Investigation of a Model for Evaluating Cognitive Decline from Facial Photographs Using AI

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K07778
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 52010:General internal medicine-related
• Research Institution: The University of Tokyo
• Principal Investigator: KAMEYAMA Yumi 東京大学, 医学部附属病院, 講師 (60505882)
• Co-Investigator(Kenkyū-buntansha): 田中 友規 東京大学, 高齢社会総合研究機構, 特任研究員 (30750343) ; 小島 太郎 東京大学, 医学部附属病院, 講師 (40401111)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 認知症 / AI / 顔写真

Outline of Final Research Achievements

An AI system, which would perform better than the human eye, we examined whether AI facial age reflects cognitive decline and MMSE score. The multiple existing AI models and "face age" AI software (Microsoft azure face API) were assessed.The Alzheimer's disease group (121 people) and normal group participants (Kashiwa cohort study of 117 people) were photographed. We evaluated their faces using multiple AI models and the Microsoft azure face API to examine the relationship between cognitive decline and MMSE.
The AI model called Xception showed high discrimination with a sensitivity of 87.3%, specificity of 94.6%, and a correct answer rate of 92.6%. The scores calculated by the AI model correlated with cognitive decline more strongly in the lower half of the face than in the upper half. The AI azure facial age correlated with "10 later judged perceived age" (r= 0.791, p= 3.88×10-27)　but it would not correlate with MMSE.

Free Research Field

認知症

Academic Significance and Societal Importance of the Research Achievements

認知症は高齢化社会において最も深刻な問題の一つであり、今後の治療戦略においては早期診断がとても重要になっている。簡単で非侵襲的で安価な認知症のスクリーニングが望まれている。我々は、顔だけで認知症をスクリーニングできることを世界で初めて示すことができたが、既存のAI azureモデルでは、顔年齢は認知機能低下を反映していなかった。日本人高齢者の顔年齢AIソフトは、日本人の顔写真で作成する必要があり、また認知機能低下を予測するより良いモデルも今後作成していく必要がある。