| Deletions are marked like this. | Additions are marked like this. |
| Line 27: | Line 27: |
|
(1) `volume difference` = 2*|A|-|B|/|A|+|B| (2) `volume overlap (Dice)` = 2*|AB|/|A|+|B|; the intersection over the mean of the segmentation volumes (3) `volume overlap (Jaccard)` = |AB|/|A+B|; the intersection over the union of the segmentation volumes |
* '''relative volume difference''' = 2*(|A|-|B|)/|A|+|B| `(the volume difference normalized by the mean)` * '''relative volume overlap (Dice)''' = 2*|A-B|/|A|+|B| `(the intersection over the mean of the segmentation volumes)` * '''relative volume overlap (Jaccard)''' = |A-B|/|A+B| `(the intersection over the union of the segmentation volumes)` Note, that the label files (e.g. aseg.mgz) need to be in a common space. You can use [[mri_convert]] with -rt nearest and -at to apply a transform to the segmentations. |
Index
Contents
Name
mri_compute_overlap - Computing three label overlap measures: volume difference, Dice and Jaccard.
Synopsis
mri_compute_overlap [options] <volume 1> <volume 2> [label numbers]
Arguments
Positional Arguments
volume 1 |
volume 2 |
[labels] list of labels to be studied, needs to be specified if -a (analyze all option) is not specified |
Optional Flagged Arguments
-a compute overlap of all labels (if missing, labels of interest should be listed) |
-s show label name for segmentation |
-l <fname> filename to write results to |
-h print help |
Description
mri_compute_overlap computes three different types of overlap measures either for all the existing labels in the input volumes or a subset of them that the users lists in the command line. The three overlap measures are the following:
relative volume difference = 2*(|A|-|B|)/|A|+|B| (the volume difference normalized by the mean)
relative volume overlap (Dice) = 2*|A-B|/|A|+|B| (the intersection over the mean of the segmentation volumes)
relative volume overlap (Jaccard) = |A-B|/|A+B| (the intersection over the union of the segmentation volumes)
Note, that the label files (e.g. aseg.mgz) need to be in a common space. You can use mri_convert with -rt nearest and -at to apply a transform to the segmentations.
Examples
References
None