US 12,456,551 B2
Atrial fibrillation risk prediction system based on heartbeat rhythm signals and application thereof
Qiang Li, Hubei (CN); Fan Lin, Hubei (CN); Peng Zhang, Hubei (CN); and Yuting Chen, Hubei (CN)
Assigned to HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY, Hubei (CN); WUHAN ZHONGKE INDUSTRIAL RESEARCH INSTITUTE OF MEDICAL SCIENCE CO., LTD, Hubei (CN); and TONGJI HOSPITAL, TONGJI MEDICAL COLLEGE, HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY, Hubei (CN)
Appl. No. 17/923,916
Filed by HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY, Hubei (CN); WUHAN ZHONGKE INDUSTRIAL RESEARCH INSTITUTE OF MEDICAL SCIENCE CO., LTD, Hubei (CN); and TONGJI HOSPITAL, TONGJI MEDICAL COLLEGE, HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY, Hubei (CN)
PCT Filed Jan. 21, 2022, PCT No. PCT/CN2022/073248
§ 371(c)(1), (2) Date Nov. 8, 2022,
PCT Pub. No. WO2023/103156, PCT Pub. Date Jun. 15, 2023.
Claims priority of application No. 202111506872.1 (CN), filed on Dec. 10, 2021.
Prior Publication US 2023/0352180 A1, Nov. 2, 2023
Int. Cl. G16H 50/30 (2018.01); A61B 5/00 (2006.01); A61B 5/352 (2021.01); G16H 40/67 (2018.01); G16H 50/20 (2018.01)
CPC G16H 50/30 (2018.01) [A61B 5/352 (2021.01); A61B 5/7203 (2013.01); G16H 50/20 (2018.01); G16H 40/67 (2018.01)] 10 Claims
OG exemplary drawing
 
1. An atrial fibrillation risk prediction system based on heartbeat rhythm signals, comprising:
a processor, wherein the processor is configured to:
extract an RR interval value between two consecutive heartbeats in the heartbeat rhythm signals and obtain the RR interval sequence, wherein the RR interval sequence is divided equally to obtain a plurality of RR interval samples, wherein the RR interval represents a time interval between two successive R waves in an electrocardiogram (ECG);
obtain several heartbeat rhythm signals with two types of labels, including a latent atrial fibrillation (LAF) and a non-atrial fibrillation (NAF), which are input into the processor to obtain several groups of RR interval samples, wherein each group comprises a plurality of RR interval samples, each RR interval sample and the corresponding label constitute a training sample configured to train and obtain an atrial fibrillation (AF) risk prediction model, wherein the AF risk prediction model comprises a cascaded convolutional neural network, a bidirectional long-short-term memory neural network, and a fully connected network; and
execute the AF risk prediction model to input heartbeat rhythm signals into the processor to obtain a plurality of RR interval samples, and input the plurality of RR interval samples into the AF risk prediction model to obtain the output probability corresponding to each RR interval sample, wherein according to different output probability thresholds, an AF risk curve between the proportion of positive samples and the probability thresholds is obtained;
wherein the area under the risk curve is calculated as the AF risk value and used to determine the AF risk of the corresponding heartbeat rhythm signals to be detected.