Enhanced Skin Disease Diagnosis through Optimized Multi-Dimensional Attention Spiking Neural Networks Utilizing Clinical Imaging and Patient Data

The early and accurate diagnosis of skin lesions is crucial for effective treatment and patient results, particularly in identifying malignant conditions. Traditional diagnostic methods often depend on manual analysis, which can be susceptible to human error due to the vast number of lesion images that need evaluation. To address these limitations, an innovative approach named, Optimized Multi-Dimensional Attention Spiking Neural Networks Utilizing Clinical Imaging and Patient Data (SDDOMDASNN) is proposed. Initially, the data is collected from PAD-UFES-20 dataset. The system begins with image preprocessing using a Learnable Edge Collaborative Filter (LECF), which effectively removes noise and preserves essential image details. A key innovation in this method is the feature extraction process, which utilizes Parameterized Local Maximum Synchrosqueezing Transforms (PLMST) to extract Bi-Sectional Texture Features, which allows the system to analyze skin texture from multiple perspectives. A Multi-Dimensional Attention Spiking Neural Network (MDASNN) is used to classification that focuses on the most relevant features of the lesion, mimicking the human brain's ability to prioritize some information. The network's performance is further enhanced by the Binary Battle Royale Optimizer Algorithm (BBROA), which fine-tunes the model to complete optimal classification accuracy. The proposed SDD-OMDASNN system is evaluated on various skin conditions, including Basal Cell Carcinoma (BCC), Squamous Cell Carcinoma (SCC), Actinic Keratosis (ACK), Seborrheic Keratosis (SEK), Bowen's disease (BOD), Melanoma (MEL), and nevus (NEV). The results demonstrate that the proposed methodology gives a higher accuracy of 99.05% and a higher recall of 98%, outperforming than other existing techniques.