development of bony lesion detection system for CT images by unsupervised deep learning

2020 Fiscal Year Final Research Report

• Project/Area Number: 18K12095
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 90130:Medical systems-related
• Research Institution: The University of Tokyo
• Principal Investigator: Hanaoka Shouhei 東京大学, 医学部附属病院, 講師 (80631382)
• Project Period (FY): 2018-04-01 – 2021-03-31
• Keywords: 医用画像処理 / 深層学習 / X線CT / 骨転移

Outline of Final Research Achievements

We developed a computer program which can estimate each voxel value of the latest CT examination from voxel values of the previous CT examination. Furthermore, it can also estimate the estimation error. Using these estimated CT value and the error, z-score of each voxel in the latest CT examination is calculated so that the anomaly-highlighted image is displayed. The proposed system was validated with a real film-reading environment. The experiment was performed with 11 radiologists and 80 datasets, and the improve of the receiver operating characteristic (ROC) curve was confirmed using the proposed temporary subtracted CT.

Free Research Field

医用画像処理

Academic Significance and Societal Importance of the Research Achievements

CT画像においてしばしば主治医や放射線科読影医によって見逃される早期のがん骨転移について、その検出を助ける骨病変抽出・強調表示手法が開発できた。これにより、がん骨転移をより早期に発見し治療できることが期待され、がん患者の予後、quality of lifeの向上に資すことができることと期待される。