沃夫巴金森懷特症候群（WPWS）是一種常見的心臟疾病，此種病人的心房和心室之間多了一條副傳導路徑（Accessory pathway）易導致心跳過速。臨床醫師可以透過心電圖（electrocardiogram, ECG）診斷受測者是否為 WPWS；若受測者為 WPWS，接著判定心電圖上不同導程的三角波（Delta wave）, 藉以判斷副傳導路徑的位置以方便進行後續的手術治療；因此偵測三角波為一項重要的研究。以往自動診斷沃夫巴金森懷特症候群系統的演算法需要判斷心電圖中的三角波型態(Positive、Negative、Isoelectric)、PR 間期（PR interval）與 QRS 複合波（QRS complex）的持續時間。然而，如此的演算法是困難且耗時，成果的準確率也不盡理想。

本論文提出基於卷積神經網路（CNN）模型的沃夫巴金森懷特症候群之三角波自動偵測演算法；先判斷心電圖資料是否為 WPWS，再進一步偵測三角波的起始位置，並在後處理視覺化三角波。本論文以臺北榮民總醫院心電圖資料集為實驗測試，探討一維（1D）與二維（2D）心電圖資料、平衡與不平衡資料量等情形下三角波自動偵測演算法的相關議題。實驗結果證實，本研究所提的 2D CNN 診斷模型準確率可達到99.52%，且以 1D CNN 偵測 delta wave 起始位置與真實標籤（Ground truth）的平均誤差為 4.13ms。

Wolff-Parkinson-White syndrome (WPWS) is a common heart disease in which an accessory pathway between the atrium and the ventricle is prone to tachycardia. WPWS diagnosis is conducted using an electrocardiogram (ECG). If WPWS is diagnosed, then the delta wave is detected using various leads to determine the position of the accessory pathway to facilitate further surgical treatment. The detection of delta waves is thus important. Existing algorithms for automatically diagnosing WPWS need to determine the delta wave pattern (positive, negative, or isoelectric), the duration of the PR interval, and the QRS complex in the ECG. However, such algorithms are difficult to implement and time-consuming, and the accuracy of their results is unsatisfactory.

This paper proposes an automatic delta wave detection algorithm based on the convolutional neural network (CNN) model. ECG data are first classified as WPWS or non-WPWS. Then, the onset position of the delta wave is detected. Finally, the delta wave is visualized in post-processing. We used the Taipei Veterans General Hospital ECG data set as an experimental data set to explore issues related to the automatic delta wave detection algorithm for one-dimensional (1D) and two-dimensional (2D) ECG data and balanced and imbalanced data. The experimental results show that the accuracy of the proposed 2D CNN can reach 99.52%, and that the average error between the delta wave onset position obtained using the proposed 1D CNN and the ground truth is 4.13 ms.

[1] Mahamat, H. A., Jacquir, S., Khalil, C., Laurent, G., & Binczak, S. (2016, August). Wolff-Parkinson-White (WPW) syndrome: The detection of delta wave in an electrocardiogram (ECG). In 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) (pp. 3809-3812). IEEE.
[2] Mahamat, H. A., Jacquir, S., Khalil, C., Laurent, G., & Binczak, S. (2016, September). Automatic detection of the Wolff-Parkinson-White syndrome from electrocardiograms. In 2016 Computing in Cardiology Conference (CinC) (pp. 417-420). IEEE
[3] Dar, M. A., SHEIKH, S. H., ABID, A. R., & MALLICK, N. H. (2012). Localization of Accessory Pathways According to AP Fitzpatrick Ecg Criteria in Patents with Wolff-Parkinson-White Syndrome in Our Population. Pakistan Heart Journal, 41(3-4).
[4] Fox, D. J., Klein, G. J., Skanes, A. C., Gula, L. J., Yee, R., & Krahn, A. D. (2008). How to identify the location of an accessory pathway by the 12-lead ECG. Heart Rhythm, 5(12), 1763-1766.
[5] Taguchi, N., Yoshida, N., Inden, Y., Yamamoto, T., Miyata, S., Fujita, M., ... & Murohara, T. (2014). A simple algorithm for localizing accessory pathways in patients with Wolff-Parkinson-White syndrome using only the R/S ratio. Journal of Arrhythmia, 30(6), 439-443.
[6] Durrer, D., Schoo, L., Schuilenburg, R. M., & Wellens, H. J. J. (2000). The role of premature beats in the initiation and the termination of supraventricular tachycardia in the Wolff-Parkinson-White syndrome. In Professor Hein JJ Wellens (pp. 1-20). Springer, Dordrecht.
[7] Gollob, M. H., Green, M. S., Tang, A. S. L., Gollob, T., Karibe, A., Hassan, A. S., ... & Bachinski, L. L. (2001). Identification of a gene responsible for familial Wolff–Parkinson–White syndrome. New England Journal of Medicine, 344(24), 1823-1831.
[8] Maron, B. J., Friedman, R. A., Kligfield, P., Levine, B. D., Viskin, S., Chaitman, B. R., ... & Soliman, E. Z. (2014). Assessment of the 12-lead ECG as a screening test for detection of cardiovascular disease in healthy general populations of young people (12–25 years of age) a scientific statement from the American Heart Association and the American College of Cardiology. Circulation, 130(15), 1303-1334.
[9] Yochum, M., Renaud, C., & Jacquir, S. (2017, April). Improvement of the Detection of the QRS Complex, T and P Waves in an Electrocardiogram Signal using 12 Leads versus 2 Leads. In COMPLEXIS (pp. 73-78).
[10] Jackman, W. M., Wang, X., Friday, K. J., Roman, C. A., Moulton, K. P., Beckman, K. J., ... & Margolis, P. D. (1991). Catheter ablation of accessory atrioventricular pathways (Wolff–Parkinson–White syndrome) by radiofrequency current. New England Journal of Medicine, 324(23), 1605-1611.
[11] Calkins, H., Langberg, J., Sousa, J., El-Atassi, R., Leon, A., Kou, W., ... & Morady, F. (1992). Radiofrequency catheter ablation of accessory atrioventricular connections in 250 patients. Abbreviated therapeutic approach to Wolff-Parkinson-White syndrome. Circulation, 85(4), 1337-1346.
[12] Jie-Zhi Cheng, Dong Ni, Yi-Hong Chou, Jing Qin, Chui-Mei Tiu, Yeun-Chung Chang, Chiun-Sheng Huang, Dinggang Shen and Chung-Ming Chen, “Computer-Aided Diagnosis with Deep Learning Architecture: Applications to Breast Lesions in US Images and Pulmonary Nodules in CT Scans,” Scientific Reports, Vol. 6, No. 24454, 2016.
[13] Sun, W., Zheng, B., & Qian, W. (2016, March). Computer aided lung cancer diagnosis with deep learning algorithms. In Medical imaging 2016: computer-aided diagnosis (Vol. 9785, p. 97850Z). International Society for Optics and Photonics.
[14] Arabasadi, Z., Alizadehsani, R., Roshanzamir, M., Moosaei, H., & Yarifard, A. A. (2017). Computer aided decision making for heart disease detection using hybrid neural network-Genetic algorithm. Computer methods and programs in biomedicine, 141, 19-26.
[15] LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. nature, 521(7553), 436.
[16] Schmidhuber, J. (2015). Deep learning in neural networks: An overview. Neural networks, 61, 85-117.
[17] Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., ... & Sánchez, C. I. (2017). A survey on deep learning in medical image analysis. Medical image analysis, 42, 60-88.
[18] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems (pp. 1097-1105).
[19] Ristovski, A., Guseva, A., Gusev, M., & Ristov, S. (2016, May). Visualization in the ECG QRS detection algorithms. In 2016 39th International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO) (pp. 202-207). IEEE.
[20] LeCun, Y., Boser, B., Denker, J. S., Henderson, D., Howard, R. E., Hubbard, W., & Jackel, L. D. (1989). Backpropagation applied to handwritten zip code recognition. Neural computation, 1(4), 541-551.
[21] Wu, Y., Yang, F., Liu, Y., Zha, X., & Yuan, S. (2018). A comparison of 1-D and 2-D deep convolutional neural networks in ECG classification. arXiv preprint arXiv:1810.07088.
[22] Saxe, J., & Berlin, K. (2015, October). Deep neural network based malware detection using two dimensional binary program features. In 2015 10th International Conference on Malicious and Unwanted Software (MALWARE) (pp. 11-20). IEEE.
[23] Willmott, C. J., & Matsuura, K. (2005). Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate research, 30(1), 79-82.
[24] Lui, H. W., & Chow, K. L. (2018). Multiclass classification of myocardial infarction with convolutional and recurrent neural networks for portable ECG devices. Informatics in Medicine Unlocked, 13, 26-33.
[25] Baloglu, U. B., Talo, M., Yildirim, O., San Tan, R., & Acharya, U. R. (2019). Classification of myocardial infarction with multi-lead ECG signals and deep CNN. Pattern Recognition Letters, 122, 23-30.