A statistical review of skin disease detection models for different skin-tones

Abstract

Abstract: Variations in the body's melanin synthesis are one of the most important factors that contribute to the development of skin diseases. As a consequence of these abnormalities, a variety of skin illnesses, including but not limited to acne, psoriasis, eczema, vitiligo, ichthyosis, and seborrheic dermatitis, may develop. Each of these illnesses manifests itself in a manner that can be seen on a person's body, making it possible to diagnose each of them using image processing methods. Scientists have developed a variety of models to represent the real workings of image processing throughout the years. Convolutional neural network (CNN), evolving fuzzy classification, constrained-syntax genetic programming, and the Naive Bayesian classifier are examples of models that belong to this group (NB). However, these models do not account for a variety of characteristics, such as geographical differences, the effect of age on the skin, gender, skin tone, skin type, and other comparable aspects when making judgments. This has led to the evaluation of a number of stratification models for skin disorders, which is discussed in this particular piece of literature. In addition, the model now incorporates performance metrics for precision, computational complexity, scalability, and other parametric adjustments. Researchers and skin care professionals are encouraged to use this knowledge to discover the algorithmic combination that is most advantageous for a particular application. Consequently, the building of a high-performance system for classifying skin illnesses and the rapid development of apps will be made feasible. Various system models, such as high-performance approaches for feature extraction and selection, illness stratification, skin segmentation, and more, may also be used. There are several system models from which to pick. This article employs a multiparametric approach, which takes into account a range of variables, such as a person's skin tone and geographical characteristics. In order to categorize the analysed algorithms, these criteria and other performance indicators are reviewed and considered. In addition, this study provides recommendations for enhancing current models via algorithmic fusion, model enhancement, incremental learning, and related approaches. The readers of this post will be able to apply these enhancements, which will improve the functionality of existing systems in a variety of use cases.