| Project/Area Number |
20K08012
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 52040:Radiological sciences-related
|
|---|
| Research Institution | Sendai National College of Technology |
|---|
Principal Investigator |
Zhang Xiaoyong 仙台高等専門学校, 総合工学科, 准教授 (90722752)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
費 仙鳳 東北文化学園大学, 工学部, 准教授 (20620470)
|
|---|
| Project Period (FY) |
2020-04-01 – 2024-03-31
|
| Project Status |
Completed (Fiscal Year 2023)
|
| Budget Amount *help |
¥4,290,000 (Direct Cost: ¥3,300,000、Indirect Cost: ¥990,000)
Fiscal Year 2022: ¥520,000 (Direct Cost: ¥400,000、Indirect Cost: ¥120,000)
Fiscal Year 2021: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2020: ¥2,470,000 (Direct Cost: ¥1,900,000、Indirect Cost: ¥570,000)
|
|---|
| Keywords | マンモグラフィー / 計算機支援診断 / 深層学習 / 説明可能なAI / 乳がん / Mammograpy / Deep Learning / Explainable AI / Computer-Aided Diagnosis / Lesion Detection / Interpretablity / Breast Cancer / Computer-Aided Detection / Artificial Intellegence |
|---|
| Outline of Research at the Start |
Deep learning (DL) has attracted extensive efforts for medical image analysis in recent years, such as breast cancer detection in mammogram.. However, since the DL worked in a “black box” fashion, its reliability became a vital problem for clinical application. For solve this problem, this research will focus on developing an interpretable DL-based computer-aided diagnosis system that can not only detect breast cancer in mammograms (making decision), but also produce a visual interpretation to provide understanding of the decision-making process (interpreting decision).
|
| Outline of Final Research Achievements |
The purpose of this study was to develop an explainable AI-based mammographic computer-aided diagnosis (CAD) system for breast cancer detection. We mainly focus on investigating the insight mechanism of AI models in reading a medical image and use it to help clinicians to understand the decision-making of AI models.
A digital mammogram dataset consisting of 30,000 cases with train labels were collected for training AI system. Several deep learning (DL)-based methods were developed for mammogram classification and mass detection in mammograms. Visualization techniques were utilized to generate visual explanations of diagnosis results. In addition, a domain shift issue related to train data was also investigated in this study.
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究では、AIのブラックボックス性を解消するため、AI内部の可視化などの技術を用いて、説明可能なAI診断システムの開発を目的とする。説明可能なAI診断システムの開発は臨床面でも非常に重要な意義をもつ。本研究では、画像解剖学的知見に基づく解析により、診断根拠の解釈を可能とすることで、正確な診断だけでなく信頼性の高い医療AIシステムの実用が可能であることを実証した。
|
