[MAINT] Store 1 timepoint for 4D generating a report with the maskers

doc/changes/latest.rst

@@ -24,6 +24,7 @@ Enhancements
 - :bdg-success:`API` Allow passing Pandas Series of image filenames to :class:`~nilearn.glm.second_level.SecondLevelModel` (:gh:`4070` by `Rémi Gau`_).
 - :bdg-info:`Plotting` Allow setting ``vmin`` in :func:`~nilearn.plotting.plot_glass_brain` and :func:`~nilearn.plotting.plot_stat_map` (:gh:`3993` by `Michelle Wang`_).
 - :bdg-success:`API` Support passing t and F contrasts to :func:`~nilearn.glm.compute_contrast` that that have fewer columns than the number of estimated parameters. Remaining columns are padded with zero (:gh:`4067` by `Rémi Gau`_).
+- :bdg-dark:`Code` Multi-subject maskers' ``generate_report`` method no longer fails with 5D data but instead shows report of first subject. User can index input list to show report for different subjects (:gh:`3935` by `Yasmin Mzayek`_).
 
 Changes
 -------

nilearn/glm/tests/test_second_level.py

@@ -598,39 +598,34 @@ def confounds():
     )
 
 
-def test_fmri_inputs(tmp_path, rng, confounds):
+def test_fmri_inputs(
+    tmp_path, rng, confounds, shape_3d_default, shape_4d_default
+):
     # Test processing of FMRI inputs
     # prepare fake data
-    p, q = 80, 10
-    X = rng.standard_normal(size=(p, q))
-    shapes = ((7, 8, 9, 10),)
-    mask, FUNCFILE, _ = write_fake_fmri_data_and_design(
-        shapes, file_path=tmp_path
+    mask, niimg, des = generate_fake_fmri_data_and_design(
+        [shape_4d_default], 1
     )
-    FUNCFILE = FUNCFILE[0]
-    func_img = load(FUNCFILE)
-    T = func_img.shape[-1]
-    des = pd.DataFrame(np.ones((T, 1)), columns=["a"])
-    des_fname = str(tmp_path / "design.csv")
-    des.to_csv(des_fname)
 
     # prepare correct input first level models
-    flm = FirstLevelModel(subject_label="01").fit(
-        FUNCFILE, design_matrices=des
-    )
+    flm = FirstLevelModel(subject_label="01").fit(niimg, design_matrices=des)
 
     # prepare correct input dataframe and lists
-    shapes = (SHAPE,)
-    _, FUNCFILE, _ = write_fake_fmri_data_and_design(
-        shapes, file_path=tmp_path
-    )
-    FUNCFILE = FUNCFILE[0]
-
+    p, q = 80, 10
+    X = rng.standard_normal(size=(p, q))
     sdes = pd.DataFrame(X[:3, :3], columns=["intercept", "b", "c"])
 
     # smoke tests with correct input
     flms = [flm, flm, flm]
 
+    shape_3d = [shape_3d_default + (1,)]
+    _, FUNCFILE, _ = write_fake_fmri_data_and_design(
+        shape_3d, file_path=tmp_path
+    )
+    FUNCFILE = FUNCFILE[0]
+    niimgs = [FUNCFILE, FUNCFILE, FUNCFILE]
+    niimg_4d = concat_imgs(niimgs)
+
     # First level models as input
     SecondLevelModel(mask_img=mask).fit(flms)
     SecondLevelModel().fit(flms)
@@ -639,11 +634,9 @@ def test_fmri_inputs(tmp_path, rng, confounds):
     SecondLevelModel().fit(flms, None, sdes)
 
     # niimgs as input
-    niimgs = [FUNCFILE, FUNCFILE, FUNCFILE]
     SecondLevelModel().fit(niimgs, None, sdes)
 
     # 4d niimg as input
-    niimg_4d = concat_imgs(niimgs)
     SecondLevelModel().fit(niimg_4d, None, sdes)
 
 
@@ -787,40 +780,27 @@ def test_fmri_img_inputs_errors(tmp_path, confounds):
 
 
 def test_fmri_inputs_for_non_parametric_inference_errors(
-    tmp_path, rng, confounds
+    tmp_path, rng, confounds, shape_3d_default, shape_4d_default
 ):
     # Test processing of FMRI inputs
-
     # prepare fake data
-    p, q = 80, 10
-    X = rng.standard_normal(size=(p, q))
-    shapes = ((7, 8, 9, 10),)
-    _, func_file, _ = write_fake_fmri_data_and_design(
-        shapes, file_path=tmp_path
-    )
-
-    func_file = func_file[0]
-
-    func_img = load(func_file)
-    T = func_img.shape[-1]
-    des = pd.DataFrame(np.ones((T, 1)), columns=["a"])
-    des_fname = str(tmp_path / "design.csv")
-    des.to_csv(des_fname)
+    _, niimg, des = generate_fake_fmri_data_and_design([shape_4d_default], 1)
 
     # prepare correct input first level models
-    flm = FirstLevelModel(subject_label="01").fit(
-        func_file, design_matrices=des
-    )
+    flm = FirstLevelModel(subject_label="01").fit(niimg, design_matrices=des)
+
     # prepare correct input dataframe and lists
-    shapes = (SHAPE,)
-    _, func_file, _ = write_fake_fmri_data_and_design(
-        shapes, file_path=tmp_path
-    )
-    func_file = func_file[0]
+    p, q = 80, 10
+    X = rng.standard_normal(size=(p, q))
+    sdes = pd.DataFrame(X[:3, :3], columns=["intercept", "b", "c"])
 
-    niimgs = [func_file, func_file, func_file]
+    shape_3d = [shape_3d_default + (1,)]
+    _, FUNCFILE, _ = write_fake_fmri_data_and_design(
+        shape_3d, file_path=tmp_path
+    )
+    FUNCFILE = FUNCFILE[0]
+    niimgs = [FUNCFILE, FUNCFILE, FUNCFILE]
     niimg_4d = concat_imgs(niimgs)
-    sdes = pd.DataFrame(X[:3, :3], columns=["intercept", "b", "c"])
 
     # test missing second-level contrast
     match = "No second-level contrast is specified."
@@ -840,7 +820,7 @@ def test_fmri_inputs_for_non_parametric_inference_errors(
 
     # test list of less than two niimgs
     with pytest.raises(TypeError, match="at least two"):
-        non_parametric_inference([func_file])
+        non_parametric_inference([FUNCFILE])
 
     # test niimgs requirements
     with pytest.raises(ValueError, match="require a design matrix"):

nilearn/maskers/__init__.py

@@ -1,5 +1,6 @@
 """The :mod:`nilearn.maskers` contains masker objects."""
 
+from ._utils import compute_middle_image
 from .base_masker import BaseMasker
 from .multi_nifti_labels_masker import MultiNiftiLabelsMasker
 from .multi_nifti_maps_masker import MultiNiftiMapsMasker
@@ -18,4 +19,5 @@
     "NiftiMapsMasker",
     "MultiNiftiMapsMasker",
     "NiftiSpheresMasker",
+    "compute_middle_image",
 ]

nilearn/maskers/_utils.py

@@ -0,0 +1,24 @@
+from nilearn import image
+
+
+def _check_dims(imgs):
+    # check dims of one image if given a list
+    if isinstance(imgs, list):
+        im = imgs[0]
+        dim = image.load_img(im).shape
+        # in case of 4D (timeseries) + 1D (subjects) return first subject
+        if len(dim) == 4:
+            return im, dim + (1,)
+        else:
+            return imgs, dim + (1,)
+    else:
+        dim = image.load_img(imgs).shape
+        return imgs, dim
+
+
+def compute_middle_image(img):
+    """Compute middle image of timeseries (4D data)."""
+    img, dim = _check_dims(img)
+    if len(dim) == 4 or len(dim) == 5:
+        img = image.index_img(img, dim[-1] // 2)
+    return img, len(dim)

nilearn/maskers/multi_nifti_masker.py

@@ -10,6 +10,7 @@
 from joblib import Memory, Parallel, delayed
 
 from nilearn import _utils, image, masking
+from nilearn.maskers import compute_middle_image
 from nilearn.maskers.nifti_masker import NiftiMasker, _filter_and_mask
 
 
@@ -240,8 +241,15 @@ def fit(self, imgs=None, y=None):
 
         self._reporting_data = None
         if self.reports:  # save inputs for reporting
-            imgs = imgs[0] if isinstance(imgs, list) else imgs
-            self._reporting_data = {"images": imgs, "mask": self.mask_img_}
+            self._reporting_data = {
+                "mask": self.mask_img_,
+                "dim": None,
+                "images": imgs,
+            }
+            if imgs is not None:
+                imgs, dims = compute_middle_image(imgs)
+                self._reporting_data["images"] = imgs
+                self._reporting_data["dim"] = dims
 
         # If resampling is requested, resample the mask as well.
         # Resampling: allows the user to change the affine, the shape or both.

nilearn/maskers/nifti_labels_masker.py

@@ -6,6 +6,7 @@
 from joblib import Memory
 
 from nilearn import _utils, image, masking
+from nilearn.maskers import compute_middle_image
 from nilearn.maskers.base_masker import BaseMasker, _filter_and_extract
 
 
@@ -290,12 +291,16 @@
             self._report_content['summary'] = regions_summary
 
             img = self._reporting_data['img']
+
             # If we have a func image to show in the report, use it
             if img is not None:
-                dim = image.load_img(img).shape
-                if len(dim) == 4:
-                    # compute middle image from 4D series for plotting
-                    img = image.index_img(img, dim[-1] // 2)
+                if self._reporting_data["dim"] == 5:
+                    msg = (
+                        "A list of 4D subject images were provided to fit. "
+                        "Only first subject is shown in the report."
+                    )
+                    warnings.warn(msg)
+                    self._report_content['warning_message'] = msg
                 display = plotting.plot_img(
                     img,
                     black_bg=False,
@@ -411,10 +416,15 @@
 
         if self.reports:
             self._reporting_data = {
-                'labels_image': self._resampled_labels_img_,
-                'mask': self.mask_img_,
-                'img': imgs,
+                "labels_image": self._resampled_labels_img_,
+                "mask": self.mask_img_,
+                "dim": None,
+                "img": imgs,
             }
+            if imgs is not None:
+                imgs, dims = compute_middle_image(imgs)
+                self._reporting_data["img"] = imgs
+                self._reporting_data["dim"] = dims
         else:
             self._reporting_data = None
 

nilearn/maskers/nifti_maps_masker.py

@@ -5,6 +5,7 @@
 from joblib import Memory
 
 from nilearn import _utils, image
+from nilearn.maskers import compute_middle_image
 from nilearn.maskers.base_masker import BaseMasker, _filter_and_extract
 
 
@@ -326,11 +327,13 @@
                 display.close()
             return embeded_images
 
-        dim = image.load_img(img).shape
-        if len(dim) == 4:
-            # compute middle image from 4D series for plotting
-            img = image.index_img(img, dim[-1] // 2)
-
+        if self._reporting_data["dim"] == 5:
+            msg = (
+                "A list of 4D subject images were provided to fit. "
+                "Only first subject is shown in the report."
+            )
+            warnings.warn(msg)
+            self._report_content["warning_message"] = msg
         # Find the cut coordinates
         cut_coords = [
             plotting.find_xyz_cut_coords(image.index_img(maps_image, i))
@@ -425,8 +428,13 @@
             self._reporting_data = {
                 "maps_image": self.maps_img_,
                 "mask": self.mask_img_,
+                "dim": None,
                 "img": imgs,
             }
+            if imgs is not None:
+                imgs, dims = compute_middle_image(imgs)
+                self._reporting_data["img"] = imgs
+                self._reporting_data["dim"] = dims
         else:
             self._reporting_data = None
 

nilearn/maskers/nifti_masker.py

@@ -8,6 +8,7 @@
 from joblib import Memory
 
 from nilearn import _utils, image, masking
+from nilearn.maskers import compute_middle_image
 from nilearn.maskers.base_masker import BaseMasker, _filter_and_extract
 
 
@@ -338,21 +339,21 @@
         img = self._reporting_data["images"]
         mask = self._reporting_data["mask"]
 
-        if img is not None:
-            dim = image.load_img(img).shape
-            if len(dim) == 4:
-                # compute middle image from 4D series for plotting
-                img = image.index_img(img, dim[-1] // 2)
-
-        else:  # images were not provided to fit
+        if img is None:  # images were not provided to fit
             msg = (
                 "No image provided to fit in NiftiMasker. "
                 "Setting image to mask for reporting."
             )
             warnings.warn(msg)
             self._report_content["warning_message"] = msg
             img = mask
-
+        if self._reporting_data["dim"] == 5:
+            msg = (
+                "A list of 4D subject images were provided to fit. "
+                "Only first subject is shown in the report."
+            )
+            warnings.warn(msg)
+            self._report_content["warning_message"] = msg
         # create display of retained input mask, image
         # for visual comparison
         init_display = plotting.plot_img(
@@ -376,13 +377,9 @@
         self._report_content["description"] += self._overlay_text
 
         # create display of resampled NiftiImage and mask
-        # assuming that resampl_img has same dim as img
         resampl_img, resampl_mask = self._reporting_data["transform"]
         if resampl_img is None:  # images were not provided to fit
             resampl_img = resampl_mask
-        elif len(dim) == 4:
-            # compute middle image from 4D series for plotting
-            resampl_img = image.index_img(resampl_img, dim[-1] // 2)
 
         final_display = plotting.plot_img(
             resampl_img,
@@ -442,7 +439,15 @@
             self.mask_img_ = _utils.check_niimg_3d(self.mask_img)
 
         if self.reports:  # save inputs for reporting
-            self._reporting_data = {"images": imgs, "mask": self.mask_img_}
+            self._reporting_data = {
+                "mask": self.mask_img_,
+                "dim": None,
+                "images": imgs,
+            }
+            if imgs is not None:
+                imgs, dims = compute_middle_image(imgs)
+                self._reporting_data["images"] = imgs
+                self._reporting_data["dim"] = dims
         else:
             self._reporting_data = None
 
@@ -485,6 +490,7 @@
                     copy=False,
                     interpolation="nearest",
                 )
+                resampl_imgs, _ = compute_middle_image(resampl_imgs)
             else:  # imgs not provided to fit
                 resampl_imgs = None