| Title |
EKG signalų tyrimas hipertenzijai prognozuoti / |
| Translation of Title |
ECG signal analysis for the prediction of hypertension. |
| Authors |
Trakšelis, Kristupas |
| Full Text |
|
| Pages |
67 |
| Keywords [eng] |
ECG ; hypertension ; signal processing ; convolutional neural networks |
| Abstract [eng] |
This study provides an overview of the problem of hypertension and ECG analysis. It analyses existing methods for the prediction of hypertension and the results obtained. The analysis highlights the problem that there is no automated machine learning method that can predict hypertension from extremely short signals using only data from a single electrode. In this work, a hypertension prediction algorithm is proposed and implemented based on SVM and oriented histogram method. The first proposal is the filtering of the electrocardiogram signal using a low-pass Butterworth filter. An algorithm from the biopeaks package is then used to segment the signal into heartbeats. The third proposal is a convolutional neural network architecture called Net. The experiments compared Net, ResNet50, ResNet152 and MobileNetV2 neural networks. The SVM algorithm was also included in the comparisons. The experiments showed a prediction accuracy of 92 % for hypertension. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
Lithuanian |
| Publication date |
2024 |
