Classifier scheme for breast clustered microcalcifications by using artificial neural networks (ref. 21,22,26)

Development of an automatic classifier scheme with purposes of indicating the case as "suspect" or "not suspect", based on investigations of digitized mammographic images. In this particular study, the focus was clustered microcalcifications previously detected by image processing techniques. The classification technique used was based on supervised artificial neural networks (ANR), employing the backpropagation learning algorithm. The classifier scheme have shown the efficacy of shape descriptors on the clusters characterization as well as the influence of features extracted from the medical reports, as age and densification. The best results indicated a 92% of correction with Az of 0.96 in the ROC curve.