Update dataset card

README.md

@@ -13,7 +13,7 @@ tags:
   - segmentation
   - ct-scan
   - projection
-pretty_name: RadGenome-ChestCT Views
+pretty_name: RadGenome-Anatomy
 size_categories:
   - 10K<n<100K
 configs:
@@ -33,24 +33,21 @@ dataset_info:
     dtype: image
   - name: image_ll
     dtype: image
-  - name: mask_labels_pa
-    sequence: string
-  - name: masks_pa
-    sequence: image
-  - name: mask_labels_ll
-    sequence: string
-  - name: masks_ll
-    sequence: image
 ---
 
-# RadGenome-ChestCT Views
+# RadGenome-Anatomy
 
-A Parquet dataset of chest CT 2D projection views with per-anatomy binary segmentation masks,
-derived from the [RadGenome-ChestCT](https://huggingface.co/datasets/RadGenome/RadGenome-ChestCT) dataset
+**RadGenome-Anatomy** is a large-scale chest radiograph anatomy segmentation dataset
+constructed from the [RadGenome-ChestCT](https://huggingface.co/datasets/RadGenome/RadGenome-ChestCT) corpus
 (originally based on [CT-RATE](https://huggingface.co/datasets/ibrahimhamamci/CT-RATE)).
+It contains **25,692 volumetric studies** (24,128 train / 1,564 validation), yielding paired
+postero-anterior (PA) and lateral (LL) projection images at **384 × 384** resolution.
 
-Each row represents one CT volume and contains its PA (posteroanterior) and LL (lateral)
-projection images, plus binary anatomy masks for all 210 anatomical structures.
+Across the two radiographic views, the dataset provides **10,790,646 fine-grained anatomy masks**
+over **210 canonical anatomy classes** and **513,860 region masks** over **10 anatomical groups**,
+for a total of **11,304,506 binary mask instances**.
+
+Each row represents one CT study and contains its PA and LL projection images.
 
 ---
 
@@ -58,18 +55,23 @@ projection images, plus binary anatomy masks for all 210 anatomical structures.
 
 | Property | Value |
 |---|---|
-| **Volumes** | 25,692 total |
-| **Views per volume** | 2 (PA + LL) |
-| **Anatomy classes** | 210 structures |
+| **Studies** | 25,692 total (24,128 train / 1,564 val) |
+| **Views per study** | 2 (PA + LL) |
+| **Image resolution** | 384 × 384 |
+| **Anatomy classes** | 210 structures (4-level hierarchy) |
+| **Region classes** | 10 body-system groups |
+| **Anatomy masks** | 10,790,646 (5,395,323 PA + 5,395,323 LL) |
+| **Region masks** | 513,860 (256,930 PA + 256,930 LL) |
 | **License** | CC-BY-4.0 |
 | **Source** | RadGenome-ChestCT / CT-RATE |
 
 ### Splits
 
-| Split | Volumes |
-|---|---|
-| train | 24,128 |
-| validation | 1,564 |
+| Split | Studies | PA projections | LL projections | Anatomy masks | Region masks |
+|---|---|---|---|---|---|
+| train | 24,128 | 24,129 | 24,129 | 10,133,770 | 482,580 |
+| validation | 1,564 | 1,564 | 1,564 | 656,876 | 31,280 |
+| **total** | **25,692** | **25,693** | **25,693** | **10,790,646** | **513,860** |
 
 ---
 
@@ -79,26 +81,33 @@ projection images, plus binary anatomy masks for all 210 anatomical structures.
 
 | Column | Type | Description |
 |---|---|---|
-| `volume_id` | `str` | Unique volume identifier, e.g. `train_1_a_1`. |
+| `volume_id` | `str` | Unique study identifier, e.g. `train_1_a_1`. |
 | `split` | `str` | Dataset split: `train` or `validation`. |
-| `image_pa` | `Image` | PA (posteroanterior, front) chest projection image (JPEG). |
-| `image_ll` | `Image` | LL (lateral, side) chest projection image (JPEG). |
-| `mask_labels_pa` | `List[str]` | Anatomy structure names for the PA masks, e.g. `["lung", "heart", ...]`. Parallel to `masks_pa`. |
-| `masks_pa` | `List[Image]` | Binary anatomy segmentation masks (one per structure) for the PA view. White pixels = structure present. |
-| `mask_labels_ll` | `List[str]` | Anatomy structure names for the LL masks. Parallel to `masks_ll`. |
-| `masks_ll` | `List[Image]` | Binary anatomy segmentation masks for the LL view. |
+| `image_pa` | `Image` | PA (posteroanterior, front) chest projection image (JPEG, 384×384). |
+| `image_ll` | `Image` | LL (lateral, side) chest projection image (JPEG, 384×384). |
 
 ### Anatomy Label Universe
 
-The dataset contains masks for **210 anatomical structures**, covering:
-- **Lungs**: left/right lung, upper/lower/middle lobes, lung nodule, tumor, effusion
-- **Heart**: chambers (atria, ventricles), myocardium, ascending aorta, pulmonary artery/vein
-- **Vessels**: aorta, inferior/superior vena cava, carotid arteries, subclavian arteries, brachiocephalic vessels, iliac vessels, renal vessels
-- **Skeleton**: ribs (1–12, left/right), thoracic vertebrae (T1–T12), cervical/lumbar vertebrae, sternum, clavicles, scapulae, humerus, femur
-- **Abdominal organs**: liver (with segments), spleen, pancreas, kidneys, gallbladder, stomach, intestine, esophagus
-- **Endocrine/other**: thyroid, adrenal glands, thymus, trachea, larynx, spinal cord
+The dataset defines **210 canonical anatomy classes** organized as a four-level hierarchy:
+*body system → organ → substructure → canonical label*. At the top level, classes are grouped
+into **10 body systems**, with a highly non-uniform per-system class count:
 
-The full ordered list is in [`label_universe.json`](./label_universe.json) at the repo root.
+| Body system | # classes | Example structures |
+|---|---|---|
+| Skeletal | 93 | ribs (1–12 L/R), thoracic vertebrae (T1–T12), cervical/lumbar vertebrae, sternum, clavicles, scapulae, humerus, femur |
+| Abdominal | 42 | liver (with segments), spleen, pancreas, kidneys, gallbladder, stomach, intestine |
+| Mediastinal | 25 | aorta, IVC/SVC, carotid/subclavian arteries, brachiocephalic vessels, iliac/renal vessels |
+| Cardiac | — | atria, ventricles, myocardium, ascending aorta, pulmonary artery/vein |
+| Pulmonary | — | left/right lung, upper/middle/lower lobes, lung nodule, tumor, effusion |
+| Airway | — | trachea, main bronchi, larynx |
+| Endocrine | — | thyroid, adrenal glands, thymus |
+| Esophageal | 2 | esophagus structures |
+| Breast | 3 | breast structures |
+| Neural / soft tissue | 5 | spinal cord, skin, muscle |
+
+These same 10 body systems also serve as the **region-mask** label set (10 classes/view).
+
+The full ordered list of canonical labels is in [`label_universe.json`](./label_universe.json) at the repo root.
 Use it to map labels to fixed class indices for consistent multi-label training.
 
 ---
@@ -112,10 +121,6 @@ from datasets import load_dataset
 
 ds = load_dataset("EvidenceAIResearch/radgenome-anatomy")
 print(ds)
-# DatasetDict({
-#     train: Dataset({features: ['volume_id', 'split', 'image_pa', 'image_ll', ...], num_rows: 24,128})
-#     validation: Dataset({features: ['volume_id', 'split', 'image_pa', 'image_ll', ...], num_rows: 1,564})
-# })
 ```
 
 ### Access images
@@ -125,57 +130,10 @@ from PIL import Image
 import io
 
 row = ds["train"][0]
-print(row["volume_id"])   # e.g. "train_1_a_1"
-
-# Decode PA image
 pa_img = Image.open(io.BytesIO(row["image_pa"]["bytes"]))
 ll_img = Image.open(io.BytesIO(row["image_ll"]["bytes"]))
-
-pa_img.show()
 ```
 
-### Working with masks
-
-Each sample contains a list of binary masks with corresponding label names:
-
-```python
-from PIL import Image
-import io, numpy as np
-
-row = ds["train"][0]
-
-# Map label → mask for the LL view
-mask_map = {
-    label: Image.open(io.BytesIO(img["bytes"]))
-    for label, img in zip(row["mask_labels_ll"], row["masks_ll"])
-}
-
-# Example: get the lung mask as a numpy array
-lung_mask = np.array(mask_map["lung"])  # shape: (H, W), values 0 or 255
-
-# Build a multi-label tensor [num_classes, H, W] using the label universe
-import json, urllib.request
-# label_universe.json is in the repo root
-universe = json.loads(open("label_universe.json").read())
-label_to_idx = {l: i for i, l in enumerate(universe)}
-
-H, W = lung_mask.shape
-multi_label = np.zeros((len(universe), H, W), dtype=np.uint8)
-for label, img in zip(row["mask_labels_ll"], row["masks_ll"]):
-    if label in label_to_idx:
-        arr = np.array(Image.open(io.BytesIO(img["bytes"])))
-        multi_label[label_to_idx[label]] = (arr > 128).astype(np.uint8)
-```
-
-### Streaming (large dataset)
-
-```python
-ds = load_dataset("EvidenceAIResearch/radgenome-anatomy", streaming=True)
-for row in ds["train"]:
-    pa_img = Image.open(io.BytesIO(row["image_pa"]["bytes"]))
-    # process ...
-    break
-```
 
 ---