Multiple Skin Lesions Diagnostics Using Cascaded Deep Convolutional Networks

In recent years, skin cancer has emerged as one of the most prevalent types of cancer in the human population. Skin lesion classification plays a crucial role in dermatology, aiding in the early objective of this study to employ deep learning techniques, including ResNet152v2, InceptionV3, and VGG16, to classify skin lesions accurately. Additionally, the study includes the segmentation of skin lesions using Deep Res U-Net and Res U-Net. Experiments are conducted to achieve optimal evaluation metrics and fine-tuned model parameters using different regularization methods and hyperparameters. The results demonstrated that modifying Res U-Net led to higher accuracy and outperformed other approaches. This emphasizes the potential of using neural networks with fewer parameters and lower computational weight for improved skin lesion classification and segmentation. Furthermore, the deep learning models employed in this study showcased their capability to capture intricate features and patterns within skin lesion images, enabling a more precise and efficient diagnosis process. By implementing these deep learning techniques, a significant enhancement is anticipated in the current state-of-the-art for skin lesion classification and segmentation. This study underscores the importance of utilizing large and diverse datasets for training deep learning models and employing appropriate pre-processing techniques like data augmentation and normalization. Moreover, the study highlights the significance of fine-tuning pre-trained models to tailor them to skin lesion classification and segmentation. Overall, this research emphasizes the potential benefits of advanced deep-learning methods in advancing the field of dermatology.