Massive Lesions Classification using Features based on Morphological Lesion Differences

Purpose of this work is the development of anautomatic classification system which could be useful for radiologistsin the investigation of breast cancer. The software has been designedin the framework of the MAGIC-5 collaboration.In the automatic classification system the suspicious regions withhigh probability to include a lesion are extracted from the image asregions of interest (ROIs). Each ROI is characterized by somefeatures based on morphological lesion differences.Some classifiers as a Feed Forward Neural Network, a K-NearestNeighbours and a Support Vector Machine are used to distinguish thepathological records from the healthy ones.The results obtained in terms of sensitivity (percentage ofpathological ROIs correctly classified) and specificity (percentage ofnon-pathological ROIs correctly classified) will be presented throughthe Receive Operating Characteristic curve (ROC). In particular thebest performances are 88% ± 1 of area under ROC curve obtainedwith the Feed Forward Neural Network.