2023 Fiscal Year Final Research Report
Pathological diagnosis based on integration of morphological images and multi-layer omics data by artificial intelligence
Project/Area Number |
21H02705
|
Research Category |
Grant-in-Aid for Scientific Research (B)
|
Allocation Type | Single-year Grants |
Section | 一般 |
Review Section |
Basic Section 49020:Human pathology-related
|
Research Institution | Keio University |
Principal Investigator |
Kanai Yae 慶應義塾大学, 医学部(信濃町), 教授 (00260315)
|
Co-Investigator(Kenkyū-buntansha) |
榊原 康文 慶應義塾大学, 理工学部(矢上), 教授 (10287427)
新井 恵吏 慶應義塾大学, 医学部(信濃町), 准教授 (40446547)
|
Project Period (FY) |
2021-04-01 – 2024-03-31
|
Keywords | 病理診断 / 人工知能 / オミックス解析 / 病理画像 |
Outline of Final Research Achievements |
The aim of this study was to build a deep learning model to predict cancer treatment response and prognosis by fusing pathological morphology images and omics information with the assistance of artificial intelligence (AI). Using micrographs and virtual slide data of renal cell carcinoma surgical specimens, a convolutional neural network model of CpG island methylation trait (CIMP)-positive renal cell carcinoma with poor prognosis was constructed. Furthermore, gradient-weighted class activation mapping was used to visualise which regions of the pathological image are focused on when discriminating CIMP-positive and CIMP-negative. Additional multilayer omics information is currently being acquired to improve the prognostic power of the model.
|
Free Research Field |
人体病理学
|
Academic Significance and Societal Importance of the Research Achievements |
病理診断は従来から、がん等の臨床症例の最終診断を担ってきたが、がんゲノム医療が社会実装された今日にあっては、従来通り形態像のみに基づく組織型分類にとどまるべきではない。病理診断学は、オミックス情報を取り込んで、ブレイクスルーを果たすべきである。可視化したCIMP判定時のAIの着眼点を形態学的診断基準に翻訳することにより、病理医が顕微鏡で見るだけでモデルと同等の治療奏効性予測・予後予測を実施できれば、「病理医とAIの創造的協働による、オミックス情報を統合した新しい病理診断の創出」の端緒となると期待される。
|