Fundamentals of artificial metaplasticity in radial basis function networks for breast cancer classification

Modern medicine generates data commonly used for the development of clinical decision support systems, whose usefulness often lies in the performance of the machine learning algorithms used for the processing of that data. Several lines of research seek to resemble artificial neural networks to biological ones by incorporating new bioinspired mechanisms. One of these mechanisms is the biological concept of metaplasticity, defined as the plasticity of synaptic plasticity and which has been shown to be directly related to learning and memory. It has also been shown that incorporating this mechanism into a multilayer perceptron improves the neural network performance in both accuracy and learning rate when diagnosing breast cancer. The early detection of breast cancer is one of the most important strategies to prevent deaths from this disease. In this work, we have modeled synaptic metaplasticity in a radial base function network, which converges faster than multilayer perceptrons, with the motivation to achieve a more accurate solution in the diagnosis of breast cancer.