[Radiology] Add ct scan level classification datasets

.gitignore

@@ -179,3 +179,4 @@ cython_debug/
 
 # numpy data
 *.npy
+!tests/eva/assets/vision/datasets/cc_ccii/CC-CCII_public/data/*

.python-version

@@ -1 +1 @@
-3.10.14
+3.12.8

pyproject.toml

@@ -73,6 +73,7 @@ vision = [
     "scipy>=1.14.0",
     "monai>=1.3.2",
     "einops>=0.8.1",
+    "medmnist>=3.0.2",
 ]
 language = [
     "datasets<4.0.0,>=2.19.0",
@@ -102,6 +103,7 @@ all = [
     "litellm>=1.61.8",
     "backoff>=2.2.1",
     "json-repair>=0.52.0",
+    "medmnist>=3.0.2",
 ]
 
 [project.scripts]

src/eva/core/data/samplers/classification/balanced.py

@@ -1,4 +1,4 @@
-""""Balanced class sampler for data loading."""
+"""Balanced class sampler for data loading."""
 
 from loguru import logger
 from typing_extensions import override

src/eva/core/data/splitting/stratified.py

@@ -1,16 +1,17 @@
 """Functions for stratified splitting."""
 
-from typing import Any, List, Sequence, Tuple
+from typing import Any, Iterable, List, Sequence, Tuple
 
 import numpy as np
 
 
 def stratified_split(
-    samples: Sequence[Any],
-    targets: Sequence[Any],
+    samples: Sequence[Any] | Iterable[Any],
+    targets: Sequence[Any] | Iterable[Any],
     train_ratio: float,
     val_ratio: float,
     test_ratio: float = 0.0,
+    groups: Sequence[Any] | Iterable[Any] | None = None,
     seed: int = 42,
 ) -> Tuple[List[int], List[int], List[int] | None]:
     """Splits the samples into stratified train, validation, and test (optional) sets.
@@ -21,19 +22,51 @@ def stratified_split(
         train_ratio: The ratio of the training set.
         val_ratio: The ratio of the validation set.
         test_ratio: The ratio of the test set (optional).
+        groups: Optional group labels for group-wise stratification.
         seed: The seed for reproducibility.
 
     Returns:
         The indices of the train, validation, and test sets.
     """
-    if len(samples) != len(targets):
+    samples_seq = samples if isinstance(samples, (list, tuple)) else list(samples)
+    targets_seq = targets if isinstance(targets, (list, tuple)) else list(targets)
+
+    if train_ratio + val_ratio + test_ratio > 1.0:
+        raise ValueError("The sum of the ratios must be lower or equal to 1")
+
+    if len(samples_seq) != len(targets_seq):
         raise ValueError("The number of samples and targets must be equal.")
-    if train_ratio + val_ratio + (test_ratio or 0) > 1.0:
-        raise ValueError("The sum of the ratios must be lower or equal to 1.")
+
+    if groups is not None:
+        groups_seq = groups if isinstance(groups, (list, tuple)) else list(groups)
+        if len(groups_seq) != len(samples_seq):
+            raise ValueError("The number of samples and groups must be equal.")
+
+        unique_groups, group_indices = np.unique(groups_seq, return_inverse=True)
+        group_targets = np.array(
+            [targets_seq[np.where(group_indices == i)[0][0]] for i in range(len(unique_groups))]
+        )
+
+        train_g, val_g, test_g = stratified_split(
+            samples=unique_groups.tolist(),
+            targets=group_targets.tolist(),
+            train_ratio=train_ratio,
+            val_ratio=val_ratio,
+            test_ratio=test_ratio,
+            seed=seed,
+        )
+
+        def map_indices(g_list):
+            if g_list is None:
+                return []
+            selected_groups = unique_groups[g_list]
+            return np.where(np.isin(groups_seq, selected_groups))[0].tolist()
+
+        return map_indices(train_g), map_indices(val_g), map_indices(test_g) or None
 
     use_all_samples = train_ratio + val_ratio + test_ratio == 1
     random_generator = np.random.default_rng(seed)
-    unique_classes, y_indices = np.unique(targets, return_inverse=True)
+    unique_classes, y_indices = np.unique(targets_seq, return_inverse=True)
     n_classes = unique_classes.shape[0]
 
     train_indices, val_indices, test_indices = [], [], []

src/eva/vision/data/datasets/__init__.py

@@ -3,12 +3,16 @@
 from eva.vision.data.datasets.classification import (
     BACH,
     BRACS,
+    CC_CCII,
     CRC,
+    LUNA25,
     MHIST,
     PANDA,
     BreaKHis,
     Camelyon16,
     GleasonArvaniti,
+    NoduleMNIST3D,
+    OrganMNIST3D,
     PANDASmall,
     PatchCamelyon,
     UniToPatho,
@@ -36,11 +40,15 @@
     "BRACS",
     "BTCV",
     "Camelyon16",
+    "CC_CCII",
     "CoNSeP",
     "CRC",
+    "LUNA25",
     "EmbeddingsSegmentationDataset",
     "FLARE22",
     "GleasonArvaniti",
+    "NoduleMNIST3D",
+    "OrganMNIST3D",
     "KiTS23",
     "LiTS17",
     "MHIST",

src/eva/vision/data/datasets/classification/__init__.py

@@ -4,9 +4,13 @@
 from eva.vision.data.datasets.classification.bracs import BRACS
 from eva.vision.data.datasets.classification.breakhis import BreaKHis
 from eva.vision.data.datasets.classification.camelyon16 import Camelyon16
+from eva.vision.data.datasets.classification.cc_ccii import CC_CCII
 from eva.vision.data.datasets.classification.crc import CRC
 from eva.vision.data.datasets.classification.gleason_arvaniti import GleasonArvaniti
+from eva.vision.data.datasets.classification.luna25 import LUNA25
 from eva.vision.data.datasets.classification.mhist import MHIST
+from eva.vision.data.datasets.classification.nodule_mnist_3d import NoduleMNIST3D
+from eva.vision.data.datasets.classification.organ_mnist_3d import OrganMNIST3D
 from eva.vision.data.datasets.classification.panda import PANDA, PANDASmall
 from eva.vision.data.datasets.classification.patch_camelyon import PatchCamelyon
 from eva.vision.data.datasets.classification.unitopatho import UniToPatho
@@ -17,9 +21,13 @@
     "BreaKHis",
     "BRACS",
     "Camelyon16",
+    "CC_CCII",
     "CRC",
     "GleasonArvaniti",
+    "LUNA25",
     "MHIST",
+    "NoduleMNIST3D",
+    "OrganMNIST3D",
     "PatchCamelyon",
     "UniToPatho",
     "WsiClassificationDataset",