Deep Learning-driven Skin Disease Diagnosis: Advancing Precision and Patient-Centered Care

Abstract

Skin diseases are in the middle of the most prevalent conditions, arising from a myriad of factors including viral infections, bacteria, allergies, and fungal pathogens. Appropriate detection of these conditions is essential for effective treatment and management. Further, Deep learning methods are employed to enable early-stage detection, with a particular emphasis on the pivotal role of feature extraction in the classification process. This research emphasizes the significance of a patient-centered approach, aiming to provide responsible and effective solutions for skin diagnoses. In pursuing more accurate and timely skin condition diagnoses, we turn to deep learning techniques, leveraging the HAM10000 dataset. Initially, we perform different prepossessing techniques on selected datasets to handle class imbalance and a Convolutional Neural Network and fine-tune hyperparameters such as with or without Dropout, CW, FL, and Using Global Average Pooling. Our technique excels in distinguishing diverse skin, Gender, localization, and Cell types with reliable evaluation metrics such as precision, recall, FI Score, and specificity. Our technique not only subsidizes the healthcare field but also underscores the potential of advanced technologies in enhancing early skin disease detection and medical decision-making.