Detection and Localisation of Prostate Abnormalities

We propose a new methodology for prostate cancer detection and localisation within the peripheral zone based on combining multiple segmentations. We extract four features using Gabor and median filters. Subsequently, we use each feature separately to generate binary segmentations taking the feature space and intensity values into account. We perform erosion on each of the segmentations to remove false positive regions. Finally, we take the intersection of all four binary segmentations, taking a model of the peripheral zone into account. The initial evaluation of this method is based on 66 MRI images from 19 patients and 84.85% of the cases were classified correctly.