Film-Screen Artefact Detection in Mammograms

Abstract

Shot noise is the result of “foreign bodies” such as dust or hairs being present on the intensifying screen or on the film during the acquisition. Shot noise is an important potential source of false positives in the detection of micro-calcifications. As such, there has been interest in its detection and labelling which is the aim of the work in this paper. In this paper, we give a clearer and better mathematical justification for our previous work in this area which uses the inherent blur in the imaging system to identify what is noise and what is real calcification. In doing so, we develop a new algorithm with a more mathematically correct approach which is iterative. This ensures that not only are individual pixels correctly marked as shot noise, but the whole area of noise is marked as such. The new algorithm due to its iterative nature is slower than the old method, but reduces the number of false positives even further in sub¬sequent CAD algorithms by being based on firmer theory and detecting the whole noise region.