Discriminatively weighted multi-scale local binary patterns

摘要

Apparatus and methods associated with detecting prostate cancer (CaP) in a magnetic resonance (MR) image of a prostate of a CaP patient are described. One example apparatus includes logics that acquire an image of a prostate, learn a weighted vector, detect salient features in the image of the prostate, and generate a heatmap that facilitates detecting CaP. An image acquisition logic acquires a T2 weighted MR image of a prostate. A learning logic learns a weighted vector based on a set of positive LBP descriptors and a set of negative LBP descriptors extracted from the image at multiple scales. A salient feature detection logic detects salient features in the image based on the weighted vector and a pixel-by-pixel weighted Hamming matching of the image. A prediction logic generates a statistical probability heatmap based on the weighted vector and the weighted Hamming matching of the image.