Improved Classification of Arsenic-Affected Skin Diseases Through Image Processing and Transfer Learning

Arsenic contamination of groundwater is a global health concern, leading to adverse health effects, including skin diseases. Early detection is crucial for prevention and treatment, although manual methods are often time-consuming and error-prone. To address this issue, deep learning methods, specifically transfer learning, offer a promising solution for accurate and efficient skin disease detection. Therefore, the research aims to propose a comprehensive framework that uses ResNet152V2 architecture along with Gaussian smoothing methods to improve the classification accuracy of skin images exposed to arsenic. ResNet152V2 model is pre-trained on large-scale image datasets, providing powerful feature extraction fine-tuned on the ArsenicSkinBD dataset. The images are preprocessed using Gaussian smoothing to reduce noise and enhance feature clarity. Specifically, the research introduces the innovative application of Gaussian smoothing along with transfer learning for skin disease classification, which has not been extensively explored in previous studies. The results show a significant increase in classification accuracy, achieving approximately 0.9904 on the testing set compared to 0.9881 without enhancements. This improvement shows the effectiveness of the method in detecting skin diseases caused by arsenic exposure. The use of Gaussian smoothing also reduces loss values on the testing set, indicating that the model becomes more efficient in optimizing its predictions. The proposed framework not only enhances detection accuracy but also supports more efficient diagnostic processes, contributing to better prevention and treatment efforts.