Fig. 13.

A selection of segmentations showing the features of the proposed method and LPA. (A) and (B) show cases where both methods perform well. (C) shows a case where the proposed method produces false positive voxels (arrow) in the GM, not present in LPA which does not consider GM. (D) shows a large infarct extending into the cortex where the extension into the cortex (arrow) is poorly segmented by LPA. (E) shows a case where small lesions are missed by LPA, despite considerable over segmentation (arrow).