Skin Disease Diagnosis using Mini-batch Fuzzy C-Means Clustering and CNN

abstract-Skin illnesses vary widely in look and intricacy, diagnosing skin diseases is a difficult undertaking. In this work, we integrate mini-batch fuzzy C-means (MBFCM) clustering plus convolutional neural networks (CNNs) to present a unique method for automated skin disease diagnosis. The suggested approach seeks to enhance the diagnosis process’s interpretability and accuracy. Initially, MBFCM clustering is used to segment the input skin photos into discrete groups, successfully capturing the structural underpinnings and patterns found in the images. The CNN architecture is then fed with the clustered picture patches in order to extract and classify features. By learning discriminative features unique to each cluster through training on the clustered data, the CNN improves its diagnostic capacity. The efficiency of the suggested strategy is demonstrated by experimental findings on a dermatology dataset that is available to the general audience. Our suggested strategy performs better in terms of accuracy and robustness for diagnosing skin diseases than traditional CNN-based techniques. Furthermore, the process of clustering offers significant understanding of the distribution of various skin diseases in the dataset, which facilitates the deciphering of the model’s conclusions.