Novelty detection for the identification of masses in mammograms

• DOI: 10.1049/cp:19950597
• Publication Date: 2005-11-09T21:40:06Z
• Authors: L. Tarassenko
• Keywords: Novelty Detection; Identification

Breast cancer is the major cause of death amongst women in 35 to 55 age group. Mammography only feasible imaging modality for screening large numbers women. With present policy, there are three million mammograms be analysed each year UK; therefore a need (as yet unmet) an automated analysis system which could highlight areas interest. In first instance, interest might simply any mass-like structures and this indeed approach reported on paper. typical many problems medicine: class real under-represented database available examples hence its prior probability will very low. As result this, few abnormalities existing databases. If neural network classifier trained using standard minimising mean-squared error (MSE) at output, ignored. We have been exploring alternative we attempt learn description normality number do not show evidence structures. The idea then test novelty against order try identify candidate masses previously unseen images interpretation sample results so far obtained database.