Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm
Also available in printed version
| المؤلف الرئيسي: | |
|---|---|
| مؤلفون آخرون: | |
| التنسيق: | Master's thesis |
| اللغة: | الإنجليزية |
| منشور في: |
Universiti Teknologi Malaysia
2025
|
| الموضوعات: | |
| الوصول للمادة أونلاين: | https://utmik.utm.my/handle/123456789/39813 |
| Abstract | Abstract here |
| _version_ | 1855605631041404928 |
|---|---|
| author | Kong, Pang Seng |
| author2 | Anita Ahmad, supervisor |
| author_facet | Anita Ahmad, supervisor Kong, Pang Seng |
| author_sort | Kong, Pang Seng |
| description | Also available in printed version |
| format | Master's thesis |
| id | utm-123456789-39813 |
| institution | Universiti Teknologi Malaysia |
| language | English |
| publishDate | 2025 |
| publisher | Universiti Teknologi Malaysia |
| record_format | DSpace |
| record_pdf | Restricted |
| spelling | utm-123456789-398132025-03-05T18:05:14Z Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm Kong, Pang Seng Anita Ahmad, supervisor Electrical engineering Also available in printed version Atrial Fibrillation (AF) is described as the most common heart condition which will cause health problems such as stroke, heart failure and etc. AF patients will increase threefold by 2050 worldwide. The electrocardiogram (ECG) machine cannot recognize the AF signal and normal signal automatically. Then, clinical observation of ECG waveform needs experienced person to observe and take long hours. So, the reliability of the technology can be used as a target for catheter ablation of atrial fibrillation. In the AF signal, there are more R peaks and the rhythm is irregular. The major objective for the design of the detection method is to analyze a practical method for analyzing the complexity of ECG signal. As ECG signal contains high frequency noise, baseline wander and interference, filtering is a must in analyzing ECG signal. No study shows the comparison between Pan-Tompkins algorithm and Discrete Wavelet Transform (DWT) algorithm in terms of R peak detection. Then, the result indicates that the DWT algorithm achieves better sensitivity of R Peak Detection compared to previous papers with 99.86% and 100% in the MIT-BIH Arrhythmia and Atrial Fibrillation databases, respectively. Furthermore, a modified windowing algorithm after DWT algorithm shows significant improvements in PQRST peak detection in terms of sensitivity and specificity compared to the windowing algorithm and other PQRST detection methods, with 100% and 94.5%, respectively. The contribution in modified windowing algorithm is eliminating the first and last R peaks which do not show full QRS complex to decrease the error detection and adjusting the window size based standard ECG waveform. This study also aims to improve the accuracy of ECG diagnosis using several types of machine learning algorithms. The features extracted after DWT and modified windowing algorithm for signal classification are RR Interval, heart rate, QRS complex, coefficient of variance (CV), normalised root mean square of successive difference (nRMSSD), and peak frequency of power spectral density (PSD). The fixed threshold values and machine learning methods are compared in the classification process to increase accuracy. However, the experiment results reveal that the machine learning classifier is more accurate than fixed threshold values in the diagnosis of normal and AF signals more. Then, the weighted K-nearest neighbour (KNN) classifier with DWT algorithm and modified windowing algorithm in holdout validation achieves the best performance in AF classification with sensitivity, specificity, and accuracy of 90%, 100% and 92.31% respectively. On this basis, it is recommended that classifying ECG signals using machine learning methods will assist in improving the research’s accuracy. In comparison to studies that utilise many features to identify AF signals or normal signals, the suggested method is simpler and only uses six features, which are fewer features. Although the accuracy result is a little less accurate, it still performs well enough to detect an AF signal in a short interval of ECG. Compared to other studies, the proposed technique has the highest specificity. Further research is needed to involve testing the proposed algorithm on a large database for better accuracy. This algorithm will be involved in hardware and implemented in real time for detecting AF ECG patients. fahmimoksen UTM 153 p. Thesis (Sarjana Falsafah (Kejuruteraan Elektrik)) - Universiti Teknologi Malaysia, 2023 2025-03-05T04:54:51Z 2025-03-05T04:54:51Z 2023 Master's thesis https://utmik.utm.my/handle/123456789/39813 vital:156232 valet-20240310-155620 ENG Closed Access UTM Complete Unpublished Completion application/pdf Universiti Teknologi Malaysia |
| spellingShingle | Electrical engineering Kong, Pang Seng Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm |
| thesis_level | Master |
| title | Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm |
| title_full | Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm |
| title_fullStr | Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm |
| title_full_unstemmed | Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm |
| title_short | Atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm |
| title_sort | atrial fibrillation detection based on electrocardiogram features using modified windowing algorithm |
| topic | Electrical engineering |
| url | https://utmik.utm.my/handle/123456789/39813 |
| work_keys_str_mv | AT kongpangseng atrialfibrillationdetectionbasedonelectrocardiogramfeaturesusingmodifiedwindowingalgorithm |