XX

2020 Fiscal Year Final Research Report

• Project/Area Number: 18K12141
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 90150:Medical assistive technology-related
• Research Institution: Tokyo Medical and Dental University
• Principal Investigator: MIYAJIMA Miho 東京医科歯科大学, 大学院医歯学総合研究科, 助教 (70616177)
• Co-Investigator(Kenkyū-buntansha): 藤原 幸一 名古屋大学, 工学研究科, 准教授 (10642514) ; 山川 俊貴 熊本大学, 大学院先端科学研究部(工), 准教授 (60510419)
• Project Period (FY): 2018-04-01 – 2021-03-31
• Keywords: てんかん発作検知 / 心拍変動解析

Outline of Final Research Achievements

We used autoencoder, which is a type of neural network, for detecting changes in heart rate variability associated with an epileptic seizure. We collected electrocardiogram data from 66 patients with focal epilepsy. The collected ictal data included focal aware seizures and focal impaired awareness seizures as well as focal to bilateral tonic-clonic seizures. We trained an autoencoder model from randomly selected 78 hours of interictal data and validated the model using the rest of episodes. The overall seizure detection sensitivity by 60 sec from clinical seizure onset was 77.6%. The area under the curve (AUC) of 0.92 was achieved. This means the level of detection performance is generally considered meaningful. The false positive rate for an unknown cause was 1.5 per hour. This results suggest that the proposed epileptic seizure detection algorithm demonstrated preferable performance focal seizures including nonconvulsive seizures.

Free Research Field

臨床てんかん学

Academic Significance and Societal Importance of the Research Achievements

心拍データのみを用いて、比較的軽い発作も含め高性能で発作検知が可能なアルゴリズムを構築できた。今後、本アルゴリズムを代表者らの有するウェアラブルてんかんモニタリングシステムのプラットフォームに実装し、プロトタイプ構築および精度検証を目指したい。
本研究の成果は、発作を検出してオンデマンド抑制するclosed-loop型治療や、発作記録に基づき治療方針を示唆する人工知能診療支援システムなど、次世代のてんかんケアにも応用可能性が高い。更に心拍や呼吸の持続モニタリング技術は、近年問題視されているてんかん突然死の病態解明にも役立つことが期待される。