Mitigating Class Imbalance in Colorectal Histopathological Image Classification Using Data Synthesis and Adaptive Loss

Histopathological image classification in colorectal cancer (CRC) is hindered by class imbalance, where limited data for some classes and overlapping features lead to degraded performance. Traditional methods struggle to adequately augment minority classes or resolve these uncertainties. To alleviate these problems, we propose a novel data synthesis and adaptive loss-based histopathological image classification framework that integrates two components: 1) a global context generative adversarial network (GCGAN) to generate realistic images for under-represented classes, thereby enriching dataset diversity; 2) an adaptive focal loss (AFL) to dynamically adjust the loss function and mitigate feature overlap. Extensive experiments on the CRC dataset demonstrate the effectiveness of our approach, outperforming other state-of-the-art (SOTA) methods. Additionally, combining GCGAN and AFL benefits these SOTA methods, with performance improvements on cervical (CC) and breast cancer (BC) datasets.