Automated Medical Diagnosis and Classification of Skin Diseases Using Efficinetnet-B0 Convolutional Neural Network

Abstract

AbstractThe skin is the biggest and most vulnerable organ in the body, so skin illnesses are common nowadays. However, the significance of skin health is sometimes underestimated. According to one survey, around 1.79% of the world population suffers from skin-related disorders. These disorders can be fatal when they are not given treatment in their early stages. As a result, skin illnesses must be recognized and diagnosed early to avert severe dangers to one's life. However, the patient may be impacted, and they are frequently subjected to extensive examinations to determine the severity of their skin issue. As a result, we must create an expert system capable of detecting illnesses at an early stage. At the moment, just a few computerized methods are available for detecting skin illnesses, but this is an era in which substantial research is being conducted and may be further expanded. In this paper, an expert system was created using the EfficientNet B-0 model, and the model can also be used to assist specialists in more effectively and efficiently identifying and diagnosing various significant skin diseases such as (Eczema, Psoriasis & Lichen Planus, Benign Tumours, Fungal Infections, and Viral Infections). In addition, we conducted a comparison of sequential Convolutional Neural Networks (CNNs), EfficientNet B0, and ResNet50. Through these models, the reasons for recognized skin illness may be defined, and therapy may be offered. The Python programming language was utilized to implement the models. The dataset was obtained via DERMNET. Using EfficientNet-B0 to train the model and predict outcomes at an epoch value of 10, we attained an accuracy of 91.36%.KeywordsExpert systemCNNsEfficientNet B0ResNet50Skin diseasesDermnet