Application of multi-task deep learning with brain MRI to psychiatric and neurological disorders

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K07593
• 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: Kyoto University
• Principal Investigator: Oishi Naoya 京都大学, 医学研究科, 准教授 (40526878)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 深層学習 / MRI / 脳 / 精神神経疾患 / マルチタスク / セグメンテーション / クラス分類

Outline of Final Research Achievements

To further improve the accuracy of segmentation and classification in neuroimaging research, we have newly developed a wide range of multi-task deep learning algorithms for brain MRI. The algorithm not only accurately identifies micro brain structures that have been suggested to be important in psychiatric and neurological disorders but are difficult to identify with conventional techniques due to their minute size, but also enables more accurate identification of foci in patients with traumatic brain injury and brain tumors by using multiple sequences of MRI data. Thus, we have demonstrated the usefulness of the multi-tasking deep learning algorithm developed in the project.

Free Research Field

医用画像工学

Academic Significance and Societal Importance of the Research Achievements

柔軟性・拡張性の高いマルチタスク深層学習アルゴリズムを新規に開発し、健常者における微小脳領域の正確な同定や精神・神経疾患における病巣の正確な同定などに成功した点は学術的意義が高いと考えられる。また、本手法はMRIに限定される技術ではなく、様々な波形（１次元）、画像（2～3次元）、動画（3～4次元）研究において応用可能であり、多岐にわたる臨床応用可能性があるという観点で社会的意義も高いと考えられる。