Automated Diagnosis of Obstetrics and Gynecology MRI Using Deep Learning

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K16780
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 52040:Radiological sciences-related
• Research Institution: Kyoto University
• Principal Investigator: Yasuhisa Kurata 京都大学, 医学研究科, 助教 (40836178)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: Deep learning / MRI / Endometrial cancer

Outline of Final Research Achievements

In this project, we realized automatic segmentation and staging of uterine endometrial cancer on MRI using convolutional neural networks, and reported the results at national and international conferences and in research papers. In parallel, we are continuing research to extend the same method to bladder cancer imaging, and have already reported some of the results in papers. We also incorporated Vision Transformer, a new deep learning method proposed during the research period, and showed that the method is applicable to medical image analysis of obstetrics and gynecology and urology. In the future, we plan to conduct a multicenter study for external validation of the developed model.

Free Research Field

画像診断（特に産婦人科領域）

Academic Significance and Societal Importance of the Research Achievements

最近ではいわゆる人工知能を用いた医用画像解析が盛んに行われているが、産婦人科領域に関する研究報告は比較的少なかった。本研究では、MRI上で子宮体癌の検出や病期診断（深達度の判定）の自動化を実現することで、産婦人科領域の画像診断における深層学習の有用性を示した。この種の研究報告が増えることで、他の領域と同様に、産婦人科画像診断の質的向上や個別化医療を目指した画像解析が進展していくと考えられる。