U-NET BASED POLYP SEGMENTATION ON KVASIR-SEG DATASET: PERFORMANCE EVALUATION AND COMPARISON WITH STATE-OF-THE-ART METHODS

Arabboev, Mukhriddin, Begmatov, Shohruh, Bobojanov, Sukhrob

Муҳандислик ва Иқтисодиёт · 2025-yil

Annotatsiya

Medical image segmentation is a fundamental task in computer-assisted diagnosis, enabling accuratedelineation of anatomical structures and pathological regions. In gastrointestinal endoscopy, precise segmentation ofpolyps is crucial for early detection and treatment planning. This study presents a U-Net-based approach for polypsegmentation using the publicly available Kvasir-SEG dataset. The proposed model was trained for 100 epochs withcustom metrics including Dice coefficient, Intersection over Union (IoU), precision, and recall. Experimental resultsdemonstrate that the model achieved an average Dice score of 0.9449, IoU of 0.9084, precision of 0.9404, and recall of0.9584, outperforming several recent state-of-the-art methods. Comparative analysis with existing approaches confirmsthe effectiveness of the vanilla U-Net architecture when combined with careful preprocessing and hyperparameter tuning.The findings highlight U-Net's continued relevance in medical image segmentation tasks and suggest directions for futurework, including integration with attention mechanisms and transformer-based architectures.

Maqola ma’lumotlari
MualliflarArabboev, Mukhriddin, Begmatov, Shohruh, Bobojanov, Sukhrob
JurnalМуҳандислик ва Иқтисодиёт
Nashr sanasi2025-12-01
Jild3
Son12
TilIngliz

Kalit so‘zlar

U-Net, Polyp Segmentation, Kvasir-SEG, Deep Learning, Medical Image Segmentation, Convolutional Neural Networks.

Ilmiy soha

Муҳандислик ва Иқтисодиёт jurnalidan boshqa maqolalar

Муҳандислик ва Иқтисодиёт — barcha maqolalar