Cleaning part 2 + DataGroup (#690)

clinicadl/data/datasets/caps_dataset.py

@@ -24,6 +24,7 @@
 from clinicadl.utils.exceptions import ClinicaDLCAPSError, ClinicaDLTSVError
 from clinicadl.utils.iotools.clinica_utils import create_subs_sess_list
 from clinicadl.utils.loading import nifti_to_tensor, pt_to_tensor
+from clinicadl.utils.typing import DataType, PathType
 
 logger = getLogger("clinicadl.caps_dataset")
 
@@ -68,10 +69,10 @@ class CapsDataset(Dataset):
 
     def __init__(
         self,
-        caps_directory: Union[str, Path],
+        caps_directory: PathType,
         preprocessing: Preprocessing = PreprocessingT1(),
         transforms: Transforms = Transforms(),
-        data: Optional[Union[pd.DataFrame, str, Path]] = None,
+        data: Optional[DataType] = None,
         label: Optional[str] = None,
         masks: Optional[list[str]] = None,
     ):
@@ -207,9 +208,7 @@ def describe(self):
             "extraction": self.extraction.model_dump(),
         }
 
-    def _get_df_from_input(
-        self, data: Optional[Union[pd.DataFrame, Path, str]]
-    ) -> pd.DataFrame:
+    def _get_df_from_input(self, data: Optional[DataType]) -> pd.DataFrame:
         """
         Generates or validates the DataFrame from the input data.
 
@@ -248,7 +247,7 @@ def _get_df_from_input(
 
         return df
 
-    def _check_data_instance(self, data: Optional[Union[pd.DataFrame, Path, str]]):
+    def _check_data_instance(self, data: Optional[DataType]):
         if isinstance(data, str):
             data = Path(data)
 
@@ -550,7 +549,7 @@ def train(self):
         """
         self.eval_mode = False
 
-    def subset(self, data: Optional[Union[pd.DataFrame, Path]] = None) -> CapsDataset:
+    def subset(self, data: Optional[DataType] = None) -> CapsDataset:
         df = self._check_data_instance(data)
 
         common_rows = pd.merge(df, self.df, how="inner")

clinicadl/data/utils.py

@@ -19,9 +19,7 @@
 SESSION_ID = "session_id"
 
 
-def df_to_tsv(
-    name: str, results_path: Path, df: pd.DataFrame, baseline: bool = False
-) -> None:
+def df_to_tsv(tsv_path: Path, df: pd.DataFrame, baseline: bool = False) -> None:
     """
     Write Dataframe into a TSV file and drop duplicates
 
@@ -46,7 +44,7 @@ def df_to_tsv(
             subset=["participant_id", "session_id"], keep="first", inplace=True
         )
     # df = df[["participant_id", "session_id"]]
-    df.to_csv(results_path / name, sep="\t", index=False)
+    df.to_csv(tsv_path, sep="\t", index=False)
 
 
 def tsv_to_df(tsv_path: Path) -> pd.DataFrame:

clinicadl/dictionary/suffixes.py

@@ -2,6 +2,8 @@
 JSON = ".json"
 LOG = ".log"
 PT = ".pt"
+PTH = ".pth"
+TAR = ".tar"
 NII = ".nii"
 GZ = ".gz"
 NII_GZ = NII + GZ

clinicadl/dictionary/words.py

@@ -5,12 +5,15 @@
 CONFIG = "config"
 COUNT = "count"
 CUDA = "cuda"
+DATA = "data"
 DESCRIPTION = "description"
 FOLD = "fold"
+GROUPS = "groups"
 ID = "id"
 IMAGE = "image"
 KFOLD = "k" + FOLD
 LABEL = "label"
+MAPS = "maps"
 MEAN = "mean"
 OBJECT = "object"
 PARTICIPANT = "participant"

clinicadl/experiment_manager/data_group.py

@@ -0,0 +1,69 @@
+import json
+from pathlib import Path
+from typing import List, Optional
+
+import pandas as pd
+
+from clinicadl.data.datasets import CapsDataset
+from clinicadl.data.utils import df_to_tsv, tsv_to_df
+from clinicadl.dictionary.suffixes import JSON, TSV
+from clinicadl.dictionary.words import DATA, GROUPS, MAPS, SPLIT, TRAIN, VALIDATION
+from clinicadl.utils.config import ClinicaDLConfig
+
+TRAIN_VAL = [TRAIN, VALIDATION]
+
+
+class DataGroup(ClinicaDLConfig):
+    maps_path: Path
+    name: str
+    split: Optional[int]
+
+    @property
+    def data_tsv(self) -> Path:
+        return self.group_split_dir / (DATA + TSV)
+
+    @property
+    def maps_json(self) -> Path:
+        return self.group_split_dir / (MAPS + JSON)
+
+    @property
+    def group_dir(self) -> Path:
+        return self.maps_path / GROUPS / self.name
+
+    @property
+    def group_split_dir(self) -> Path:
+        if self.name in TRAIN_VAL:
+            return self.group_dir / (SPLIT + "-" + str(self.split))
+        else:
+            return self.group_dir
+
+    @property
+    def df(self) -> pd.DataFrame:
+        return tsv_to_df(self.data_tsv)
+
+    @property
+    def caps_dir(self) -> Path:
+        dict_ = self._read_json()
+        return Path(dict_["cap_directory"])
+
+    def exists(self) -> bool:
+        return self.data_tsv.is_file() and self.maps_json.is_file()
+
+    def create(self, caps_dataset: CapsDataset):
+        df_to_tsv(self.data_tsv, caps_dataset.df)
+        self._write_json(caps_dataset)
+
+    def _write_json(self, caps_dataset: CapsDataset):
+        dict_ = {}  # TODO: to complete
+        with self.maps_json.open(mode="w") as file:
+            json.dump(dict_, file)
+
+    def _read_json(self):
+        if self.maps_json.is_file():
+            with self.maps_json.open(mode="r") as file:
+                dict_ = json.load(file)
+                return dict_
+
+        raise FileNotFoundError(
+            f"Could not find the `maps.json` file for the data grou : {self.name} (in {self.maps_path})"
+        )

clinicadl/experiment_manager/maps_reader.py

@@ -0,0 +1,83 @@
+from pathlib import Path
+from typing import Optional, Union
+
+import pandas as pd
+import torch
+
+from clinicadl.data.datasets import CapsDataset
+from clinicadl.data.utils import tsv_to_df
+from clinicadl.dictionary.suffixes import PTH, TAR
+from clinicadl.dictionary.words import BEST, GROUPS, PARTICIPANT_ID, SPLIT, TMP
+from clinicadl.experiment_manager.data_group import DataGroup
+from clinicadl.model import ClinicaDLModel
+from clinicadl.splitter.split import Split
+from clinicadl.utils.exceptions import (
+    ClinicaDLConfigurationError,
+    ClinicaDLDataLeakageError,
+)
+from clinicadl.utils.typing import PathType
+
+
+class MapsReader:
+    def __init__(self, maps_path: PathType) -> None:
+        self.maps_path = Path(maps_path)
+
+    def _create_data_group(
+        self, name: str, caps_dataset: CapsDataset, split: Optional[int] = None
+    ) -> DataGroup:
+        """
+        Check that a data_group is not already written and writes the characteristics of the data group
+        (TSV file with a list of participant / session + JSON file containing the CAPS and the preprocessing).
+        """
+        data_group = DataGroup(name=name, split=split, maps_path=self.maps_path)
+        if data_group.exists():
+            raise ClinicaDLConfigurationError(
+                f"Data group {data_group.name} already exists, please give another name to your data group"
+            )
+
+        data_group.create(caps_dataset)
+        return data_group
+
+    def _load_data_group(self, name: str, split: Optional[int] = None) -> DataGroup:
+        """creates a new data_group."""
+        data_group = DataGroup(name=name, split=split, maps_path=self.maps_path)
+        if data_group.exists():
+            return data_group
+
+        raise ClinicaDLConfigurationError(
+            f"Could not find data group {data_group.name}"
+        )
+
+    def get_train_val_df(self):
+        """Loads the train and validation data groups."""
+        path = self.maps_path / GROUPS / "train+validation.tsv"
+        return tsv_to_df(path)
+
+    def get_model(self) -> ClinicaDLModel:
+        return ClinicaDLModel()  # type: ignore
+
+    def _write_network_weights(self):
+        """TO COMPLETE"""
+        pass
+
+    def _write_optim_weights(self):
+        """TO COMPLETE"""
+        pass
+
+    def write_tensor(self):
+        """TO COMPLETE"""
+        pass
+
+    def split_path(self, split: int):
+        return self.maps_path / (SPLIT + "-" + str(split))
+
+    def optimizer_path(self, split: int, resume: bool = False) -> Path:
+        """TO COMPLETE"""
+
+        return self.split_path(split) / TMP / ("optimizer" + PTH + TAR)
+
+    def checkpoint_path(self, split: int, resume: bool = False):
+        return self.split_path(split) / TMP / ("checkpoint" + PTH + TAR)
+
+    def model_path(self, split: int, metric: str):
+        return self.split_path(split) / (BEST + "-" + metric) / ("model" + PTH + TAR)

clinicadl/model/__init__.py

@@ -0,0 +1 @@
+from .clinicadl_model import ClinicaDLModel

clinicadl/predictor/config.py → clinicadl/predictor/old_predictor/config.py

@@ -41,7 +41,6 @@ class PredictConfig(BaseModel):
     data: DataConfig
     validation: ValidationConfig
     computational: ComputationalConfig
-    dataloader: DataLoaderConfig
     split: SplitConfig
     transforms: TransformsConfig
 

clinicadl/predictor/old_predictor.py → clinicadl/predictor/old_predictor/old_predictor.py

@@ -1059,7 +1059,7 @@ def _compute_output_tensors(
         Compute the output tensors and saves them in the MAPS.
 
         Args:
-            dataset (clinicadl.data.datasets.caps_dataset.CapsDataset): wrapper of the data set.
+            dataset (clinicadl.data.datasets.CapsDataset): wrapper of the data set.
             data_group (str): name of the data group used for the task.
             split (int): split number.
             selection_metrics (list[str]): metrics used for model selection.

clinicadl/predictor/predictor.py

@@ -1,12 +1,71 @@
-from clinicadl.data.caps_dataset import CapsDataset
-from clinicadl.experiment_manager.experiment_manager import ExperimentManager
+from pathlib import Path
+from typing import Optional
+
+import pandas as pd
+import torch
+
+from clinicadl.data.datasets import CapsDataset
+from clinicadl.data.utils import tsv_to_df
+from clinicadl.dictionary.words import GROUPS, PARTICIPANT_ID
+from clinicadl.experiment_manager import ExperimentManager
+from clinicadl.experiment_manager.maps_reader import MapsReader
+from clinicadl.model import ClinicaDLModel
+from clinicadl.splitter.split import Split
+from clinicadl.utils.exceptions import (
+    ClinicaDLConfigurationError,
+    ClinicaDLDataLeakageError,
+)
 
 
 class Predictor:
-    def __init__(self, manager: ExperimentManager):
+    def __init__(self, reader: MapsReader, model: ClinicaDLModel):
         """TO COMPLETE"""
-        pass
+        self.reader = reader
+        self.model = model
 
-    def predict(self, dataset_test: CapsDataset, split: int):
+    def predict(self, dataset_test: CapsDataset, split: Split):
         """TO COMPLETE"""
         pass
+
+    def _check_leakage(self, dataset_test: CapsDataset):
+        """Checks that no intersection exist between the participants used for training and those used for testing."""
+
+        df_train_val = self.reader.get_train_val_df()
+        df_test = dataset_test.df
+
+        participants_train = set(df_train_val[PARTICIPANT_ID].values)
+        participants_test = set(df_test[PARTICIPANT_ID].values)
+        intersection = participants_test & participants_train
+
+        if len(intersection) > 0:
+            raise ClinicaDLDataLeakageError(
+                "Your evaluation set contains participants who were already seen during "
+                "the training step. The list of common participants is the following: "
+                f"{intersection}."
+            )
+
+    def test(self):
+        """Computes the predictions and evaluation metrics."""
+        pass
+
+    def _test_loader(self):
+        """Launches the testing task on a dataset wrapped by a DataLoader and writes prediction TSV files."""
+        pass
+
+    def _compute_latent_tensor(self):
+        """Compute the output tensors and saves them in the MAPS."""
+        pass
+
+    @torch.no_grad()
+    def _compute_output_nifti(self):
+        """omputes the output nifti images and saves them in the MAPS."""
+        pass
+
+    @torch.no_grad()
+    def _compute_output_tensors(self):
+        """Compute the output tensors and saves them in the MAPS."""
+        pass
+
+    def _ensemble_prediction(self):
+        """Computes the results on the image-level."""
+        pass

clinicadl/splitter/make_splits/kfold.py

@@ -11,6 +11,7 @@
 from clinicadl.splitter.splitter.kfold import KFoldConfig
 from clinicadl.tsvtools.tsvtools_utils import extract_baseline
 from clinicadl.utils.exceptions import ClinicaDLConfigurationError, ClinicaDLTSVError
+from clinicadl.utils.typing import DataType, PathType
 
 
 def _validate_stratification(
@@ -98,8 +99,8 @@ def preprocess_stratification(
 
 
 def make_kfold(
-    data: Union[pd.DataFrame, Path, str],
-    output_dir: Optional[Union[Path, str]] = None,
+    data: DataType,
+    output_dir: Optional[PathType] = None,
     subset_name: str = "validation",
     valid_longitudinal: bool = False,
     n_splits: PositiveInt = 5,

clinicadl/splitter/make_splits/single_split.py

@@ -13,6 +13,7 @@
 from clinicadl.splitter.splitter.single_split import SingleSplitConfig
 from clinicadl.tsvtools.tsvtools_utils import extract_baseline
 from clinicadl.utils.exceptions import ClinicaDLConfigurationError, ClinicaDLTSVError
+from clinicadl.utils.typing import DataType, PathType
 
 logger = getLogger("clinicadl.splitter.single_split")
 
@@ -128,8 +129,8 @@ def _chi2_test(x_test: List[int], x_train: List[int]) -> float:
 
 
 def make_split(
-    data: Union[pd.DataFrame, Path, str],
-    output_dir: Optional[Union[Path, str]] = None,
+    data: DataType,
+    output_dir: Optional[PathType] = None,
     n_test: PositiveFloat = 100,
     subset_name: str = "test",
     p_categorical_threshold: float = 0.50,