Establishment of the new analytical method for the grade of breast cancer that combined MRI and deep learning: development to custom-made therapy

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K08017
• 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: Kanazawa University
• Principal Investigator: Kawashima Hiroko 金沢大学, 保健学系, 教授 (70293355)
• Co-Investigator(Kenkyū-buntansha): 宮地 利明 金沢大学, 保健学系, 教授 (80324086)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: MRI / 深層学習 / 乳癌 / 悪性度

Outline of Final Research Achievements

This study analyzed enormous information to be obtained from MRI using deep learning and was intended to establish the new analytical method for the grade of breast cancer to be connected directly with custom-made therapy. We targeted a prediction of the pathological complete response of preoperative chemotherapy. We investigated it if we put clinical data, contrast-enhanced MRI data, characteristic quantity in the first-order from MRI, characteristic quantity in the second-order from MRI together how whether it was highest-precision. As a result, for the prediction of the pathological complete response, the thing that became highest-precision was found when we utilized all information of clinical data, MRI data, the characteristic quantity in the first-order from MRI, and the characteristic quantity in the second-order from MRI.

Free Research Field

放射線科学

Academic Significance and Societal Importance of the Research Achievements

乳癌はそのサブタイプによって大筋の治療方針が決定されるが，１つのサブタイプの中でも生物学的悪性度には幅がある．MRIは乳癌患者の術前検査として広く定着しているが，乳癌の悪性度をMRIの数少ないパラメータで正確に予測することは難しい．今回，ディープラーニングでMRI情報を解析し，乳癌の悪性度を予測することを試みた．その結果，術前化学療法の病理学的完全奏功を治療開始前に予測するためには，臨床データ，基本的なMRIデータに加え，造影MRIデータから抽出した第一段階の特徴量および第二段階の特徴量のすべての情報を活用した場合に最も精度が高くなることがわかり，ディープラーニングの可能性が示された．