A ROBUST DEEP LEARNING APPROACH FOR AUTOMATED DETECTION AND CLASSIFICATION OF GASTROINTESTINAL TRACT DISEASES IN ENDOSCOPY IMAGES

Abstract

Gastrointestinal (GI) disorders represent a major global health burden, and accurate early diagnosis is critical for effective treatment and improved patient outcomes. Endoscopic imaging is a primary diagnostic tool; however, manual interpretation is subjective and prone to errors. This study presents a deep learning-based approach for automated classification of GI diseases using EfficientNet-B1 and ResNet-101 models on the Kvasir-V2 dataset. Image preprocessing and supervised learning techniques were applied to enhance feature extraction and model generalization. Experimental results indicate that EfficientNet-B1 outperformed ResNet-101, achieving an accuracy of 94.71% with high precision, recall, and F1-score values. The proposed framework demonstrates the effectiveness of deep learning in supporting gastroenterologists by providing reliable and accurate diagnostic predictions. This approach has the potential to improve clinical workflows and reduce diagnostic variability in endoscopic examinations.