Eyelid basal cell carcinoma classification using shallow and deep learning artificial neural networks

Results of exhaustive application of neural network pattern recognition and classification of cases suffering from eyelid basal cell carcinoma are reported. Recognition and classification were based on shallow and deep learning methods; namely, multi-layer error backpropagation and convolution neural networks were utilized. The processed material consisted of full-face, or half-face photographs of healthy subjects and patients suffering from eyelid basal cell carcinoma. Various training and learning methods were used and the efficiency of the proposed algorithms was evaluated using as performance metrics the accuracy score, that is, the ratio of the number of the correctly classified cases over the total number of cases under examination. With respect to the accuracy, some of the proposed algorithms reached up to 100%.