Smart assisted diagnosis solution with multi-sensor Holter

详细信息    查看全文

• 作者：Rongfang Bie^a ; ^{rfbie@bnu.edu.cn}Author Vitae ; Guangzhi Zhang^a ; ^{zgz@mail.bnu.edu.cn}Author Vitae ; Yunchuan Sun^b ; ^{yunch@bnu.edu.cn}Author Vitae ; Shuaijing Xu^a ; ^{xushuaijing@mail.bnu.edu.cn}Author Vitae ; Zhuorong Li^c ; ^{lizhuorong@bnu.edu.cn}Author Vitae ; Houbing Song^d ; ^{Houbing.Song@mail.wvu.edu}Author Vitae
• 关键词：Electrocardiography ; Sensor ; Clustering ; Machine learning ; Holter monitoring
• 刊名：Neurocomputing
• 出版年：2017
• 出版时间：12 January 2017
• 年：2017
• 卷：220
• 期：Complete
• 页码：67-75
• 全文大小：1883 K
• 卷排序：220

文摘

Cardiovascular disease has become an increasingly serious threat to human health. Holter monitoring is essential in the prevention and treatment of cardiovascular disease. When combined with a variety of sensors, a traditional Holter becomes a mobile health device. Based on Holter data and sensor data, this paper proposes a supplementary diagnosis and treatment program. The solution consists of three main steps: I. Perform segmentation of ECG (Electrocardiography) data and conduct feature extraction. Build the personalized ECG templates, apply factor graph and max-sum algorithm for precise template matching, and realize the feature extraction and representation of ECG data. II. Use the action sensor data for action classification and identification. As an important factor of health monitoring, body movements are categorized as resting, walking, going upstairs and downstairs, flipping and sudden change. The improved classification algorithm achieves high accuracy identification. III. Combine ECG data and action data for clustering analysis. The proposed solution improves the affinity propagation algorithm and allow doctors to supervise the clustering procedure. A knowledge matrix is introduced accordingly, thus achieving an iterable clustering optimization. With the help of these components, the innovative approach is able to integrate the Holter data and sensor data, meanwhile, doctors are encouraged to participate in the process of clustering algorithm. Our system is capable of not only assisting doctors in quickly determining the most valuable information, but also of building a personalized private repository for patients. The experimental results indicate that the proposed system is an efficient, accurate, and interactive auxiliary diagnostic and a therapeutic support tool.