RegionalStats#

class petpal.utils.stats.RegionalStats(input_image_path: str, segmentation_image_path: str, label_map_option: str | dict)#

Run statistics on each region in a parametric 3D PET kinetic model or other image.

Options:

RegionalStats.mean: Mean value within each region
RegionalStats.std: Standard devation of values within each region.
RegionalStats.nvox: Number of voxels in each region.
RegionalStats.max: Maximum value in each region.
RegionalStats.min: Minimum value in each region.
RegionalStats.median: Median value in each region.
RegionalStats.get_stats(stats_func): Get a generic statistic run on each region. Runs function stats_func on each region, which must take a 1D array as the only argument.

Example

import numpy as np
from petpal.utils.stats import RegionalStats
from petpal.utils.image_io import write_dict_to_json

# Set up class
input_image_path = 'sub-001_ses-01_space-mpr_desc-SUVR_pet.nii.gz'
segmentation_image_path = 'sub-001_ses-01_seg.nii.gz'
region_stats_obj = RegionalStats(input_image_path = input_image_path,
                                 segmentation_image_path = segmentation_image_path,
                                 label_map_option = 'freesurfer')

# Preset statistic: get mean in each region
region_means = region_stats_obj.mean
write_dict_to_json(region_means,'sub-001_ses-01_RegionMeanSUVR.json')

# Create function to get 95th percentile value for each region
def calc_95th_percentile(arr: np.ndarray):
    return np.percentile(arr,95)
region_95th = region_stats_obj.get_stats(calc_95th_percentile)
write_dict_to_json(region_95th,'sub-001_ses-01_Region95thPercentileSUVR.json')

Variables:

pet_img – 3D PET image on which to get statistics for each region.
seg_img – Segmentation image in same space as pet_img defining regions.
label_map – Dictionary that assigns labels to regions in seg_img.

get_voxels(label: str) → numpy.ndarray#

Get the voxel array for the selected label.

Parameters:: label (str) – Name of the region from which to extract voxels.
Returns:: voxel_arr (np.ndarray) – Voxels in region as a flattened array.

get_stats(stats_func: collections.abc.Callable, dtype: object = float) → dict#

Get stats for all regions. Applies stats_func to the voxel_arr to return a value for each region. Set dtype depending on what type the output is (e.g. float or int).

Parameters:

stats_func (Callable) – The function to run on each region’s voxels. Must take a 1D array as the only required positional argument and return a value or np.ndarray.
dtype (object) – The dtype that stats_func returns, if not float. Typically float or int. Default float.

Returns:

region_stats (dict) – The statistics for each region of interest.

property mean: dict#: Get mean value for each region.

property std: dict#: Get standard deviation of values for each region.

property nvox: dict#: Get number of voxels in each region.

property max: dict#: Get maximum value in each region.

property min: dict#: Get minimum value in each region.

property median: dict#: Get median value in each region.

RegionalStats#

This Page