Abstract
Segmenting polyps in colonoscopy images is essential for the earlyidentification and diagnosis of colorectal cancer, a significant cause ofworldwide cancer deaths. Prior deep learning based models such as Attentionbased variation, UNet variations and Transformer-derived networks have hadnotable success in capturing intricate features and complex polyp shapes. Inthis study, we have introduced the DeepLabv3++ model which is an enhancedversion of the DeepLabv3+ architecture. It is designed to improve the precisionand robustness of polyp segmentation in colonoscopy images. We have utilizedThe proposed model incorporates diverse separable convolutional layers andattention mechanisms within the MSPP block, enhancing its capacity to capturemulti-scale and directional features. Additionally, the redesigned decoderfurther transforms the extracted features from the encoder into a moremeaningful segmentation map. Our model was evaluated on three public datasets(CVC-ColonDB, CVC-ClinicDB, Kvasir-SEG) achieving Dice coefficient scores of96.20%, 96.54%, and 96.08%, respectively. The experimental analysis shows thatDeepLabV3++ outperforms several state-of-the-art models in polyp segmentationtasks. Furthermore, compared to the baseline DeepLabV3+ model, our DeepLabV3++with its MSPP module and redesigned decoder architecture, significantly reducedsegmentation errors (e.g., false positives/negatives) across small, medium, andlarge polyps. This improvement in polyp delineation is crucial for accurateclinical decision-making in colonoscopy.