Evaluation of a two-stage model for detection of abnormalities in digital mammograms

Describes the characterization and development of a model for the detection of abnormalities in digital mammograms. The underlying detection model was previously introduced by Lai et al. (1989) for the detection of circumscribed masses in mammograms. This has been further developed into a two stage process whereby, in the first stage, the objective is to seek out all potential abnormalities present on the digitized mammogram while in the second stage these potential abnormalities are further investigated to establish whether they are genuine abnormalities or false alarms. This paper examines various spatial and textural techniques which have been developed in order to evaluate and then improve the performance of this detection model. The aim is ultimately to identify techniques which enhance the performance of this two-stage model of detection, and then to adapt the model for the purpose of analyzing mammograms for circumscribed masses and/or other masses