2024-03-28T18:08:40Zhttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_oaipmhoai:ipsj.ixsq.nii.ac.jp:002117682023-04-27T10:00:04Z01164:04402:10541:10615
CNNを用いた心電図の自動診断支援に関する研究Research on CNN-based automatic diagnosis support for electro-cardiogramjpn知的情報処理http://id.nii.ac.jp/1001/00211662/Technical Reporthttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=211768&item_no=1&attribute_id=1&file_no=1Copyright (c) 2021 by the Institute of Electronics, Information and Communication Engineers This SIG report is only available to those in membership of the SIG.法政大学理工学部本田, 祥之藤井, 章博清水, 宏泰心電図の自動診断について,既存の研究では,心電図データを時系列のある画像として扱う.このため,自動診断を行う場合,計算量が大きくなり,入力可能な心電図の時間も短くなる場合が多い.本研究では,心電図波形を 2 次元波形画像として扱うことにより,機械学習を用いた心電図波形の診断支援手法を提案する.これにより,心電図波形の前処理にかかる計算量を軽減し,入力できる心電図の時間についての改善も行っている.In this study, we proposed a method to determine normal and abnormal ECG waveforms and to classify them into multiple classes. In this method, the ECG is not treated as time-series waveform data, but as two-dimensional waveform image data, and classified. We used four classes from the MIT-BIH arrhythmia database: N class, VEB class, SVEB class, and F class, and classified them using CNN, VGG16, VGG19, and ResNet. The maximum accuracy was achieved with VGG16 and VGG19. 99.22% of sensitivity, 99.52% of specificity, and 99.3% of accuracy were obtained.AA11135936研究報告知能システム(ICS)2021-ICS-2033162021-06-252188-885x2021-06-17