智慧型心房顫動病患診斷決策支援系統設計之研究

Study on Design of Intelligent Diagnostic Decision Support Systems in Atrial Fibrillation Patients

張祐禎(Yiu-Jen Chang)

指導教授 ： 何文獻

高雄醫學大學/健康科學院/醫務管理暨醫療資訊學系碩士班/碩士(2018年)

摘要

心房顫動是一種陣發性且多數人發作時無明顯症狀的心臟疾病，其未發作時的心電圖與正常人無明顯差異，導致難以察覺以及診斷，而心房顫動若未及早治療，容易導致病情惡化且提高中風的可能性。
本論文使用Physionet網站(https://physionet.org)所提供的心房顫動心電圖資料庫，經過訊號濾波處理後，利用多種智慧化參數擷取方法得到心電圖中的特徵，並結合P波形態學參數與心率變異參數進行人工智慧模型建模。
本論文除了比較多種人工智慧模型的表現，也列出參數擷取方法的優劣，本論文採用Stacking集成學習法結合各種不同模型，最後得到92%的正確率(accuracy)、88%的敏感度(sensitivity)、96%的特異度(specificity)、95.7%的陽性預測值(positive predictive value)、88.9%的陰性預測值(negative predictive value)、0.9231的F1評分(F1 score)以及0.911的AUROC值(area under receiver operating characteristic curve)，並實作出一智慧型心房顫動診斷決策支援系統，使其能在臨床上輔助醫師診斷心房顫動。

關鍵字

心房顫動 ； 人工智慧 ； 決策支援系統

並列摘要

Atrial fibrillation is a paroxysmal heart disease with no obvious symptoms when most people attack. There is no obvious difference between the patients and the normal people on ECG when the patients are not under attacked, which makes it difficult to detect and diagnose. If the atrial fibrillation is not treated early, it will possibly lead to deterioration of the disease and increase the possibility of stroke.
This study used Physionet's atrial fibrillation electrocardiogram database (AFPDB), after signal filtering, using a variety of intelligent parameter extraction methods to extract ECG features, combined with P-wave morphological parameters and heart rate variation parameters for artificial intelligence modeling.
In addition to comparing of performance with different models, the advantages and disadvantages of the parameter extraction methods are also listed. Finally, this study used the Stacking method combined with various models to build an intelligent decision support system for atrial fibrillation diagnosis, and achieved accuracy of 92%, sensitivity of 88%, specificity of 96%, positive predictive value of 95.7%, negative predictive value of 88.9%, F1 score of 0.9231 and area under receiver operating characteristic curve (AUROC) of 0.911. It can help clinicians diagnose atrial fibrillation.

並列關鍵字

Atrial Fibrillation ； artificial intelligence ； decision support systems

參考文獻

[1] Andrikopoulos, G. K., Dilaveris, P. E., Richter, D. J., Gialafos, E. J., Synetos, A. G., & Gialafos, J. E. (2000). Increased variance of P wave duration on the electrocardiogram distinguishes patients with idiopathic paroxysmal atrial fibrillation. Pacing and Clinical Electrophysiology, 23(7), 1127-1132.

[2] Asgari, S., Mehrnia, A., & Moussavi, M. (2015). Automatic detection of atrial fibrillation using stationary wavelet transform and support vector machine. Computers in Biology and Medicine, 60, 132-142.

[3] Aytemir, K., ÖZER, N., Atalar, E., Sade, E., AKSÖYEK, S., ÖVÜNÇ, K., Oto, A., ÖZMEN, F., & Kes, S. (2000). P Wave Dispersion on 12‐Lead Electrocardiography in Patients with Paroxysmal Atrial Fibrillation. Pacing and Clinical Electrophysiology, 23(7), 1109-1112.

[4] Benjamin, E. J., Levy, D., Vaziri, S. M., D'agostino, R. B., Belanger, A. J., & Wolf, P. A. (1994). Independent risk factors for atrial fibrillation in a population-based cohort: the Framingham Heart Study. Jama, 271(11), 840-844.

[5] Berger, R. D., Akselrod, S., Gordon, D., & Cohen, R. J. (1986). An efficient algorithm for spectral analysis of heart rate variability. IEEE Transactions on biomedical engineering(9), 900-904.