Patch-Based Segmentation with Spatial Consistency: Application to MS Lesions in Brain MRI

Algorithm 1

The lesion detection and segmentation algorithm.

Input: An MR brain image.

Determine a subset of training brains that are most similar to the given test.

for    do

 (i) Set .

 (ii) For each patch centered at pixel in the test brain image, find similar patches in the training set using

     the distance measure defined in (7).

 (iii) Compute the weight of each selected patched using (4) and compute the weighted average lesion label decision using (5).

end for

Output: A voxel-labeled lesion map of the brain image.