Explainable hybrid AlexNet-EfficientViT architecture for accurate gallbladder disease diagnosis

Abstract

Due to their various clinical symptoms and the difficulties in using medical imaging, gallbladder illnesses such inflammation, gallstones, and malignancy are still difficult to diagnose. However, intelligent systems that can automatically analyze medical images and produce therapeutically significant insights have been made possible by recent developments in image and signal processing. In order to significantly improve diagnostic accuracy, we present a novel diagnostic framework in this paper that combines Explainable AI (XAI), deep neural networks, and sophisticated optimization approaches. Transformer-based architectures are employed to detect and classify subtle imaging patterns, while the Lion optimizer contributes to both precision and computational efficiency. To address one of the key barriers to clinical implementation, we incorporate XAI methods that provide transparency and interpretable explanations for the model’s predictions. Achieving an outstanding accuracy of 99.86%, the proposed framework establishes a new benchmark in medical image analysis and demonstrates strong potential as a practical clinical decision support tool, effectively bridging the gap between state-of-the-art computational methods and real-world healthcare applications.