Neural snakes: An active contour model using neural networks for mammogram segmentation
Alexei Stapinski · The Atrium (University of Guelph) · 2003
Computer-Aided Diagnosis of Mammograms is a developing research area. The use of a computer system as a second reader for the Radiologists is appealing as a means of increasing diagnosis accuracy. Multiple steps are present in Computer-Aided Diagnosis systems, including a segmentation of the breast from the rest of the mammogram. A fully automated method for a true border mammogram segmentation is detailed. The method makes use of Active Contours or "Snakes" as a means of locating the true border. A new derivation of Active Contour, which uses a neural network and is hence dubbed the Neural Snake, is researched and created. Both Snakes are tested, and their behaviour is compared. Future work on Neural Snakes is detailed.