Incorporating the Breast Imaging Reporting and Data System Lexicon with a Fully Convolutional Network for Malignancy Detection on Breast Ultrasound.

Yung-Hsien Hsieh,Seng-Tong Dai,Dar-Ren Chen,Hsin-Ya Huang,Fang-Rong Hsu,Wei-Chung Shia

doi:10.3390/diagnostics12010066

Abstract

In this study, we applied semantic segmentation using a fully convolutional deep learning network to identify characteristics of the Breast Imaging Reporting and Data System (BI-RADS) lexicon from breast ultrasound images to facilitate clinical malignancy tumor classification. Among 378 images (204 benign and 174 malignant images) from 189 patients (102 benign breast tumor patients and 87 malignant patients), we identified seven malignant characteristics related to the BI-RADS lexicon in breast ultrasound. The mean accuracy and mean IU of the semantic segmentation were 32.82% and 28.88, respectively. The weighted intersection over union was 85.35%, and the area under the curve was 89.47%, showing better performance than similar semantic segmentation networks, SegNet and U-Net, in the same dataset. Our results suggest that the utilization of a deep learning network in combination with the BI-RADS lexicon can be an important supplemental tool when using ultrasound to diagnose breast malignancy.

Highlights

Publisher’s Note: MDPI stays neutralBreast ultrasound (US) imaging is an important and common examination for the clinical diagnosis of breast cancer
We focused on the ability of semantic segmentation, combining deep network and the Breast Imaging Reporting and Data System (BI-RADS) lexicon, to facilitate multi-target segmentation of US images by comparing the similarity of this prediction result to that of the radiology report drafted by experienced physicians
This visualization is clinically impactful, for physicians and radiologists, because it can show all detected US image features that are synonymous with the BI-RADS malignant lexicon at a glance; this considerably decreases the effort of visually reading the image

Summary

Introduction

Breast ultrasound (US) imaging is an important and common examination for the clinical diagnosis of breast cancer. It is a non-radiation imaging method, well tolerated by patients that can be integrated into interventional procedures [1]. Deep learning has undergone rapid development with various and deeper network architecture and currently plays an important role in medical imaging analysis and computeraid diagnosis. Deep learning generates a standardized analysis with objective and consistent results, and it can discover significant, hidden, provide a powerful reference in the clinic, and decrease observer bias. Image Pixel Count: The total number of pixels in each lexicon in all images. Image Count: Total number of images in each lexicon

Objectives

Methods

Results

Discussion

Conclusion