A Novel Skin Disease Classification Framework Using Template Matching and Stacked Ensemble Learning

Skin disorders are common health problems around the world, and they typically carry underlying hazards that can lead to physical and mental illness. Accurately detecting skin diseases from clinical images presents a substantial issue in medical image management. The subjective and time-consuming manual diagnosis processes of medical specialists underline the need for automatic skin disease prediction to help patients and dermatologists plan their treatments more quickly. The primary contribution of this research is to present a stacked ensemble learning-based autonomous skin disease detection system capable of detecting disease spread at an early stage and assessing disease severity. This paper also discusses how the skin disease detection system will be implemented utilizing unique segmentation using template matching and hybrid feature extraction strategies to extract important features for classification. An initial evaluation of the proposed framework will be carried out with the assistance of machine learning techniques such as K-nearest neighbors, random forest classifiers and support vector machines. Skin diseases such as melanoma, melanocytic nevi, eczema, and psoriasis are taken into consideration while applying the suggested strategy, which also includes the utilization of multiple classifiers and stacked ensemble learning. Additionally, accuracy, recall, and F-score are utilized to evaluate the proposed work in comparison to conventional machine learning classifiers. The findings indicate that the suggested method, which makes use of stacked ensemble learning, achieves 98.2 % of accuracy proving it to be higher comparing with the conventional machine learning classifier performance.