Deep learning based classification of breast tumors with shear-wave elastography

• Published in: Ultrasonics 2016-12, Vol.72, p.150-157
• Main Authors: Zhang, Qi, Xiao, Yang, Dai, Wei, Suo, Jingfeng, Wang, Congzhi, Shi, Jun, Zheng, Hairong
• Format: Article
• Language: English
• Subjects: Adult; Architecture; Biopsy; Breast; Breast Neoplasms - diagnostic imaging; Breast Neoplasms - pathology; Breast tumors; Classification; Computer-aided diagnosis; Construction; Deep learning; Diagnosis, Differential; Elasticity Imaging Techniques - methods; Female; Humans; Image Interpretation, Computer-Assisted - methods; Learning; Neural Networks (Computer); Point-wise gated Boltzmann machine; Retrospective Studies; Sensitivity and Specificity; Shear-wave elastography; Texture; Tumors; Ultrasonography, Mammary - methods
• ISSN: 0041-624X; 1874-9968
• DOI: 10.1016/j.ultras.2016.08.004

•We automatically extract learned-from-data features from SWE of breast tumors by DL.•Our DL model includes point-wise gated Boltzmann machine and RBM.•Our DL model gets 93.4% accuracy, 88.6% sensitivity, 97.1% specificity and 0.947 AUC.•The DL model is superior to the model using statistical features. This study aims to build a deep learning (DL) architecture for automated extraction of learned-from-data image features from the shear-wave elastography (SWE), and to evaluate the DL architecture in differentiation between benign and malignant breast tumors. We construct a two-layer DL architecture for SWE feature extraction, comprised of the point-wise gated Boltzmann machine (PGBM) and the restricted Boltzmann machine (RBM). The PGBM contains task-relevant and task-irrelevant hidden units, and the task-relevant units are connected to the RBM. Experimental evaluation was performed with five-fold cross validation on a set of 227 SWE images, 135 of benign tumors and 92 of malignant tumors, from 121 patients. The features learned with our DL architecture were compared with the statistical features quantifying image intensity and texture. Results showed that the DL features achieved better classification performance with an accuracy of 93.4%, a sensitivity of 88.6%, a specificity of 97.1%, and an area under the receiver operating characteristic curve of 0.947. The DL-based method integrates feature learning with feature selection on SWE. It may be potentially used in clinical computer-aided diagnosis of breast cancer.