Development of diagnosis support system for Parkinson's disease by voice analysis

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K07428
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 52020:Neurology-related
• Research Institution: Fujita Health University
• Principal Investigator: Ito Shinji 藤田医科大学, 医学部, 教授 (40572079)
• Co-Investigator(Kenkyū-buntansha): 渡辺 宏久 藤田医科大学, 医学部, 教授 (10378177) ; 加藤 昇平 名古屋工業大学, 工学(系)研究科(研究院), 教授 (70311032)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: パーキンソン病 / 音声解析 / 人工知能 / 機械学習 / 診断支援システム / 発話特徴 / 早期診断 / 遠隔医療

Outline of Final Research Achievements

Parkinson's disease (PD) is a common disease affecting people aged 65 and older, with a prevalence of approximately 1 in 100 people, and drug therapy since the early stage of the disease is effective. To facilitate early diagnosis at primary medical institutions and early treatment based on differential diagnosis by neurologists, we built a system that uses AI to capture speech characteristics specific to PD. We established a method to distinguish between PD and healthy subject with an F-value of 0.906 using only speech of 3 everyday words from 116 PD patients and 94 healthy subjects collected in the registry of the department of neurology at our hospital. Furthermore, we have been detected differences in the speech characteristics between PD, PD-related diseases, and healthy subjects. In addition, we established a basis for clarifying the correlation between motor symptoms, cognitive function, autonomic dysorder, various biomarkers, etc. and speech characteristics in PD.

Free Research Field

脳神経内科学

Academic Significance and Societal Importance of the Research Achievements

本研究では専門的な診察手技や、複雑・高価な機器を要さず収集・評価できる”音声”を用いて、早期にPDの可能性に気づき、健常者との鑑別に役立つ簡便な診断補助システムを構築した。またPDとPD関連疾患との鑑別における有用性も示された。今後さらにPD患者の歩行や日常動作における運動障害悪化や、認知障害進行に伴う音声変化の特徴を捉えうる見込みで、地域医療機関からの音声情報のfeedbackが、専門医による治療強化時機の把握に役立つ可能性がある。さらに音声解析という簡便な手法は、地域の現場において医師、看護師、療法士及び介護従事者を含む全ての関連職種間で有用な診療情報を共有するモデルを提示した。