Computer-aided diagnosis system for breast ultrasound images using deep learning

Computer-aided diagnosis system for breast ultrasound images using deep learning

S4 Ultrasound in Medicine & Biology A small pilot study was performed to assess feasibility of shearwave elastography during treatment process. Resu...

37KB Sizes 0 Downloads 49 Views

S4

Ultrasound in Medicine & Biology

A small pilot study was performed to assess feasibility of shearwave elastography during treatment process. Results: The Histogram Width Tissue Characterization (HWTC) between 35 and 50 year age group shows significant difference between the epidermal and subcutaneous fat tissue characterization. The epidermis layer of 50 year old female has less gray scale levels with reduced HWTC rate. Conclusion: Further research is encouraged to include large sample populations. Following collagen filler treatment or other aesthetic methods, follow up sonography and elastography methods may prove to provide a measurable outcome in treatment. There is limited knowledge regarding elastography, shear wave elastography and strain elastography, which provides a large field of novel research to be delved upon in future. It is expected in future technology for tissue elasticity imaging there will be sophisticated improvements such as three dimensional and quantitative elasticity images based on tissue elasticity modulus.

Computer-aided diagnosis system for breast ultrasound images using deep learning Hiroki Tanaka,1 Shih-Wei Chiu,1 Takanori Watanabe,2 Setsuko Kaoku,3 Takuhiro Yamaguchi1 1 Tohoku University Graduate School of Medicine, Sendai, Miyagi, Japan, 2 National Hospital Organization Sendai Medical Center, Sendai, Miyagi, Japan, 3 National Hospital Organization Osaka National Hospital, Osaka, Japan Introduction: Ultrasonography has been recommended as an adjunctive modality to mammography, which is insufficient for accurate diagnosis of breast cancer in women with dense breast tissue. However, ultrasonography has the disadvantage of being operator dependent, requiring proficiency in reading ultrasound (US) images and increasing the false positive rate. To overcome these problems, we aimed to develop a computer-aided diagnosis (CAD) system for classification of breast malignant and benign masses on ultrasonography based on the convolutional neural network (CNN), a state-of-the-art deep learning technique. Methods: From a large ultrasound (US) image database, managed by Japan Association of Breast and Thyroid Sonology (JABTS), we collected images of 1536 breast masses (897 malignant and 639 benign) confirmed by pathological examinations, with each breast mass captured from various angles using the US imaging probe. We constructed an ensemble network by combining two CNN models (VGG192 and ResNe1523) fine-tuned on balanced training data with the data augmentation which is a popular technique for synthetically generating new samples from the original and used the mass level classification method enabling the CNN to classify a given mass using all views. We explored the grounds of classification by generating a heatmap capable of presenting important regions used by the CNN for classification in humans. Results: On an independent 154 test masses (77 malignant and 77 benign), our network showed outstanding classification performance with a sensitivity of 90.9%, a specificity of 87.0%, and an AUC of 0.951 compared with the two CNN models. In addition, our study indicated that not only breast masses but also surrounding tissues are important regions for correct classification. Conclusion: This CNN-based CAD system is expected to assist doctors from the viewpoint of second opinion and improve the diagnosis of breast cancer in clinical practice. Acknowledgments—We would like to thank SAS Institute Japan Ltd for their technical support and offering the development environment for deep learning, which was founded by SAS Institute Inc.

Volume 45, Number S1, 2019 References: 1. Suzuki K. Overview of deep learning in medical imaging. Radiol Phys Technol. 2017;10:257-273. 2. Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. Int Conf on Learning Representation (ICLR). 2015;1-13. 3. He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition Proc of the IEEE Conf on Computer Vision and Pattern Recognition. 2015;770-778.

Grading fatty liver and detecting liver pathological features - Is there consensus between sonographers and radiologists? David Sheng-Liang Yang,1 Michal Schneider,2 Paul Lombardo2 1 Department of Ultrasound, Canberra Imaging Group, Canberra, ACT, Australia, 2 Department of Medical Imaging and Radiation Sciences, Monash University, Clayton, VIC, Australia Introduction: Hepatic steatosis is the leading cause of chronic liver disease. Reliable detection and staging of liver disease is important to facilitate diagnosis and treatment. The aim of this study was to evaluate the interobserver agreement between trainee sonographers, qualified sonographers and radiologists in grading non-alcoholic fatty liver disease (NAFLD) and detecting common liver pathological features in B-mode images. Methods: 150 B-mode liver ultrasound images from 50 adult patients referred for an abdominal ultrasound were obtained retrospectively from a PACS system. The images were independently graded for the severity of hepatic steatosis (normal, mild, moderate or severe) and the detection of incidental 0findings, focal fatty sparing, liver surface irregularity and rounded liver edge (present or absent) by 17 qualified, six trainee sonographers and six radiologists. Fleiss’ kappa statistics were used to calculate interobserver agreement among participants in grading common liver pathological features. Intraclass correlation coefficient (ICC) were used to calculate the level of absolute agreement among participants in grading NAFLD. Results: The interobserver agreement rates among trainee sonographers for the detection of incidental findings, focal fatty sparing, liver surface irregularity and rounded liver edge were: k = 0.243, 0.486, 0.155 and 0.079 respectively. Among qualified sonographers, the agreement rates were: k = 0.323, 0.428, 0.167 and 0.152 respectively. Among radiologists, the agreement rates were: k = 0.156, 0.266, 0.015 and 0.154 respectively. The intraclass correlation coefficient average scores among trainee sonographers, qualified sonographers, all the sonographers combined and radiologists were: 0.927, 0.978, 0.985 and 0.954 respectively. Conclusion: Visual assessment of common liver pathology in B-mode imaging has low interobserver agreement among sonographers and radiologists, but excellent inter-rater reliability in grading NAFLD. The low agreement levels are likely caused by a lack of standardised assessment criteria. The development of a standardised criteria for staging NAFLD and common liver pathological features are recommended. Key Words: Hepatic steatosis, interobserver agreement, radiologists, sonographers, grading. 2D Shear Wave Elastography in risk prediction of indeterminate thyroid nodules compared to histological findings Pedro P. Moraes, Marcelo M. Straus, Marcelo M. Schelini, Rosa R. Sigrist, Maria Cristina M.C. Chammas Hospital das Clınicas School of Medicine - University of S~ ao Paulo, e PAULO, Brazil SAO