Machine learning and deep learning in healthcare: advancing cardiac arrhythmia classification in healthcare analytics

Abstract

Cardiac arrhythmias, a global leading disease cause, necessitate rapid, efficient diagnosis. Shifting from traditional manual electrocardiogram analysis to machine learning approaches offers enhanced efficiency and accuracy in detection. However, literature research has shown that long training times and a lack of practical suitability have made implementation difficult to date. Three prototypes were developed and tested; the results were then used to optimize the most promising model further. The optimized CNN achieved an overall classification accuracy of 98.53%. The results are tested for their applicability in a practical context, evaluated, and compared against existing approaches, resulting in above-average classification outcomes.