95% while the overall sensitivity is roughly 88% and the false positive has remained at a manageable 4 per volume.
    CARS 2004. Computer Assisted Radiology and Surgery. Proceedings of the 18th International Congress and Exhibition, Chicago, USA, June 23-26, 2004; 06/2004
  • Source
    Senthil Periaswamy, Hany Farid
    [Show abstract] [Hide abstract]
    ABSTRACT: We have developed a general-purpose registration algorithm for medical images and volumes. This method models the transformation between images as locally affine but globally smooth. The model also explicitly accounts for local and global variations in image intensities. This approach is built upon a differential multiscale framework, allowing us to capture both large- and small-scale transformations. We show that this approach is highly effective across a broad range of synthetic and clinical medical images.
    IEEE Transactions on Medical Imaging 08/2003; 22(7):865-74. · 3.80 Impact Factor

Publication Stats

411 Citations
25.76 Total Impact Points

Institutions

  • 2009–2012
    • National Institutes of Health
      • Radiology and Imaging Sciences Department
      Bethesda, MD, United States
  • 1999–2003
    • Dartmouth College
      • Department of Computer Science
      Hanover, NH, United States
  • 1998
    • Dartmouth–Hitchcock Medical Center
      Lebanon, New Hampshire, United States