Differences between revisions 8 and 15 (spanning 7 versions)
⇤ ← Revision 8 as of 2018-09-20 10:33:30 →
Size: 3428
Editor: AndrewHoopes
Comment:
← Revision 15 as of 2018-09-24 11:21:53 → ⇥
Size: 3962
Editor: AndrewHoopes
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
= mri_overlap = = mri_seg_overlap =
Line 3: Line 3:
This page proposes a new program, called `mri_overlap`, that will replace the redundant (and buggy/slow) `mri_compute_seg_overlap` and `mri_compute_overlap`. A primary new feature is the ability to save a detailed overlap report to a JSON file. This page proposes a new program, called `mri_seg_overlap`, that will replace the redundant (and buggy/slow) `mri_compute_seg_overlap` and `mri_compute_overlap`. A primary new feature is the ability to save a detailed overlap report to a JSON file.
Line 6: Line 6:
USAGE: USAGE
 mri_seg_overlap [options] <vol1> <vol2>
Line 8: Line 9:
    mri_overlap <vol1> <vol2> [options] DESCRIPTION
 
 This program computes the structural overlap between two segmentation
 volumes. The following statistics can be calculated for each label:
 
      dice coefficient - F1-measure of the similarity between
                              the two labels.
      jaccard coefficient - Intersecting volume over the combined
                              volume of the two labels.
      volume difference - Fractional difference in volumes of
                              the labels.
 
 By default, only dice scores are computed, but this can be
 reconfigured with the --measures flag.
 
 Each structure in the input volumes will be considered by default, but
Line 10: Line 26:
DESCRIPTION: ---- /!\ '''Edit conflict - other version:''' ----
 an alternative list of labels can be specified with the --label or -labelfile
Line 12: Line 29:
    This program computes the overlap between two segmentation volumes. The following
    statistics can be calculated for each label:		 ---- /!\ '''Edit conflict - your version:''' ----
 an alternative list of labels can be specified with the --label or --labelfile
Line 15: Line 32:
        dice coefficient - F1-measure of the similarity between the two labels
        jaccard coefficient - intersecting volume over the combined volume of the two labels
        volume difference - fractional difference in volumes of the labels		 ---- /!\ '''End of edit conflict''' ----
 flags. Additionally, the --seg flag will compute the overlap between
 the following 12 major anatomical structures:
 
      Cerebral White Matter (2,41)
      Cerebral Cortex (3,42)
      Hippocampus (17,53)
      Caudate (11,50)
      Putamen (12,51)
      Pallidum (13,52)
      Amygdala (18,54)
      Thalamus Proper (10,49)
      Lateral Ventricle (4,43)
      Third and Fourth Ventricles (14,15)
      Inf Lateral Vent (5,44)
      Accumbens Area (26,58)
  
Line 19: Line 50:
    By default, only dice scores are computed, but this can be reconfigured
    with the --measures flag.		 OPTIONAL ARGUMENTS
 -o, --out [FILE]
  Save detailed overlap report to a JSON file.
Line 22: Line 54:
    Each structure in the input volumes will be considered by default, but
    an alternative list of labels can be specified with the -l flag. Additionally,
    the --seg flag will compute the overlap between the following 12 major
    anatomical structures:		  -m, --measures [MEASURE...]
  List of measures to compute. By default, only dice scores are
  computed. Options include: dice, jaccard, voldiff
Line 27: Line 58:
        Hippocampus (17,53)
        Caudate (11,50)
        Putamen (12,51)
        Pallidum (13,52)
        Amygdala (18,54)
        Thalamus Proper (10,49)
        Lateral Ventricle (4,43)
        Third and Fourth Ventricles (14,15)
        Inf Lateral Vent (5,44)
        Cerebral White Matter (2,41)
        Cerebral Cortex (3,42)
        Accumbens Area (26,58)		  -l, --labels [LABEL...]
  Space-separated list of label values to include. By default,
  the overlap will be computed between all labels available in
  both input volumes.
Line 40: Line 63:
OPTIONAL ARGUMENTS:  -n, --names [LABELNAME...]
  Custom label names corresponding to the values specified with
  the --labels flag.
Line 42: Line 67:
    -o, --out
        Save detailed overlap report to a JSON file.		  -f, --labelfile [FILE]
  Text file specifying the label values to include. This file
  must be in the format of a freesurfer lookup-table, in which
  each line contains a label value and, optionally, a
  corresponding label name.
Line 45: Line 73:
    -m, --measures
        List of measures to compute. By default, only dice scores
        are computed. Options include: dice, jaccard, voldiff		  -x, --no-names
  Do not report label names.
Line 49: Line 76:
    -l, --labels
        Space-separated list of label values to include. By default, the overlap
        will be computed between all labels available in both input volumes.		  -s, --seg
  Compute overlap between the the major segmentation structures
  listed above.
Line 53: Line 80:
    -f, --labelfile
        File specifying a list of labels to include. This file can either be a
        space-separated list of label values or in the format of a freesurfer
        color lookup-table.		  -q, --quiet
  Quiet mode - do not print results to stdout.
Line 58: Line 83:
    -s, --seg
        Compute overlap between the major segmentation structures listed above.

EXAMPLE:

    mri_overlap aseg.mgz mseg.mgz -l 17 53 -m dice voldiff -o overlap.json

    This example will compute the dice scores and volume differences of
    hippocampal segmentations between the two input volumes. Results will
    be saved to overlap.json.
Line 72: Line 87:
The proposed JSON output would look something like this: NOTE: if the `--seg` flag is used, an additional statistic called `weighted-subcortical-mean`, which ignores wm and cortex, will be reported for each measure.
Line 76: Line 91:
    "names": {
        "1": "structure 1",
        "2": "structure 2"
        ...
    },
Line 82: Line 92:
        "/path/to/manual_label.mgz",
        "/other/path/to/aseg.mgz"		         "/path/to/aseg.mgz",
        "/path/to/mseg.mgz"
Line 87: Line 97:
            "mean": 0.9142,
            "std": 0.0232,
Line 90: Line 98:
                "2": 0.9143,
                "3": 0.8739,
                "4": 0.9002,		                 "10": 0.9075892566820576,
                "11": 0.8953675962203272,
                "12": 0.9146751935375295,
Line 94: Line 102:
            }             },
            "mean": 0.6805390452588164,
            "std": 0.3237900375396044,
            "weighted-mean": 0.9340674985029207
Line 97: Line 108:
            "mean": 0.8902,
            "std": 0.0144,
Line 100: Line 109:
                "2": 0.8921,                 "10": 0.8308131920466723,
                "11": 0.8105570623826414,
                "12": 0.8427663203597457,
Line 102: Line 113:
            }
        },
        "voldiff": {
            "mean": 0.0221,
            "std": 0.0032,
            "labels": {
                "2": 0.0123,
                ...
            }		             },
            "mean": 0.5892475078628221,
            "std": 0.3029904264784899,
            "weighted-mean": 0.8801749135061013
Line 112: Line 118:
    },
    "names": {
        "10": "Left-Thalamus",
        "11": "Left-Caudate",
        "12": "Left-Putamen",
        ...

mri_seg_overlap

This page proposes a new program, called mri_seg_overlap, that will replace the redundant (and buggy/slow) mri_compute_seg_overlap and mri_compute_overlap. A primary new feature is the ability to save a detailed overlap report to a JSON file. 

USAGE
        mri_seg_overlap [options] <vol1> <vol2>

DESCRIPTION
        
        This program computes the structural overlap between two segmentation 
        volumes. The following statistics can be calculated for each label:
        
             dice coefficient     -  F1-measure of the similarity between
                                     the two labels.
             jaccard coefficient  -  Intersecting volume over the combined
                                     volume of the two labels.
             volume difference    -  Fractional difference in volumes of
                                     the labels.
        
        By default, only dice scores are computed, but this can be 
        reconfigured with the --measures flag.
        
        Each structure in the input volumes will be considered by default, but

---- /!\ '''Edit conflict - other version:''' ----
        an alternative list of labels can be specified with the --label or -labelfile

---- /!\ '''Edit conflict - your version:''' ----
        an alternative list of labels can be specified with the --label or --labelfile

---- /!\ '''End of edit conflict''' ----
        flags. Additionally, the --seg flag will compute the overlap between 
        the following 12 major anatomical structures:
        
             Cerebral White Matter (2,41)
             Cerebral Cortex (3,42)
             Hippocampus (17,53)
             Caudate (11,50)
             Putamen (12,51)
             Pallidum (13,52)
             Amygdala (18,54)
             Thalamus Proper (10,49)
             Lateral Ventricle (4,43)
             Third and Fourth Ventricles (14,15)
             Inf Lateral Vent (5,44)
             Accumbens Area (26,58)
         

OPTIONAL ARGUMENTS
        -o, --out [FILE]
                Save detailed overlap report to a JSON file.

        -m, --measures  [MEASURE...]
                List of measures to compute. By default, only dice scores are 
                computed. Options include: dice, jaccard, voldiff

        -l, --labels  [LABEL...]
                Space-separated list of label values to include. By default, 
                the overlap will be computed between all labels available in 
                both input volumes.

        -n, --names  [LABELNAME...]
                Custom label names corresponding to the values specified with 
                the --labels flag.

        -f, --labelfile  [FILE]
                Text file specifying the label values to include. This file 
                must be in the format of a freesurfer lookup-table, in which 
                each line contains a label value and, optionally, a 
                corresponding label name.

        -x, --no-names
                Do not report label names.

        -s, --seg
                Compute overlap between the the major segmentation structures 
                listed above.

        -q, --quiet
                Quiet mode - do not print results to stdout.

JSON

NOTE: if the --seg flag is used, an additional statistic called weighted-subcortical-mean, which ignores wm and cortex, will be reported for each measure. 

{
    "inputs": [
        "/path/to/aseg.mgz",
        "/path/to/mseg.mgz"
    ],
    "measures": {
        "dice": {
            "labels": {
                "10": 0.9075892566820576,
                "11": 0.8953675962203272,
                "12": 0.9146751935375295,
                ...
            },
            "mean": 0.6805390452588164,
            "std": 0.3237900375396044,
            "weighted-mean": 0.9340674985029207
        },
        "jaccard": {
            "labels": {
                "10": 0.8308131920466723,
                "11": 0.8105570623826414,
                "12": 0.8427663203597457,
                ...
            },
            "mean": 0.5892475078628221,
            "std": 0.3029904264784899,
            "weighted-mean": 0.8801749135061013
        }
    },
    "names": {
        "10": "Left-Thalamus",
        "11": "Left-Caudate",
        "12": "Left-Putamen",
        ...
    }
}

mri_overlap (last edited 2018-09-24 11:22:21 by AndrewHoopes)