Neural network diagnosis of malignant skin cancers using principal component analysis as a preprocessor

This paper presents an artificial neural network which is used to separate the malignant melanoma from benign categories of skin cancers based on cancer shapes and their relative color. To reduce the computational complexities, while increasing the possibility of not being trapped in local minima of the back-propagation neural network, we applied PCA (principal component analysis) to the originally training patterns, and utilized a cross entropy error function between the output and the target patterns. By using this method, more built-in features of the cancer image through its color and the cancer shapes could be used as the input of the system, leading to higher accuracy of finding the differences between malignant cancer from the benign one. Using this approach, for reasonably balance of training/testing sets, above 91,8% of correct classification of malignant and benign cancers could be obtained.