Feasibility of multi-layered perceptron network in discriminating breast magnetic resonance imaging lesions

The feasibility of MLP networks in reliably discriminating between benign and malignant lesions is demonstrated in this pilot study. MLP networks can be trained on similar information that a radiologist would use to interpret lesions on DCE-MRI study of breast and perform as well as a trained radiologist. In the future this work will be extended to a larger data set, and the feasibility of other neural networks such as radial basis function will be tested.