Detection of Myocardial Infarction using Spectral Textures through ECG

Myocardial infarction (MI), commonly known as a heart attack, can cause nonrecoverable impairment to the cardiac muscles, which may even result in fatality. Therefore, it is crucial to diagnose MI accurately and reliably. This article proposes a novel framework for detecting MI using a 12-channel Electrocardiogram (ECG) signal. The method filters the signal to eliminate motion artifacts and high-frequency noise, followed by the extraction of novel spectral texture patterns and Mel frequency cepstral coefficients (MFCCs) from all ECG signal channels. These features are combined to form a comprehensive feature representation. ANOVA is then applied to reduce feature dimensions, and various state-of-the-art classification methods such as support vector machines (SVM), Artificial Neural Networks (ANN), AdaBoost (AB), and K-nearest neighbors (KNN) are utilized. The AB classifier achieved the highest performance with 98.7% accuracy and 98.4% recall using 10-fold cross-validation. The proposed method is reliable and holds the potential for clinical adoption in hospitals.