SLS-cGAN: An Automated Skin Lesions Segmentation Approach Based on cGAN

With the increase in dermatological lesions, there is a growing need for accurate and efficient computer-based image analysis and diagnostic methods. Accurately segmenting these lesions’ pigmented areas is a crucial part of this process. This study introduces Skin lesion segmentation based on the Conditional Generative Adversarial Networks (SLS-cGAN) model for segmenting skin lesions in medical imaging. This model combines the capabilities of UNets and GANs to improve the quality of the model in the medical image segmentation process. Upon evaluation using the HAM10000 dataset, the proposed model demonstrated significant effectiveness, achieving an accuracy of 0.972, sensitivity of 0.943, Dice coefficient of 0.952, and Intersection over Union of 0.915. The SLS-cGAN model shows superiority over other methods outlined in the literature.