BIRADS features-oriented semi-supervised deep learning for breast ultrasound computer-aided diagnosis.

Researchers

Journal

Modalities

Models

Abstract

We propose a novel BIRADS-SSDL network that integrates clinically-approved breast lesion characteristics (BIRADS features) into a task-oriented Semi-Supervised Deep Learning (SSDL) for accurate diagnosis on ultrasound (US) images with a small training dataset. Breast US images are converted to BIRADS-oriented Feature Maps (BFMs) using a distance-transformation coupled with a Gaussian filter. Then, the converted BFMs are used as the input of an SSDL network, which performs unsupervised Stacked Convolutional Auto-Encoder (SCAE) image reconstruction guided by lesion classification. This integrated multi-task learning allows SCAE to extract image features with the constraints from the lesion classification task, while the lesion classification is achieved by utilizing the SCAE encoder features with a convolutional network. We trained BIRADS-SSDL network with an alternative learning strategy by balancing reconstruction error and classification label prediction error. We compared the performance of the BIRADS-SSDL network with conventional SCAE and SSDL methods that use original images as inputs, as well as with an SCAE that use BFMs as inputs. Experimental results on two breast US datasets show that BIRADS-SSDL ranked the best among the four networks, with classification accuracy around 94.23±3.33% and 84.38±3.11%. In the case of experiments across two datasets collected from two different institution/and US devices, the developed BIRADS-SSDL is generalizable across the different US devices and institutions without overfitting to a single dataset and achieved satisfactory results. Furthermore, we investigate the performance of the proposed method by varying model training strategies, lesion boundary accuracy, and Gaussian filter parameter. Experimental results showed that pre-training strategy can help to speed up model convergence during training but no improvement of classification accuracy on testing dataset. The proposed method could achieve a satisfactory performance when there are slight deviation of the lesion’s boundary. Compared with state-of-the-art methods, BIRADS-SSDL could be promising for effective breast US lesion CAD using small datasets.
© 2020 Institute of Physics and Engineering in Medicine.