How Intra-source Imbalanced Datasets Impact The Performance Of Deep Learning For Covid-19 Diagnosis Using Chest X-ray Images

This study examines the impact of intra-source imbalance (ISI) on deep learning-based methods (DLMs) for COVID-19 diagnosis. The findings show that using an intra-source imbalanced dataset causes a serious training bias, while the deep learning model performs reliably when trained on an intra-source balanced dataset. This study provides clear evidence that intra-source balance is essential for training data to minimize the risk of poor performance of DLMs.