curvanato 0.0.0

Anatomical curvature processing for medical images.

curvanato

local curvature quantification for anatomical data

Installation & Testing

To install the package locally, navigate to the package root and run:

pip install .

Usage

import ants
import antspynet
import antspyt1w
import curvanato
import re
import os  # For checking file existence
import pandas as pd
import numpy as np
os.environ["ITK_GLOBAL_DEFAULT_NUMBER_OF_THREADS"] = "32"
# ANTPD data from open neuro
fn='.//bids/sub-RC4111/ses-1/anat/sub-RC4111_ses-1_T1w.nii.gz' # easy
fn='.//bids/sub-RC4103/ses-1/anat/sub-RC4103_ses-1_T1w.nii.gz'
fn='./bids//sub-RC4110/ses-2/anat/sub-RC4110_ses-2_T1w.nii.gz'
if os.path.exists(fn):
    t1=ants.image_read( fn )
    t1=ants.resample_image( t1, [0.5, 0.5, 0.5], use_voxels=False, interp_type=0 )
    hoafn = re.sub( ".nii.gz", "_hoa.nii.gz" , fn )
    if not os.path.exists(hoafn):
        hoa = antspynet.harvard_oxford_atlas_labeling(t1, verbose=True)['segmentation_image']
        ants.image_write( hoa, hoafn)
    hoa = ants.image_read( hoafn )
    citfn = re.sub( ".nii.gz", "_cit168.nii.gz" , fn )
    if not os.path.exists(citfn):
        t1b = antspynet.brain_extraction( t1, modality="t1threetissue" )['segmentation_image'].threshold_image(1,1)
        t1r = ants.rank_intensity( t1 * t1b )
        cit = antspyt1w.deep_cit168( t1r, verbose=True)['segmentation']
        ants.image_write( cit, citfn)
    cit = ants.image_read( citfn )

###################################################################################
###################################################################################
###################################################################################
###################################################################################
ctype='cit'
tcaudL=curvanato.load_labeled_caudate( option='hmt', binarize=False, label=[1,3,5] )
tcaudR=curvanato.load_labeled_caudate( option='hmt', binarize=False, label=[2,4,6] )
vlab=None
leftside=True
gr=0
subd=0
otherside=True
if otherside:
    ccfn = [
        re.sub( ".nii.gz", "_"+ctype+"Rkappa.nii.gz" , fn ), 
        re.sub( ".nii.gz", "_"+ctype+"R.nii.gz" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Rthk.nii.gz" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Rkappa.csv" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Rkappa.png" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Rthk.png" , fn ) ]
    pcaud=[3,4]
    plabs=[4]
    if ctype == 'cit':
        mytl=18
    xx = curvanato.t1w_caudcurv(  cit, target_label=mytl, ventricle_label=vlab, 
        prior_labels=pcaud, prior_target_label=plabs, subdivide=subd, grid=gr,
        priorparcellation=tcaudR,  plot=True,
        verbose=True )
    ants.plot( xx[0], xx[1], crop=True, axis=2, nslices=21, ncol=7, filename=ccfn[4] )
    ants.plot( xx[0], xx[2], crop=True, axis=2, nslices=21, ncol=7, filename=ccfn[5] )
    for j in range(3):
        ants.image_write( xx[j], ccfn[j] )
    xx[3].to_csv( ccfn[3] )

if leftside:
    mytl=2
    ccfn = [
        re.sub( ".nii.gz", "_"+ctype+"Lkappa.nii.gz" , fn ), 
        re.sub( ".nii.gz", "_"+ctype+"L.nii.gz" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Lthk.nii.gz" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Lkappa.csv" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Lkappa.png" , fn ),
        re.sub( ".nii.gz", "_"+ctype+"Lthk.png" , fn ) ]
    print("Begin " + fn + " caud kap")
    pcaud=[1,2]
    plabs=[2]
    xx = curvanato.t1w_caudcurv( cit, target_label=2, ventricle_label=vlab, 
        prior_labels=pcaud, prior_target_label=plabs, subdivide=subd, grid=gr,
        priorparcellation=tcaudL,  plot=True, searchrange=20,
        verbose=True )
    ants.plot( xx[0], xx[1], crop=True, axis=2, nslices=21, ncol=7, filename=ccfn[4] )
    ants.plot( xx[0], xx[1], crop=True, axis=2, nslices=21, ncol=7, filename=ccfn[5] )
    for j in range(3):
        ants.image_write( xx[j], ccfn[j] )
    xx[3].to_csv( ccfn[3] )

Example data

this package has been tested on ANTPD data from openneuro.

could also try data here which included repeated T1w acquisitions on same subjects but with different parameters. however, last time i tried this, the link was not working.