Closed thenineteen closed 3 years ago
According to the color table generated by NiftyWeb, of which we have an adaptation in this repo, there are 156 possible structures/parcels/segments (after discarding the first five).
That doesn't mean all are present in the parcellation/segmentation we are using. There is some information about this in my comments in #184.
The parcellation we are using seems to have 137 different structures, discarding the first five as before:
In [1]: import numpy as np
In [2]: import nibabel as nib
In [3]: data = nib.load('slicer/Resources/Image/MNI_152_gif.nii.gz').get_data()
In [4]: len(np.unique(data[data > 16]))
Out[4]: 137
I get different numbers depending on the exact query, but some of these GIF parcellations must be ignored then if we only have 137:
The below gives 146
from mega_analysis import Semiology, Laterality
patient = Semiology('Tonic',
symptoms_side=Laterality.LEFT,
dominant_hemisphere=Laterality.LEFT,
normalise_to_localising_values=False, # default is False
)
patient.granular = False
patient.top_level_lobes = True
heatmap = patient.get_num_datapoints_dict()
len(heatmap)
Additionally, there are 115 parcellations when I looked at marginal probabilities for each GIF, of which the following are zero:
39 | 40 | 41 | 42 | 72 | 73 | 74 |
---|
leaving around 108 non-zero
What is the total number of different GIF parcellations supported by our programme? per hemisphere and in total please
e.g. 80 per hemisphere = 160 in total.
I need this to calculate the naive baseline probability of a semiology to localise to a given brain region, as benchmark for summary forrest plots