Gastro Intestinal Disease Detection Using RegNet, MobileNetV2, EfficientNetB2, and ResNet-152 with Transformers

The early detection of gastrointestinal (GI) diseases is crucial for effective treatment and improved patient outcomes. The application of advanced deep learning techniques has shown promise in detecting gastrointestinal diseases through the analysis of medical images. This study presents a novel hybrid deep learning approach by integrating RegNet with Swin Transformer, aiming to enhance the accuracy and contextual understanding of medical endoscopic images. Additionally, other model combinations-MobileNetV2 + Swin Transformer, EfficientNetB2, and ResNet152 + Vision Transformer-were also evaluated to benchmark performance. We fine-tuned our proposed model, RegNet with Swin Transformer. This combination allows us to recognize important details in medical images while also understanding the overall context, which is crucial for accurate diagnoses. By merging RegNet’s ability to recognize patterns and Swin Transformer’s skill in understanding context, our system can effectively identify key details in medical images. Our method achieved an accuracy of $90 \%$, demonstrating that this approach can improve the early detection of gastrointestinal disorders, ultimately leading to better care and outcomes for patients. This research demonstrates that combining scalable CNN backbones with hierarchical vision transformers can significantly improve the detection and classification of GI diseases. The findings support the integration of such hybrid models into clinical decision-making tools.