Automated Segmentation and Screening of Gastrointestinal Diseases from Endoscopic Images

Gastrointestinal (GI) diseases, which include Polyps, Ulcerative colitis and Esophagitis are a pressing global health problem. Early diagnosis and accuracy will be key for effective treatment and better patient outcomes. Traditional endoscopic procedures are time-consuming and highly dependent on the expertise of clinicians. Recent advances in AI and deep learning offer solutions that can automate the analysis of images, thereby accelerating and increasing accuracy. This paper presents an automated system to segment and classify GI diseases using deep learning models from endoscopic images. It uses the Kvasir dataset, which includes four classes: Normal, Polyps, Esophagitis, and Ulcerative Colitis. This proposed system utilizes YOLOv11 and U-Net for segmentation and InceptionV3 and EfficientNet for classification. The results proved that the system successfully enhanced the diagnosis accuracy rate of clinicians and lessened clinician workload; thereby, this allowed early GI diseases detection.