SqueezeNet Fusion: Enhancing Rhythmic Heart Disease Classification through Integrated Pattern Mining and Deep Learning

Abstract

The measurement of the heart’s electrical activity, known as Electrocardiogram (ECG), is commonly employed for detecting heart diseases due to its non-invasive and straightforward nature. Studying the fusion of action impulse patterns generated by the specialized cardiac tissues of the heart is a key aspect of the analysis. Carefully scrutinizing the electrical signal produced with each heartbeat allows for the identification of any abnormalities in the heart. The integration of data mining technology in healthcare has significantly enhanced knowledge, making data mining an increasingly preferred option in the medical sector. In this study, a hybridized algorithm is employed for pattern mining to classify heartbeats. The initial step involves pre-processing input data from the ECG signal using a median filter, followed by the extraction of features. These extracted features encompass both medical and statistical aspects. The subsequent phase entails activity pattern mining based on the Optimization Search Algorithm (OSA), a hybridized optimization algorithm. Finally, heartbeats are classified using the rule matrix generated from activity pattern mining. This classification is performed by SqueezeNet, trained through the proposed model Optimization Search Algorithm and SqueezeNet (OSA-SN). Furthermore, performance of this reasearch, such as precision, sensitivity, and specificity with maximal values of 0.91, 0.94, 0.93.