A blockchain-based deep learning system with optimization for skin disease classification

Recently, there has been a rapid advancement in the usage of smart healthcare systems, especially for the diagnosis of diseases based on imaging technologies, although providing security is challenging. In this scenario, Blockchain is found to be an innate secure method for sharing information based on automated cryptographic techniques for ensuring uniformity of data. In this research, a Blockchain-based system is devised for classifying skin diseases based on an optimized deep-learning method. The skin illness is classified based on the skin images, which are stored in Blockchain in a data format. Here, classification is accomplished using a LeNet, whose structural optimization is done by the proposed Transit Circle Inspired Optimization. Further, a novel segmentation technique is developed for excerpting the lesions from the surrounding skin. Here, DeepJoint segmentation modified by the inclusion of Kumar-Hassebrooks distance is utilized for segmentation. Thereafter, data augmentation and feature extraction are executed, and the features determined are subjected to the LeNet for classification. Here, a novel technique named Transit Circle Inspired Optimization is developed for modifying the weights of the LeNet. Additionally, the Transit Circle Inspired Optimization-LeNet attains an accuracy of 0.925, True Positive Rate of 0.930, True Negative Rate of 0.915, False Negative Rate of 0.070, and False Positive Rate of 0.085. The results show that the proposed method has better performance and a high detection rate.