Classification of Skin Lesion With Features Extraction Using Quantum Chebyshev Polynomials and Autoencoder From Wavelet-Transformed Images

Skin, a vital organ acting as a protective barrier to the external environment, plays a pivotal role in overall human health. Early detection of skin diseases is essential, as untreated conditions can escalate to serious issues such as skin cancer. This study presents an innovative automated system designed for efficient classification of skin lesions, addressing the growing demand for advanced biomedical image analysis. Leveraging the power of Deep Learning, the proposed model incorporates several pre-processing techniques such as wavelet transformations, pooling methods, and normalization to enhance image clarity and remove extraneous artifacts. Two distinct feature extractors are used to extract key features: Quantum Chebyshev polynomials for initial feature extraction, followed by an Autoencoder (AE) for feature refinement and dimensionality reduction. These optimized features are classified using Long Short-Term Memory (LSTM). The experimental evaluation of the proposed model includes analysis with five different optimizers: Adam, RMSprop, SGD, Adadelta, and Adagrad, accross two widely recognized datasets, ISIC2017 and HAM10000. The resutlts reveals that the Adam optimizer consistently yields the highest scores across multiple evaluation matrices. For the ISIC2017 dataset, the model achieves 98.87% accuracy, 98.23% precision, 98.26% recall, F1-score 98.24%, and 98.16% specificity. The HAM10000 dataset exhibits even more remarkable metrics, with 99.58% accuracy, 97.84% precision, 97.49% recall, 97.66% F1-score, and 97.74% specificity. The proposed model surpasses the current state-of-the-art in skin lesion classification and holds the potential to serve as a valuable tool for medical professionals, aiding in the automated classification of skin cancer.