Discriminatively weighted multi-scale local binary patterns

A non-transitory computer-readable storage medium storing computer executable instructions that when executed by a computer cause the computer to perform a method, the method comprising:

• accessing an input image, where the input image is a digitized image comprising a set of pixels and intensity information about the set of pixels;
• labelling a non-empty subset of pixels in the input image as a positive set;
• labelling a non-empty subset of pixels in the input image as a negative set, where the negative set does not include the positive set;
• computing a first sum of positive local binary pattern (LBP) descriptors as a function of the positive set;
• computing a second sum of negative and positive LBP descriptors as a function of the negative set and the positive set;
• computing a weighted vector from the first sum of positive LBP descriptors and the second sum of negative and positive LBP descriptors;
• extracting a multiscale LBP at a co-ordinate from the input image at a plurality of scales;
• finding the Hamming distance of the multiscale LBP at the plurality of scales;
• computing a weighted sum as a function of the transpose of the weighted vector and a column vector of the Hamming distance of the multiscale LBP;
• assigning the weighted sum for the co-ordinate to a statistical probability heatmap of the input image;
• updating a memory to store the statistical probability heatmap, and
• controlling a display to display the statistical probability heatmap, where the statistical probability heatmap facilitates distinguishing regions of the input image associated with the positive set and regions of the input image associated with the negative set.

An apparatus comprising:

• a processor;
• a memory;
• an input/output interface;
• a set of logics;
• and
• an interface to connect the processor, the memory, the input/output interface and the set of logics, the set of logics comprising:
• an image acquisition logic that acquires an image of a region of tissue;
• a learning logic that learns a weighted vector based, at least in part, on the image;
• a salient feature detection logic that detects salient features in the image, based, at least in part, on the weighted vector, and
• a prediction logic that generates a heatmap that distinguishes salient features from non-salient features in the image.

An apparatus comprising:

• a processor;
• a memory;
• an input/output interface;
• a set of logics;
• and
• an interface to connect the processor, the memory, the input/output interface and the set of logics, the set of logics comprising:
• a first logic that generates a weighted vector;
• a second logic that determines a similarity between a template pixel and another pixel in an image based, at least in part on the weighted vector;
• and
• a third logic that facilitates predicting cancer by producing a heatmap of the image based, at least in part, on the weighted vector and the similarity.