Creation of an analysis platform and AI to detect lung fibrosis CT images with high risk of developing lung cancer.

• Project/Area Number: 18K15553
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 52040:Radiological sciences-related
• Research Institution: Saga University
• Principal Investigator: Egashira Ryoko 佐賀大学, 医学部, 助教 (70457464)
• Project Period (FY): 2018-04-01 – 2024-03-31
• Project Status: Completed (Fiscal Year 2023)
• Budget Amount: ¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000); Fiscal Year 2021: ¥780,000 (Direct Cost: ¥600,000、Indirect Cost: ¥180,000); Fiscal Year 2020: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000); Fiscal Year 2019: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000); Fiscal Year 2018: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
• Keywords: 間質性肺炎における肺癌発生 / 肺線維症 / 人工知能 / 間質性肺炎 / 間質性肺炎に合併した肺癌の自動セグメンテーション / 間質性肺炎に合併した肺癌の自動ラベリング / 肺癌 / 画像診断 / Radiogenomics

Outline of Final Research Achievements

The aim of the research was to detect image findings of lung fibrosis that may be a risk factor for lung cancer development by training an artificial intelligence (AI) on "CT images of lung fibrosis that developed lung cancer" and "CT images of lung fibrosis that did not develop lung cancer after long-term follow-up" in collaboration with an engineering research collaborator. The model was built using the Encoder part of SimSiam and set up so that images with similar features were placed on the latent space. Using the past images of the area where lung cancer occurred in the course of the disease and the same area (positive images) and the images of the course of the area that did not develop cancer (negative images) as search images, it was found that the detection of high-risk areas for carcinogenesis was possible from similar image search.

Academic Significance and Societal Importance of the Research Achievements

肺線維症のCT画像において，肺癌が発生する領域には，発生しない領域と比べ何らかの特徴が存在し，それを把握しておくことにより，リスクの高い患者さんを早期発見し，厳重に経過観察することが可能となると考えられる．

Research Products

• [Presentation] 肺癌合併間質性肺炎患者のCT画像における機械学習を用いた非癌部画像情報を用いた発癌肺の判別法の検討 (2021)
  • Author(s): 吉田直樹
  • Organizer: 生体医工学会大会