[ENH] Skip processed visits for t1-volume-tissue-segmentation

clinica/pipelines/t1_volume_tissue_segmentation/t1_volume_tissue_segmentation_pipeline.py

@@ -3,6 +3,8 @@
 from nipype import config
 
 from clinica.pipelines.engine import Pipeline
+from clinica.utils.bids import Visit
+from clinica.utils.filemanip import extract_visits
 
 cfg = dict(execution={"parameterize_dirs": False})
 config.update_config(cfg)
@@ -63,6 +65,78 @@ def get_output_fields(self) -> List[str]:
             "t1_mni",
         ]
 
+    def get_processed_visits(self) -> list[Visit]:
+        """Return a list of visits for which the pipeline is assumed to have run already.
+
+        Before running the pipeline, for a given visit, if both the PET SUVR registered image
+        and the rigid transformation files already exist, then the visit is added to this list.
+        The pipeline will further skip these visits and run processing only for the remaining
+        visits.
+        """
+        from functools import reduce
+
+        from clinica.utils.filemanip import extract_visits
+        from clinica.utils.input_files import (
+            t1_volume_dartel_input_tissue,
+            t1_volume_native_tpm,
+            t1_volume_native_tpm_in_mni,
+        )
+        from clinica.utils.inputs import clinica_file_reader
+
+        if not self.caps_directory.is_dir():
+            return []
+        visits = [
+            set(extract_visits(x[0]))
+            for x in [
+                clinica_file_reader(
+                    self.subjects,
+                    self.sessions,
+                    self.caps_directory,
+                    pattern,
+                )
+                for pattern in [
+                    t1_volume_dartel_input_tissue(tissue_number=i)
+                    for i in self.parameters["dartel_tissues"]
+                ]
+            ]
+        ]
+        visits.extend(
+            [
+                set(extract_visits(x[0]))
+                for x in [
+                    clinica_file_reader(
+                        self.subjects,
+                        self.sessions,
+                        self.caps_directory,
+                        pattern,
+                    )
+                    for pattern in [
+                        t1_volume_native_tpm(tissue_number=i)
+                        for i in self.parameters["tissue_classes"]
+                    ]
+                ]
+            ]
+        )
+        visits.extend(
+            [
+                set(extract_visits(x[0]))
+                for x in [
+                    clinica_file_reader(
+                        self.subjects,
+                        self.sessions,
+                        self.caps_directory,
+                        pattern,
+                    )
+                    for pattern in [
+                        t1_volume_native_tpm_in_mni(tissue_number=i, modulation=False)
+                        for i in self.parameters["tissue_classes"]
+                    ]
+                ]
+            ]
+        )
+
+        return sorted(list(reduce(lambda x, y: x.intersection(y), visits)))
+
     def _build_input_node(self):
         """Build and connect an input node to the pipeline.
 

clinica/utils/testing_utils.py

@@ -10,6 +10,8 @@
 
 from clinica.pipelines.dwi.utils import DWIDataset
 
+from .spm import SPMTissue, get_spm_tissue_from_index
+
 __all__ = [
     "build_test_image_cubic_object",
     "build_bids_directory",
@@ -209,15 +211,84 @@ def _build_subjects(directory: Path, configuration: dict) -> None:
                 for pipeline_name, pipeline_config in configuration[
                     "pipelines"
                 ].items():
-                    (directory / "subjects" / sub / ses / pipeline_name).mkdir(
-                        exist_ok=True
-                    )
+                    if pipeline_name.startswith("t1_volume"):
+                        (directory / "subjects" / sub / ses / "t1").mkdir(exist_ok=True)
+                    else:
+                        (directory / "subjects" / sub / ses / pipeline_name).mkdir(
+                            exist_ok=True
+                        )
                     if pipeline_name == "t1_linear":
                         _build_t1_linear(directory, sub, ses, pipeline_config)
                     if pipeline_name == "pet_linear":
                         _build_pet_linear(directory, sub, ses, pipeline_config)
                     if pipeline_name == "t1":
                         _build_t1(directory, sub, ses, configuration)
+                    if pipeline_name == "t1_volume_tissue_segmentation":
+                        _build_t1_volume_tissue_segmentation(
+                            directory, sub, ses, pipeline_config
+                        )
+
+
+def _build_t1_volume_tissue_segmentation(
+    directory: Path, sub: str, ses: str, config: dict
+) -> None:
+    """Build a fake t1-volume-tissue-segmentation file structure in a CAPS directory."""
+    segmentation_folder = (
+        directory / "subjects" / sub / ses / "t1" / "spm" / "segmentation"
+    )
+    segmentation_folder.mkdir(parents=True, exist_ok=True)
+    common_filename_part = f"{sub}_{ses}_T1w_segm-"
+    _build_t1_volume_tissue_segmentation_native_space(
+        segmentation_folder, common_filename_part, config
+    )
+    _build_t1_volume_tissue_segmentation_dartel_input(
+        segmentation_folder, common_filename_part, config
+    )
+    _build_t1_volume_tissue_segmentation_normalized_space(
+        segmentation_folder, common_filename_part, config
+    )
+
+
+def _extract_tissues_from_config(config: dict, key: str) -> list[SPMTissue]:
+    if (tissue_classes := config.get(key, None)) is not None:
+        return [get_spm_tissue_from_index(i) for i in tissue_classes]
+    return [tissue for tissue in SPMTissue]
+
+
+def _build_t1_volume_tissue_segmentation_native_space(
+    segmentation_folder: Path, common_filename_part: str, config: dict
+) -> None:
+    (segmentation_folder / "native_space").mkdir(exist_ok=True)
+    for tissue in _extract_tissues_from_config(config, "tissue_classes"):
+        (
+            segmentation_folder
+            / "native_space"
+            / f"{common_filename_part}{tissue.value}_probability.nii.gz"
+        ).touch()
+
+
+def _build_t1_volume_tissue_segmentation_dartel_input(
+    segmentation_folder: Path, common_filename_part: str, config: dict
+) -> None:
+    (segmentation_folder / "dartel_input").mkdir(exist_ok=True)
+    for tissue in _extract_tissues_from_config(config, "dartel_tissues"):
+        (
+            segmentation_folder
+            / "dartel_input"
+            / f"{common_filename_part}{tissue.value}_dartelinput.nii.gz"
+        ).touch()
+
+
+def _build_t1_volume_tissue_segmentation_normalized_space(
+    segmentation_folder: Path, common_filename_part: str, config: dict
+) -> None:
+    (segmentation_folder / "normalized_space").mkdir(exist_ok=True)
+    for tissue in _extract_tissues_from_config(config, "tissue_classes"):
+        (
+            segmentation_folder
+            / "normalized_space"
+            / f"{common_filename_part}{tissue.value}_space-Ixi549Space_modulated-off_probability.nii.gz"
+        ).touch()
 
 
 def _build_t1_linear(directory: Path, sub: str, ses: str, config: dict) -> None:

test/unittests/pipelines/t1_volume_tissue_segmentation/test_t1_volume_tissue_segmentation_pipeline.py

@@ -1,4 +1,9 @@
-from clinica.utils.testing_utils import build_caps_directory
+import shutil
+
+from packaging.version import Version
+
+from clinica.utils.bids import Visit
+from clinica.utils.testing_utils import build_bids_directory, build_caps_directory
 
 
 def test_t1_volume_tissue_segmentation_info_loading(tmp_path):
@@ -28,7 +33,6 @@ def test_t1_volume_tissue_segmentation_info_loading(tmp_path):
 
 def test_t1_volume_tissue_segmentation_dependencies(tmp_path, mocker):
     from packaging.specifiers import SpecifierSet
-    from packaging.version import Version
 
     from clinica.pipelines.t1_volume_tissue_segmentation.t1_volume_tissue_segmentation_pipeline import (
         T1VolumeTissueSegmentation,
@@ -53,3 +57,158 @@ def test_t1_volume_tissue_segmentation_dependencies(tmp_path, mocker):
             ThirdPartySoftware.SPM, SpecifierSet(">=12"), Version("12.7219")
         ),
     ]
+
+
+def test_t1_volume_tissue_segmentation_get_processed_visits_empty(tmp_path, mocker):
+    from clinica.pipelines.t1_volume_tissue_segmentation.t1_volume_tissue_segmentation_pipeline import (
+        T1VolumeTissueSegmentation,
+    )
+
+    mocker.patch(
+        "clinica.utils.check_dependency._get_spm_version",
+        return_value=Version("12.7219"),
+    )
+    bids = build_bids_directory(
+        tmp_path / "bids", {"sub-01": ["ses-M000", "ses-M006"], "sub-02": ["ses-M000"]}
+    )
+    caps = build_caps_directory(tmp_path / "caps", {})
+
+    pipeline = T1VolumeTissueSegmentation(
+        bids_directory=str(bids),
+        caps_directory=str(caps),
+        parameters={
+            "tissue_classes": (1, 2, 3),
+            "dartel_tissues": (1, 2, 3),
+        },
+    )
+    assert pipeline.get_processed_visits() == []
+
+
+def test_t1_volume_tissue_segmentation_get_processed_visits(tmp_path, mocker):
+    """Test the get_processed_visits for T1VolumeTissueSegmentation.
+
+    We build a CAPS dataset with the following structure:
+
+    caps2
+    ├── dataset_description.json
+    └── subjects
+        └── sub-01
+            └── ses-M000
+                └── t1
+                    └── spm
+                        └── segmentation
+                            ├── dartel_input
+                            │         └── sub-01_ses-M000_T1w_segm-XXXXXXXX_dartelinput.nii.gz
+                            ├── native_space
+                            │         └── sub-01_ses-M000_T1w_segm-YYYYYYYY_probability.nii.gz
+                            └── normalized_space
+                                └── sub-01_ses-M000_T1w_segm-YYYYYYYY_space-Ixi549Space_modulated-off_probability.nii.gz
+
+    And this for several subjects and sessions.
+    We can control the values for XXXXX and YYYYY through the lists of tissues that we pass
+    to the CAPS generator (XXXX is controlled by 'dartel_tissues', while YYYY is controlled by 'tissue_classes').
+
+    The purpose of this test is to verify that, depending on what the user wants in terms of tissues,
+    the pipeline correctly identifies the already processed visits as the ones having ALL images for
+    the tissues of interest. If there is at least one image missing, then the visit will be processed
+    again (and therefore won't be listed in the list of "processed" visits).
+    """
+    from clinica.pipelines.t1_volume_tissue_segmentation.t1_volume_tissue_segmentation_pipeline import (
+        T1VolumeTissueSegmentation,
+    )
+
+    mocker.patch(
+        "clinica.utils.check_dependency._get_spm_version",
+        return_value=Version("12.7219"),
+    )
+    bids = build_bids_directory(
+        tmp_path / "bids",
+        {"sub-01": ["ses-M000", "ses-M006"], "sub-02": ["ses-M000", "ses-M012"]},
+    )
+    caps = build_caps_directory(
+        tmp_path / "caps",
+        {
+            "pipelines": {
+                "t1_volume_tissue_segmentation": {
+                    "tissue_classes": (1, 2, 3, 4),
+                    "dartel_tissues": (2, 4, 5, 6),
+                }
+            },
+            "subjects": {
+                "sub-01": ["ses-M006"],
+                "sub-02": ["ses-M000", "ses-M012"],
+            },
+        },
+    )
+    pipeline = T1VolumeTissueSegmentation(
+        bids_directory=str(bids),
+        caps_directory=str(caps),
+        parameters={
+            "tissue_classes": (1, 2, 5, 6),
+            "dartel_tissues": (2, 4, 6),
+        },
+    )
+    # No visit considered already processed since we are asking for tissues 1, 2, 5, and 6
+    # and the CAPS folder only contains tissues 1, 2, 3, and 4
+    assert pipeline.get_processed_visits() == []
+
+    pipeline = T1VolumeTissueSegmentation(
+        bids_directory=str(bids),
+        caps_directory=str(caps),
+        parameters={
+            "tissue_classes": (1, 2, 3, 4),
+            "dartel_tissues": (1, 2, 3),
+        },
+    )
+    # No visit considered already processed since we are asking for dartel tissues 1, 2, and 3
+    # and the CAPS folder only contains tissues 2, 4, 5, and 6 (1 and 3 are missing...)
+    assert pipeline.get_processed_visits() == []
+
+    pipeline = T1VolumeTissueSegmentation(
+        bids_directory=str(bids),
+        caps_directory=str(caps),
+        parameters={
+            "tissue_classes": (1, 2, 3),
+            "dartel_tissues": (2, 4, 5),
+        },
+    )
+    # All visits are considered processed since we are asking for tissues that are present in the CAPS folder
+    assert pipeline.get_processed_visits() == [
+        Visit("sub-01", "ses-M006"),
+        Visit("sub-02", "ses-M000"),
+        Visit("sub-02", "ses-M012"),
+    ]
+
+    # Delete the folder "dartel_input" altogether for subject 02 session M000 (but keep the other folders)
+    shutil.rmtree(
+        tmp_path
+        / "caps"
+        / "subjects"
+        / "sub-02"
+        / "ses-M000"
+        / "t1"
+        / "spm"
+        / "segmentation"
+        / "dartel_input"
+    )
+    # Check that subject 02 session M000 is not considered a processed visit anymore
+    assert pipeline.get_processed_visits() == [
+        Visit("sub-01", "ses-M006"),
+        Visit("sub-02", "ses-M012"),
+    ]
+
+    # Delete a single file in the "native_space" folder for subject 01 session M006 (keep other files and folders)
+    (
+        tmp_path
+        / "caps"
+        / "subjects"
+        / "sub-01"
+        / "ses-M006"
+        / "t1"
+        / "spm"
+        / "segmentation"
+        / "native_space"
+        / "sub-01_ses-M006_T1w_segm-graymatter_probability.nii.gz"
+    ).unlink()
+    # Check that subject 01 session M006 is not considered a processed visit anymore
+    assert pipeline.get_processed_visits() == [Visit("sub-02", "ses-M012")]