Percorrer por autor "Silva, Miguel Almeida e"
A mostrar 1 - 1 de 1
Resultados por página
Opções de ordenação
- New approach of skin cancer detection using a deep learning methodPublication . Silva, Miguel Almeida e; Parlak, Onur; Rodrigues, Pedro Miguel de LuísSkin cancer occurs when normal skin cell growth becomes abnormal4often triggered by ultraviolet light exposure4and early detection is critical for effective treatment. This thesis investigates a novel diagnostic approach that targets specific amino acids associated with skin cancer cells. Focusing on asparagine, aspartic acid, and glutamine, it was employed three electrodes made of gold, platinum, and carbon to measure these biomarkers using the Square Wave Voltammetry method. The resulting signals were converted into spectrograms to serve as input for a deep learning model. A total of 1,350 spectrogram samples (150 per amino acid per electrode) were used to train and evaluate the model. For the gold electrode, the model achieved an accuracy of 82% (F1 score: 81%, precision: 84%, recall: 82%, Cohen9s kappa: 73%, ROC AUC: 93%), excelling in the identification of aspartic acid and glutamine while underperforming on asparagine. The platinum electrode attained an overall accuracy of 80% with perfect classification of aspartic acid, whereas the carbon electrode reached 78% accuracy, performing best for asparagine. In tests with amino acid mixtures, the gold electrode reliably detected all components, whereas the platinum and carbon electrodes showed selective misclassifications. These findings highlight that electrode material significantly influences spectral pattern recognition and suggest that combining optimized electrode selection with deep learning can enhance early skin cancer diagnosis. Furthermore, the study demonstrates that increasing sample volume improves model accuracy, providing a promising foundation for future diagnostic tools.