Published Paper Details:

Medical image segmentation, particularly for polyps in gastrointestinal images, presents significant challenges because of variations in color and morphology observed during colonoscopy imaging. In our study, we focused on this crucial area, utilizing a dataset comprising gastrointestinal polyp images. We employed advanced deep learning techniques, specifically FCN, Dual U-Net with ResNet Encoder, U-Net, and Unet_Resnet, to tackle the segmentation task effectively. To enhance the effectiveness of these algorithms, we incorporated data augmentation techniques, which involve creating modified versions of the initial images to enrich the training dataset. We evaluated the effectiveness of these algorithms using standard metrics like Dice Similarity Coefficient (DSC) and Intersection Over Union (IOU), which quantify the correctness of segmentation compared to ground truth. Among the models tested, the Dual U-Net with ResNet Encoder emerged as the most successful, achieving a DSC of 0.87 and IOU of 0.80. This model outperformed other approaches, including U-Net, FCN, and Unet_Resnet, demonstrating its superior performance in segmenting gastrointestinal polyp images.