Merge pull request #439 from jdkent/fix/resampling

[FIX,REF] start changing how to handle resampling

Author: tyarkoni

nimare/base.py

@@ -9,7 +9,10 @@
 from abc import ABCMeta, abstractmethod
 from collections import defaultdict
 
+import nibabel as nb
 import numpy as np
+from nilearn._utils.niimg_conversions import _check_same_fov
+from nilearn.image import concat_imgs, resample_to_img
 
 from .results import MetaResult
 from .utils import get_masker
@@ -283,14 +286,44 @@ def __init__(self, *args, **kwargs):
             mask = get_masker(mask)
         self.masker = mask
 
+        self.resample = kwargs.get("resample", False)
+
+        # defaults for resampling images (nilearn's defaults do not work well)
+        self._resample_kwargs = {"clip": True, "interpolation": "linear"}
+        self._resample_kwargs.update(
+            {k.split("resample__")[1]: v for k, v in kwargs.items() if k.startswith("resample__")}
+        )
+
     def _preprocess_input(self, dataset):
         """Preprocess inputs to the Estimator from the Dataset as needed."""
         masker = self.masker or dataset.masker
+
+        mask_img = masker.mask_img or masker.labels_img
+        if isinstance(mask_img, str):
+            mask_img = nb.load(mask_img)
+
         for name, (type_, _) in self._required_inputs.items():
             if type_ == "image":
+                # If no resampling is requested, check if resampling is required
+                if not self.resample:
+                    check_imgs = {img: nb.load(img) for img in self.inputs_[name]}
+                    _check_same_fov(**check_imgs, reference_masker=mask_img, raise_error=True)
+                    imgs = list(check_imgs.values())
+                else:
+                    # resampling will only occur if shape/affines are different
+                    # making this harmless if all img shapes/affines are the same
+                    # as the reference
+                    imgs = [
+                        resample_to_img(nb.load(img), mask_img, **self._resample_kwargs)
+                        for img in self.inputs_[name]
+                    ]
+
+                # input to NiFtiLabelsMasker must be 4d
+                img4d = concat_imgs(imgs, ensure_ndim=4)
+
                 # Mask required input images using either the dataset's mask or
                 # the estimator's.
-                temp_arr = masker.transform(self.inputs_[name])
+                temp_arr = masker.transform(img4d)
 
                 # An intermediate step to mask out bad voxels. Can be dropped
                 # once PyMARE is able to handle masked arrays or missing data.

nimare/tests/conftest.py

@@ -6,6 +6,8 @@
 
 import pytest
 import nibabel as nib
+import numpy as np
+from nilearn.image import resample_img
 
 import nimare
 from nimare.tests.utils import get_test_data_path
@@ -79,3 +81,43 @@ def mni_mask():
     return nib.load(
         os.path.join(get_resource_path(), "templates", "MNI152_2x2x2_brainmask.nii.gz")
     )
+
+
+@pytest.fixture(scope="session")
+def testdata_ibma_resample(tmp_path_factory):
+    tmpdir = tmp_path_factory.mktemp("testdata_ibma_resample")
+
+    # Load dataset
+    dset_file = os.path.join(get_test_data_path(), "test_pain_dataset.json")
+    dset_dir = os.path.join(get_test_data_path(), "test_pain_dataset")
+    mask_file = os.path.join(dset_dir, "mask.nii.gz")
+    dset = nimare.dataset.Dataset(dset_file, mask=mask_file)
+    dset.update_path(dset_dir)
+
+    # create reproducible random number generator for resampling
+    rng = np.random.default_rng(seed=123)
+    # Move image contents of Dataset to temporary directory
+    for c in dset.images.columns:
+        if c.endswith("__relative"):
+            continue
+        for f in dset.images[c].values:
+            if (f is None) or not os.path.isfile(f):
+                continue
+            new_f = f.replace(
+                dset_dir.rstrip(os.path.sep), str(tmpdir.absolute()).rstrip(os.path.sep)
+            )
+            dirname = os.path.dirname(new_f)
+            if not os.path.isdir(dirname):
+                os.makedirs(dirname)
+            # create random affine to make images different shapes
+            affine = np.eye(3)
+            np.fill_diagonal(affine, rng.choice([1, 2, 3]))
+            img = resample_img(
+                nib.load(f),
+                target_affine=affine,
+                interpolation="linear",
+                clip=True,
+            )
+            nib.save(img, new_f)
+    dset.update_path(tmpdir)
+    return dset

nimare/tests/test_meta_ibma.py

@@ -2,8 +2,10 @@
 Test nimare.meta.ibma (image-based meta-analytic estimators).
 """
 import os.path as op
+from contextlib import ExitStack as does_not_raise
 
 from nilearn.input_data import NiftiLabelsMasker
+import pytest
 
 import nimare
 from nimare.correct import FDRCorrector, FWECorrector
@@ -135,3 +137,24 @@ def test_ibma_with_custom_masker(testdata_ibma):
     meta.fit(testdata_ibma)
     assert isinstance(meta.results, nimare.results.MetaResult)
     assert meta.results.maps["z"].shape == (5,)
+
+
+@pytest.mark.parametrize(
+    "resample,resample_kwargs,expectation",
+    [
+        (False, {}, pytest.raises(ValueError)),
+        (None, {}, pytest.raises(ValueError)),
+        (True, {}, does_not_raise()),
+        (
+            True,
+            {"resample__clip": False, "resample__interpolation": "continuous"},
+            does_not_raise(),
+        ),
+    ],
+)
+def test_ibma_resampling(testdata_ibma_resample, resample, resample_kwargs, expectation):
+    meta = ibma.Fishers(resample=resample, **resample_kwargs)
+    with expectation:
+        meta.fit(testdata_ibma_resample)
+    if isinstance(expectation, does_not_raise):
+        assert isinstance(meta.results, nimare.results.MetaResult)