RF: Use `mc/prefiltered_func_data_mcf_conf_hp.nii.gz` as motion confounds if it exists; save classifier probabilities to `fix4melview` file

• If the mc/prefiltered_func_data_mcf_conf_hp.nii.gz already exists in the input data directory, use it as the motion confounds, instead of loading and high-pass filtering the MCFLIRT results from mc/prefiltered_func_data_mcf.par. This is required to support HCP scripts which perform their own filtering of motion confounds, and save them to mc/prefiltered_...._hp.nii.gz. Reported on the FSL mailing list: https://www.jiscmail.ac.uk/cgi-bin/wa-jisc.exe?A2=FSL;199b80ab.2409
• Classifier probabilities are now saved to the fix4melview label file. The original FIX was adjusted in 1.069 to do this, but this feature was never ported to pyfix. This was probably because the version of FIX on the FMRIB cluster was never updated beyond version 1.067.

pyfix/tests/__init__.py

@@ -11,6 +11,11 @@ import               sys
 import               zipfile
 from unittest import mock
 
+import numpy as np
+
+from fsl.data.image      import Image
+from fsl.utils.image.roi import roi
+
 from pyfix import fix, feature
 from pyfix.util import indir
 
@@ -90,9 +95,13 @@ def link_directory(src, dest, exclude=None):
             os.symlink(op.join(src, filename), op.join(dest, filename))
 
 
-def unzip(infile, destdir=None):
+def unzip(infile, destdir=None, paths=None):
     with zipfile.ZipFile(infile) as f:
-        f.extractall(destdir)
+        if paths is None:
+            f.extractall(destdir)
+        else:
+            for path in paths:
+                f.extract(path, destdir)
 
 
 def main(cmd=''):
@@ -132,3 +141,42 @@ def mock_features(feature_hook=None):
 
     with mock.patch('pyfix.feature.FEATURE_EXTRACTORS', funcs):
         yield
+
+
+def compare_arrays(got, exp, err=0.01):
+    exprange = exp.max() - exp.min()
+    maxerr   = np.abs(got - exp).max() / exprange
+    assert maxerr < err
+
+
+def compare_4D_data(got, exp, err=0.01, corr=0.99):
+
+    assert np.all(got.shape == exp.shape)
+    compare_arrays(got, exp, err)
+
+    got      = got.reshape(-1, got.shape[-1])
+    exp      = exp.reshape(-1, exp.shape[-1])
+    nvoxels  = got.shape[0]
+    corrs    = np.zeros(nvoxels)
+
+    for i in range(nvoxels):
+        corrs[i] = np.corrcoef(got[i], exp[i])[0, 1]
+
+    assert corrs.min() > corr
+
+
+def crop(data, params):
+    xlo, xlen, ylo, ylen, zlo, zlen = params[:6]
+    return data[xlo:xlo + xlen,
+                ylo:ylo + ylen,
+                zlo:zlo + zlen]
+
+
+def crop_image(infile, outfile, params):
+
+    xlo, xlen, ylo, ylen, zlo, zlen = params[:6]
+    image   = Image(infile)
+    cropped = roi(image, [(xlo, xlo + xlen),
+                          (ylo, ylo + ylen),
+                          (zlo, zlo + zlen)])
+    cropped.save(outfile)