Benign and malignant breast tumors classification based on region growing and CNN segmentation

Rahimeh Rouhi, Mehdi Jafari, Shohreh Kasaei, Peiman Keshavarzian

Abstract: Highlights•CNN templates are generated using a genetic algorithm to segment mammograms.•An adaptive threshold is computed in region growing process by using ANN and intensity features.•In tumor classification, CNN produces better results than region growing.•MLP produces the highest classification accuracy among other classifiers.•Results on DDSM images are more appropriate than those of MIAS.