AbstractID: 8926 Title: How do Stochastic Noise and Breast Structure affect Detection
Performance in Digital Mammography?
We have investigated how the observer detection threshold for two simulated mass lesions varies with stochastic noise and breast
structure. A digital mammography system was used to acquire radiographs of an anthropomorphic breast phantom, with mammograms
acquired with and without the mass lesions. A difference image was generated corresponding to the lesion alone, which was added at a
reduced intensity to a non-lesion digital mammogram during a 4-Alternate Forced-Choice (4-AFC) experiment. The lesion intensity
that corresponded to a 92% correct performance level in the 4-AFC experiments was determined (I92%). Stochastic noise in the image
was estimated on a per pixel basis using a Monte Carlo model of the mammographic system. Breast structure was quantified using a
measure of the similarity of the background breast structure and the lesion to be detected. Values of I92% , the stochastic noise and the
similarity measure were obtained at 12 different locations in the image of the anthropomorphic phantom. The results indicated that
lesion intensity required for detection was about 25% lower in the lower noise regions. The similarity measure, based on local
structure, was a better predictor of I92% than the calculated stochastic noise.