Siping Chen

Shenzhen University, Bao'an, Guangdong, China

Are you Siping Chen?

Claim your profile

Publications (100)96.45 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: Fascicle orientation is one of the most widely used parameters for quantifying muscle function in mechanical analysis, clinical diagnosis, and rehabilitation assessment. Ultrasonography has frequently been used as a reliable way to measure the changes in fascicle orientation of human muscles non-invasively. Conventionally, most such measurements are conducted by a manual analysis of ultrasound images. This manual approach is time consuming, subjective and not suitable for measuring dynamic changes. In this study, we developed an automated tracking method based on a frequency domain Radon transform. The goal of the study was to evaluate the performance of the proposed method by comparing it with the manual approach and by determining its repeatability. A real-time B-mode ultrasound scanner was used to examine the medial gastrocnemius muscle during contraction. The coefficient of multiple correlation (CMC) was used to quantify the level of agreement between the two methods and the repeatability of the proposed method. The two methods were also compared by linear regression and a Bland–Altman analysis. The findings indicated that the results obtained using the proposed method were in good agreement with those obtained using the manual approach (CMC = 0.94 ± 0.03, difference = −0.23 ± 0.68°) and were highly repeatable (CMC = 0.91 ± 0.04). In conclusion, the new method presented here may provide an accurate, highly repeatable, and efficient approach for estimating fascicle orientation during muscle contraction.
    Biomedical Signal Processing and Control 07/2015; 20. DOI:10.1016/j.bspc.2015.04.016 · 1.53 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, a multi-scale convolutional network (MSCN) and graph partitioning based method is proposed for accurate segmentation of cervical cytoplasm and nuclei. Specifically, deep learning via MSCN is explored to extract scale invariant features and then segment regions centered at each pixel. The coarse segmentation is refined by an automated graph partitioning method based on the pre-trained feature. The texture, shape, and contextual information of the target objects are learned to localize the appearance of distinctive boundary, which is also explored to generate markers to split the touching nuclei. For further refinement of the segmentation, a coarse-to-fine nucleus segmentation framework is developed. The computational complexity of the segmentation is reduced by using superpixel instead of raw pixels. Extensive experimental results demonstrate that the proposed cervical nucleus cell segmentation delivers promising results and outperforms existing methods.
    IEEE transactions on bio-medical engineering 05/2015; DOI:10.1109/TBME.2015.2430895 · 2.23 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Acquisition of the standard plane is the prerequisite of biometric measurement and diagnosis during the ultrasound (US) examination. In this paper, a new algorithm is developed for the automatic recognition of the fetal facial standard planes (FFSPs) such as the axial, coronal, and sagittal planes. Specifically, densely sampled root scale invariant feature transform (RootSIFT) features are extracted and then encoded by Fisher vector (FV). The Fisher network with multi-layer design is also developed to extract spatial information to boost the classification performance. Finally, automatic recognition of the FFSPs is implemented by support vector machine (SVM) classifier based on the stochastic dual coordinate ascent (SDCA) algorithm. Experimental results using our dataset demonstrate that the proposed method achieves an accuracy of 93.27% and a mean average precision (mAP) of 99.19% in recognizing different FFSPs. Furthermore, the comparative analyses reveal the superiority of the proposed method based on FV over the traditional methods.
    PLoS ONE 05/2015; 10(5). DOI:10.1371/journal.pone.0121838 · 3.53 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, piezoelectric single crystalline PMNT layer with thickness ranging from several microns to several tens microns has been demonstrated on PMMA substrate for miniature ultrasound probes applications. By micromachining of the PMNT-on-substrate wafers, 2 mm diameter aperture ultrasonic transducer is fabricated. The ultrasonic pulse-echo characteristics of the transducer were measured and the results show that the non-matching layer single element transducer with 6 Mrayl backing layer has a center frequency of 45 MHz and a -6dB bandwidth of 25%. Experimental results are in good agreement with those predicted by the KLM model. This novel processing of PMNT-on-substrate micromachining can be applied to various MEMS applications such as acoustic sensor, ultrasonic transducer, resonator, pressure sensor, etc.
    Journal of Medical Imaging and Health Informatics 04/2015; 5(2). DOI:10.1166/jmihi.2015.1402 · 0.62 Impact Factor
  • Source
    SPIE Medical Imaging: Digital Pathology; 03/2015
  • [Show abstract] [Hide abstract]
    ABSTRACT: Past neuroimaging studies have focused on identifying specialized functional brain systems for processing different components of reading, such as orthography, phonology, and semantics. More recently, a few experiments have been performed to look into the integration and interaction of distributed neural systems for visual word recognition, suggesting that lexical processing in alphabetic languages involves both ventral and dorsal neural pathways originating from the visual cortex. In the present functional magnetic resonance imaging study, we tested the multiple pathways model with Chinese character stimuli and examined how the neural systems interacted in reading Chinese. Using dynamic causal modeling, we demonstrated that visual word recognition in Chinese engages the ventral pathway from the visual cortex to the left ventral occipitotemporal cortex, but not the dorsal pathway from the visual cortex to the left parietal region. The ventral pathway, however, is linked to the superior parietal lobule and the left middle frontal gyrus (MFG) so that a dynamic neural network is formed, with information flowing from the visual cortex to the left ventral occipitotemporal cortex to the parietal lobule and then to the left MFG. The findings suggest that cortical dynamics is constrained by the differences in how visual orthographic symbols in writing systems are linked to spoken language. Hum Brain Mapp, 2015. © 2015 Wiley Periodicals, Inc. © 2015 Wiley Periodicals, Inc.
    Human Brain Mapping 03/2015; DOI:10.1002/hbm.22792 · 6.92 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Since most image classification tasks involve discriminative information (i.e., saliency), this paper proposes a new bag-of-phrase (BoP) approach to incorporate this information. Specifically, saliency map and local features are first extracted from edge-based dense descriptors. These features are represented by histogram and mined with discriminative learning technique. Image score calculated from the saliency map is also investigated to optimize a support vector machine (SVM) classifier. Both feature map and kernel trick methods are explored to enhance the accuracy of the SVM classifier. In addition, novel inter- and intra-class histogram normalization methods are investigated to further boost the performance of the proposed method. Experiments using several publicly available benchmark datasets demonstrate that the proposed method achieves promising classification accuracy and superior performance over state-of-the-art methods.
    Pattern Recognition 02/2015; 48(8). DOI:10.1016/j.patcog.2015.02.004 · 2.58 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this letter, a 2.5 MHz 320-element 1.5D active matrix phased-array ultrasound transducer was designed and fabricated using high-temperature relaxor-based piezoelectric ternary Pb(In1/2Nb1/2)O3-Pb(Mg1/3Nb2/3)O3-PbTiO3 (PIN-PMN-PT) single crystals. The experiment and finite element analysis with the PZFlex package software were developed to design and optimize the 1.5D transducer. The FEM results show that the −6 dB fractional bandwidth is up to 83% using two λ/3.75 matching layers. A 64×5 active element PIN-PMN-PT single-crystal phased array was fabricated and tested. The measured center frequency and −6 dB fractional bandwidth were 2.5 MHz and 71%, respectively. Elevation beam profiles were obtained using a hydrophone, which showed that the measured −6 dB slice thickness was approximately 3 mm throughout the entire range of field of interest. The PIN-PMN-PT single crystal with over 30 degree usage temperature and higher Ec field should satisfy the demands of advanced high-density ultrasound transducer applications.
    Materials Letters 02/2015; 145. DOI:10.1016/j.matlet.2015.01.118 · 2.27 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The viscoelastic properties of the human cornea can provide valuable information for clinical applications such as the early detection of corneal diseases, better management of corneal surgery and treatment and more accurate measurement of intra-ocular pressure. However, few techniques are capable of quantitatively and non-destructively assessing corneal biomechanics in vivo. The cornea can be regarded as a thin plate in which the vibration induced by an external vibrator propagates as a Lamb wave, the properties of which depend on the thickness and biomechanics of the tissue. In this study, pulses of ultrasound radiation force with a repetition frequency of 100 or 200 Hz were applied to the apex of corneas, and the linear-array transducer of a SonixRP system was used to track the tissue motion in the radial direction. Shear elasticity and viscosity were estimated from the phase velocities of the A0 Lamb waves. To assess the effectiveness of the method, some of the corneas were subjected to collagen cross-linking treatment, and the changes in mechanical properties were validated with a tensile test. The results indicated that the shear modulus was 137 ± 37 kPa and the shear viscosity was 3.01 ± 2.45 mPa·s for the group of untreated corneas and 1145 ± 267 kPa and was 0.16 ± 0.11 mPa·s for the treated group, respectively, implying a significant increase in elasticity and a significant decrease in viscosity after collagen cross-linking treatment. This result is in agreement with the results of the mechanical tensile test and with reports in the literature. This initial investigation illustrated the ability of this ultrasound-based method, which uses the velocity dispersion of low-frequency A0 Lamb waves, to quantitatively assess both the elasticity and viscosity of corneas. Future studies could discover ways to optimize this system and to determine the feasibility of using this method in clinical situations.
    Ultrasound in Medicine & Biology 01/2015; 41(5). DOI:10.1016/j.ultrasmedbio.2014.12.017 · 2.10 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Shear wave based ultrasound elastography utilizes mechanical excitation or acoustic radiation force to induce shear waves in deep tissue. The tissue response is monitored to obtain elasticity information about the tissue. During the past two decades, tissue elasticity has been extensively studied and has been used in clinical disease diagnosis. However, biological soft tissues are viscoelastic in nature. Therefore, they should be simultaneously characterized in terms of elasticity and viscosity. In this study, two shear wave-based elasticity imaging methods, shear wave dispersion ultrasound vibrometry (SDUV) and acoustic radiation force impulsive (ARFI) imaging, were compared. The discrepancy between the measurements obtained by the two methods was analyzed, and the role of viscosity was investigated. To this end, four types of gelatin phantoms containing 0%, 20%, 30% and 40% castor oil were fabricated to mimic different viscosities of soft tissue. For the SDUV method, the shear elasticity μ1 was 3.90 ± 0.27 kPa, 4.49 ± 0.16 kPa, 2.41 ± 0.33 kPa and 1.31 ± 0.09 kPa; and the shear viscosity μ2 was 1.82 ± 0.31 Pa•s, 2.41 ± 0.35 Pa•s, 2.65 ± 0.13 Pa•s and 2.89 ± 0.14 Pa•s for 0%, 20%, 30% and 40% oil, respectively in both cases. For the ARFI measurements, the shear elasticity μ was 7.30 ± 0.20 kPa, 8.20 ± 0.31 kPa, 7.42 ± 0.21 kPa and 5.90 ± 0.36 kPa for 0%, 20%, 30% and 40% oil, respectively. The SDUV results demonstrated that the elasticity first increased from 0% to 20% oil and then decreased for the 30% and 40% oil. The viscosity decreased consistently as the concentration of castor oil increased from 0% to 40%. The elasticity measured by ARFI showed the same trend as that of the SDUV but exceeded the results measured by SDUV. To clearly validate the impact of viscosity on the elasticity estimation, an independent measurement of the elasticity and viscosity by dynamic mechanical analysis (DMA) was conducted on these four types of gelatin phantoms and then compared with SDUV and ARFI results. The shear elasticities obtained by DMA (3.44 ± 0.31 kPa, 4.29 ± 0.13 kPa, 2.05 ± 0.29 kPa and 1.06 ± 0.18 kPa for 0%, 20%, 30% and 40% oil, respectively) were lower than those by SDUV, whereas the shear viscosities obtained by DMA (2.52 ± 0.32 Pa·s, 3.18 ± 0.12 Pa·s, 3.98 ± 0.19 Pa·s and 4.90 ± 0.20 Pa·s for 0%, 20%, 30% and 40% oil, respectively) were greater than those obtained by SDUV. However, the DMA results showed that the trend in the elasticity and viscosity data was the same as that obtained from the SDUV and ARFI. The SDUV results demonstrated that adding castor oil changed the viscoelastic properties of the phantoms and resulted in increased dispersion of the shear waves. Viscosity can provide important and independent information about the inner state of the phantoms, in addition to the elasticity. Because the ARFI method ignores the dispersion of the shear waves, namely viscosity, it may bias the estimation of the true elasticity. This study sheds further light on the significance of the viscosity measurements in shear wave based elasticity imaging methods. Copyright © 2014 World Federation for Ultrasound in Medicine & Biology. Published by Elsevier Inc. All rights reserved.
    Ultrasound in Medicine & Biology 12/2014; 41(2). DOI:10.1016/j.ultrasmedbio.2014.09.028 · 2.10 Impact Factor
  • Nonlinear Dynamics 12/2014; 78(4):2897-2907. DOI:10.1007/s11071-014-1634-4 · 2.42 Impact Factor
  • Source
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, a new method for automatic vaginal bacteria cell segmentation and classification is proposed. Segmentation algorithm based on superpixel is first investigated to segment region of interest of the input image into cells. Feature extraction based on the segmented regions is trained by supervised deep learning method. Four types of different bacteria are studied for classification. Our experimental results show the classification result yields an accuracy of 99%, sensitivity of 100% and specificity of 98.04%. Compared to the state-of-the-arts method, better segmentation results have been achieved. Furthermore, our comparative analysis also shows that deep learning method outperforms traditional methods such as neural network and support vector machine.
    Journal of Medical Imaging and Health Informatics 10/2014; 4(5). DOI:10.1166/jmihi.2014.1320 · 0.62 Impact Factor
  • Computers in Industry 10/2014; 69. DOI:10.1016/j.compind.2014.09.006 · 1.46 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Acquisition of the standard plane is crucial for medical ultrasound diagnosis. However, this process requires substantial experience and a thorough knowledge of human anatomy. Therefore it is very challenging for novices and even time consuming for experienced examiners. We proposed a hierarchical, supervised learning framework for automatically detecting the standard plane from consecutive 2-D ultrasound images. We tested this technique by developing a system that localizes the fetal abdominal standard plane from ultrasound video by detecting three key anatomical structures: the stomach bubble, umbilical vein and spine. We first proposed a novel radial component-based model to describe the geometric constraints of these key anatomical structures. We then introduced a novel selective search method which exploits the vessel probability algorithm to produce probable locations for the spine and umbilical vein. Next, using component classifiers trained by random forests, we detected the key anatomical structures at their probable locations within the regions constrained by the radial component-based model. Finally, a second-level classifier combined the results from the component detection to identify an ultrasound image as either a "fetal abdominal standard plane" or a "non- fetal abdominal standard plane." Experimental results on 223 fetal abdomen videos showed that the detection accuracy of our method was as high as 85.6% and significantly outperformed both the full abdomen and the separate anatomy detection methods without geometric constraints. The experimental results demonstrated that our system shows great promise for application to clinical practice.
    Ultrasound in Medicine & Biology 09/2014; 40(11). DOI:10.1016/j.ultrasmedbio.2014.06.006 · 2.10 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Research on how lexical tone is neuroanatomically represented in the human brain is central to our understanding of cortical regions subserving language. Past studies have exclusively focused on tone perception of the spoken language, and little is known as to the lexical tone processing in reading visual words and its associated brain mechanisms. In this study, we performed two experiments to identify neural substrates in Chinese tone reading. First, we used a tone judgment paradigm to investigate tone processing of visually presented Chinese characters. We found that, relative to baseline, tone perception of printed Chinese characters were mediated by strong brain activation in bilateral frontal regions, left inferior parietal lobule, left posterior middle/medial temporal gyrus, left inferior temporal region, bilateral visual systems, and cerebellum. Surprisingly, no activation was found in superior temporal regions, brain sites well known for speech tone processing. In activation likelihood estimation (ALE) meta-analysis to combine results of relevant published studies, we attempted to elucidate whether the left temporal cortex activities identified in Experiment one is consistent with those found in previous studies of auditory lexical tone perception. ALE results showed that only the left superior temporal gyrus and putamen were critical in auditory lexical tone processing. These findings suggest that activation in the superior temporal cortex associated with lexical tone perception is modality-dependent. Hum Brain Mapp, 2014. © 2014 Wiley Periodicals, Inc.
    Human Brain Mapping 09/2014; 36(1). DOI:10.1002/hbm.22629 · 6.92 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Currently, the evaluation of placental maturity has mainly focused on subjective measure, which highly depends on the observation and experiences of the clinicians and not reliable. This paper proposes a new method for grading placenta maturity in B-mod ultrasound (US) images automatically based on local intensity order pattern (LIOP) and fisher vector (FV). After extracting invariant LIOP feature from the affine covariant region, the feature is encoded by FV to improve the classification accuracy and reduce the processing time. Experimental results show the effectiveness of the proposed method with an accuracy of 0.9375, a sensitivity of 0.9804 and a specificity of 0.9375 for the placental maturity grading. Moreover, experimental results demonstrate that the LIOP feature outperforms the traditional SIFT feature for grading.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Compared to 1D phased array probes with a fixed focus in elevation, multi-row arrays can significantly improved the slice thickness throughout image by expanding aperture and dynamic focusing in elevation. This paper describes the design and measurement of a 64 × 5 element 1.5D ultrasonic transducer that enables dynamic focusing and apodization in the elevation direction. We manufactured a transducer with an aperture size of 24 mm × 16 mm using a widely-used piezoelectric ceramic (PZT-5H) as the piezoelectric vibrator. The measured center frequency and -6 dB fractional bandwidth of the 1.5D transducer were 3 MHz and 79%, respectively. A two-way insertion loss of -58 dB was obtained at the average center frequency. We carried out a sound field simulation and measured the actual transmitting (one-way) sound field data by using a hydrophone. In this way the sound beam profile in elevation direction was obtained, showing the -6 dB slice thickness measured was about 2 mm throughout the whole range of interest (from near to far field). This provided a much greater focal depth and much better imaging resolution in the elevation direction than can be achieved by using 1D probe.
    Sensors and Actuators A Physical 08/2014; 214. DOI:10.1016/j.sna.2014.04.028 · 1.94 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, a superpixel and convolution neural network (CNN) based segmentation method is proposed for cervical cancer cell segmentation. Since the background and cytoplasm contrast is not relatively obvious, cytoplasm segmentation is first performed. Deep learning based on CNN is explored for region of interest detection. A coarse-to-fine nucleus segmentation for cervical cancer cell segmentation and further refinement is also developed. Experimental results show that an accuracy of 94.50% is achieved for nucleus region detection and a precision of 0.9143±0.0202 and a recall of 0.8726±0.0008 are achieved for nucleus cell segmentation. Furthermore, our comparative analysis also shows that the proposed method outperforms the related methods.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Automation-assisted reading (AAR) techniques have the potential to reduce errors and increase productivity in cervical cancer screening. The sensitivity of AAR relies heavily on automated segmentation of abnormal cervical cells, which is handled poorly by current segmentation algorithms. In this paper, a global and local scheme based on graph cut approach is proposed to segment cervical cells in images with a mix of healthy and abnormal cells. For cytoplasm segmentation, the multi-way graph cut is performed globally on the a* channel enhanced image, which can be effective when the image histogram presents a non-bimodal distribution. For segmentation of nuclei, especially when they are abnormal, we propose to use graph cut adaptively and locally, which allows the combination of intensity, texture, boundary and region information. Two concave points-based approaches are integrated to split the touching-nuclei. As part of an ongoing clinical trial, preliminary validation results obtained from 21 cervical cell images with non-ideal imaging condition and pathology show that our segmentation method achieved 93% accuracy for cytoplasm, and 88.4% F-measure for abnormal nuclei, outperforming state of the art methods in terms of accuracy. Our method has the potential to improve the sensitivity of AAR in screening for cervical cancer.
    Computerized medical imaging and graphics: the official journal of the Computerized Medical Imaging Society 07/2014; DOI:10.1016/j.compmedimag.2014.02.001 · 1.50 Impact Factor

Publication Stats

102 Citations
96.45 Total Impact Points


  • 2006–2015
    • Shenzhen University
      • College of Information Engineering
      Bao'an, Guangdong, China
  • 2011–2012
    • Zhejiang University
      • College of Biomedical Engineering and Instrument Science
      Hang-hsien, Zhejiang Sheng, China
  • 2009
    • Xi'an Jiaotong University
      • Key Laboratory of Biomedical Information Engineering of Ministry of Education
      Ch’ang-an, Shaanxi, China