Faster Region Convolutional Neural Networks Applied to Ultrasonic Images for Breast Lesion Detection and Classification
Kaizhen Wei, Boyang Wang, Jafar Saniie · 2020
According to the global cancer data report of the International Agency for Research on Cancer, breast cancer already becomes a common and widespread disease around the world and is also the leading cause of death from cancer in women. Compared with mammograms and magnetic resonance imaging, breast ultrasound imaging is a universal and inexpensive test method for breast lesion screening without additional risk from radiation. However, most radiologists still need support and assistance from the computer-aided diagnosis systems for the detection and classification of breast cancer in ultrasonic images by improving the differentiation between malignant and benign lesions. In this paper, we propose a lesion detection and classification algorithm for breast cancer using Faster Region Convolutional Neural Networks (Faster R-CNN) applied to ultrasonic images. The proposed algorithm is capable of locating breast lesions with a bounding box and characterizing the detected lesion as malignant or benign. An open-access series of ultrasonic data of breast lesions is used for training and testing the Faster R-CNN, and the examination result shows that the algorithm is capable of precisely locate and characterize the breast lesion with the accuracy of more than 95%.