Toward New-Generation AI-Based CAD System: Development of Interpretable Deep Learning-Based CAD System for Breast Cancer Diagnosis Using Mammogram

本研究的目的是开发一个可解释的基于深度学习（DL）的计算机辅助诊断（CAD）系统，用于乳房x光检查中的乳腺癌诊断。在2021财年的基础上，我们在2022财年取得了以下进展：(1)在上一财年收集的四组乳腺x线照片数据集上进行了DL模型在病变检测中的评价实验。(2)提出了一种新的训练方法，利用临床医生的逐像素注释和显著性图来提高深度学习模型的准确性，并进行了测试。(3)在国际相关期刊上发表论文2篇。

The purpose of this research is to develop an interpretable deep learning (DL)-based computer-aided diagnosis (CAD) system for breast cancer diagnosis in mammogram. On the base of the achivement of FY2021, we achived the following progresses in the FY2022.(1) Experments for evaluation of DL models in lesion detection has been conducted on four mammogram data sets, which were collected in the previous FY.(2) A new training method, which utilized the clinicians's pixel-wise anotation and saliency maps to improve the DL model acuuracy, was proposed and tested.(3) Two papers has been published in the related international journals.