Deep learning-based analysis of probe-based confocal laser microscopic findings to develop a new diagnostic methods of diffuse lung diseases

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K08216
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 53030:Respiratory medicine-related
• Research Institution: Fujita Health University
• Principal Investigator: Imaizumi Kazuyoshi 藤田医科大学, 医学部, 教授 (50362257)
• Co-Investigator(Kenkyū-buntansha): 寺本 篤司 藤田医科大学, 保健学研究科, 教授 (00513780)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: びまん性肺疾患 / 共焦点レーザー顕微鏡 / 深層学習 / 気管支鏡

Outline of Final Research Achievements

We investigated the possibility of classifying and diagnosing diffuse lung diseases using AI analyses of peripheral lung autofluorescence microscopic video images obtained using probe-based confocal laser endomicroscopy (pCLE). pCLE images of interstitial pneumonia (UIP and fibrotic NSIP) were analyzed using texture analysis and machine learning methods. When 74 types of image features were extracted, and the Naive Bayes method was used as a machine learning method, the classification accuracy rate was 64.7% on an individual image basis and 67% on a case basis. AI-based differentiation of interstitial pneumonia disease types from pCLE images could be a promising technology that can be a novel and non-invasive diagnostic tool for differential diagnosis of diffuse lung diseases.

Free Research Field

呼吸器内科学

Academic Significance and Societal Importance of the Research Achievements

プローブ型共焦点レーザー顕微内視鏡 (pCLE)による末梢肺自家蛍光顕微鏡画像は、気管支鏡生検鉗子口を通して肺の末梢に挿入し直接肺の微細構造を観察できる簡便な方法である。肺内には肺胞壁　気管支壁の弾性線維や肺胞マクロファージに自家蛍光があり、これらを直接観察することで肺胞壁肥厚や弾性線維の破壊程度、あるいは肺胞内の炎症状況の微細像がリアルタイムで観察できる。得られる画像は動画データであり、症例毎・疾患毎の画像特徴を肉眼的に抽出するのは困難であるが、本研究でAIを用いた解析が可能であることが示唆されたことは、今後、新しい非浸襲的なびまん性肺疾患の病理診断の確立につながる。