Abstract

The smart capsule endoscopy has emerged as a transformative tool for non-invasive gastrointestinal diagnostics. It presents a novel framework for real-time ulcer monitoring and targeted drug delivery using an AI-enhanced capsule system and leveraging the combined strength of EffivientNet-B7 and vision transformer (ViT). Our model achieves a high classification accuracy of 98.7% in detecting gastric ulcers from endoscopic imagery. The capsule integrates onboard imaging and wireless transmission, and a micro-reservoir drug release mechanism. To enable autonomous therapeutic intervention upon ulcer detection. The hybrid deep learning architecture enhances feature extraction and spatial attention. To ensure robust performance across varied ulcer morphologies and lighting conditions. This approach not only improves diagnostic precision but also introduces a closed-loop treatment paradigm. Paving the way for intelligent and patient-centric gastrointestinal care.