Performance improvement of deep learning based multi-class ECG classification model using limited medical dataset

Medical data often exhibit class imbalance, which poses a challenge in classification tasks. To solve this problem, augmentation techniques are used to balance the data. However, methods not always reliable when applied bio-signals. Also, bio-signal such as ECG has limitation of standardized or normalized methods. The present study endeavors tackle difficulties associated with imbalanced and limited medical datasets. Our is compare different approaches for addressing imbalance datasets, evaluate efficacy various models overcoming these challenges. end, three experiments configurations were considered, that is, change loss function (Experiment A), amount each B), grouping C). Inception-V3 was our main model, dataset groups utilized: an large data, balanced subclass bundled small We propose improved method using focal classification. F1 score 0.96 Inception net 0.86 environment same ratio.