Classification of mammographic features using RBF-SA
DOI:
https://doi.org/10.5585/exacta.v4i2.765Keywords:
Mammography. Neural networks. Optimization. Performance. Simulated annealing.Abstract
We present in this work a new type of classes discriminator based upon nonlinear and combinational optimization techniques: radial basis functions-simulated annealing (RBF-SA). The combinational optimization method is used here as a preestimation of some parameters of the network classifier. We compare the classifier performance with and without pre-estimation. For training the classifiers, adopting the leave-one-out procedure, we have used case examples such as mammographic masses (malignant and benign). The classifier is trained with shape factors and edge-sharpness measures extracted from 57 regions of interest (ROI) (37 malignant and 20 benign), manually delineated, that describe mammographic masses and tumor features in terms of polygonal models for shape factors (compactness [CC], Fourier description [FF], fractional concavity [FCC] and speculated index [SI]) and edge sharpness-acutance (A) . The classifier performance is compared in terms of the area under the receive operating characteristic (ROC) curve – (A). Higher values of A correspond to a better performance of classifier. Experiments with mammographic tumor and masses show that the best result of 0.9776 is obtained with RBF-SA when RBF parameters such as centers and spread matrix are pre-estimated, which is significantly better than the results obtained with no pre-estimation or only pre-estimation of the RBF centers, which are, 0.7071 and 0.9552 respectively.Downloads
Download data is not yet available.