Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{206800,
        author = {Tejashwini K and Aravind Naik},
        title = {An Interpretable Deep Learning Framework for Gastrointestinal Abnormality Detection in Capsule Endoscopy},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {no},
        pages = {497-508},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=206800},
        abstract = {Explainable Artificial Intelligence (XAI) has gained significance in modern medicine, especially in medical imaging applications where transparency and credibility are critical considerations. Gastrointestinal diagnosis using capsule endoscopy yields a massive volume of images; therefore, an automated analysis process would be extremely helpful in this field. This paper proposes a deep-learning-based XAI model for detection of gastrointestinal abnormalities in capsule endoscopy images. The suggested model employs the technique of knowledge distillation along with explainability methods to ensure high prediction accuracy and transparency in decision making. The proposed framework uses a teacher-student network based on ResNets to distill the knowledge of a larger, more complicated teacher model into a smaller, interpretable student model. Grad-CAM is used to make this model interpretable and provides information about the parts of the image that have the greatest impact on the model's prediction. This can help the physician understand the model's decisions. The performance of the suggested model is evaluated using the Capsule Vision 2024 challenge dataset. It comprises different categories of gastrointestinal abnormalities and shows a significant imbalance between classes. The experiment reveals that the suggested distillation model outperforms other baseline approaches and achieves an accuracy of 96.52% and weighted F1-score of 96.37%. Grad-CAM also proves that the model focuses on meaningful regions consistently.},
        keywords = {Capsule Endoscopy, Explainable AI, Knowledge Distillation, Grad-CAM, ResNet, Medical Imaging Technology, Deep Learning, GI Anomalies, Class Imbalance, CNN understanding},
        month = {July},
        }