Intelligent detection of melanoma growth stage based on the analysis of the thermal response of skin

Abstract

In response to a thermal stimulation, skin tumours produce different temperature variations in comparison with healthy tissues. In this paper, we exploit this fact to design an intelligent melanoma detection system that estimates the development stage of a skin tumour. This system is based on a feed-forward artificial neural network, which receives a signal of the thermal response measured from the skin surface, and predicts the growth stage of the tumour. The measurement should be performed during the heat recovery after removing a cold stimulus. In order to train and test the neural network, here, these signals are provided by FEM-based simulations that solve the Pennes' bioheat transfer equation. Also, the signals are processed quantitatively and their convenient features are extracted. The achieved accuracy of 96% shows that the thermal response as the distinguishing criterion is an appropriate choice for the early diagnosis of the melanoma type of skin cancer.