refactor evaluation and reintroduce baseline docker workflow

Author: ChristianHinge

.github/workflows/baseline-docker.yml

@@ -0,0 +1,49 @@
+name: Build & Push Baseline Image
+
+on:
+  push:
+    branches: [main]
+    tags: ["v*"]
+    paths:
+      - "src/baseline/**"
+      - ".github/workflows/baseline-docker.yml"
+
+jobs:
+  build-push:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Log in to GHCR
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ghcr.io/bic-mac-challenge/baseline
+          tags: |
+            type=semver,pattern={{version}}
+            type=semver,pattern={{major}}.{{minor}}
+            type=sha,prefix=sha-,format=short
+            type=raw,value=latest,enable={{is_default_branch}}
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Build and push
+        uses: docker/build-push-action@v6
+        with:
+          context: src/baseline
+          push: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

src/baseline/dataset.py

@@ -3,6 +3,9 @@
 
 
 def get_case_features(features_dir):
+    """Return all available model inputs for a case: NAC-PET, topogram, combined and chunked
+    DIXON MRI (in/out-phase), MRI face mask, and metadata (sex, age, height, weight).
+    These are everything a participant model may use to predict the pseudo-CT."""
     paths = {
         "nacpet":                  os.path.join(features_dir, "nacpet.nii.gz"),
         "topogram":                os.path.join(features_dir, "topogram.nii.gz"),
@@ -18,6 +21,8 @@ def get_case_features(features_dir):
 
 
 def get_case_ct_labels(ct_label_dir):
+    """Return label paths for a case. The target to predict is `ct` (Hounsfield Units),
+    along with body/organ segmentations and a prediction mask."""
     return {
         "ct":               os.path.join(ct_label_dir, "ct.nii.gz"),
         "body_seg":         os.path.join(ct_label_dir, "body_seg.nii.gz"),
@@ -27,6 +32,7 @@ def get_case_ct_labels(ct_label_dir):
 
 
 def get_dataset(data_dir):
+    """Build a list of cases from a directory of subjects, each combining features and CT labels."""
     cases = []
     for sub in sorted(os.listdir(data_dir)):
         case_dir = os.path.join(data_dir, sub)

src/evaluation/eval.py src/evaluation/eval_case.pysrc/evaluation/eval.py renamed to src/evaluation/eval_case.py

@@ -22,7 +22,7 @@
 )
 
 
-def evaluate_subject(subject_path, pred_pet_path=None, pred_ct_path=None):
+def evaluate_case(subject_path, pred_pet_path=None, pred_ct_path=None):
     """
     Run metrics for a single subject.
 
@@ -103,7 +103,7 @@ def main():
     if args.pred_pet is None and args.pred_ct is None:
         parser.error("At least one of --pred_pet or --pred_ct must be provided.")
 
-    results = evaluate_subject(args.subject_path, args.pred_pet, args.pred_ct)
+    results = evaluate_case(args.subject_path, args.pred_pet, args.pred_ct)
 
     print("\n================ Evaluation Results ================")
     print(f"Subject: {os.path.basename(args.subject_path)}")

src/evaluation/eval_dataset.py

@@ -26,11 +26,11 @@
 
 import numpy as np
 
-from eval import evaluate_subject
+from eval_case import evaluate_case
 from metrics import compute_brain_outlier_score
 
 
-def evaluate(dataset_path, pred_dir, subjects=None):
+def eval_dataset(dataset_path, pred_dir, subjects=None):
     """
     Evaluate predictions across multiple subjects.
 
@@ -70,7 +70,7 @@ def evaluate(dataset_path, pred_dir, subjects=None):
         pred_pet     = os.path.join(pred_dir, subject_id, "pet.nii.gz")
         pred_ct      = os.path.join(pred_dir, subject_id, "ct.nii.gz")
 
-        results = evaluate_subject(subject_path, pred_pet, pred_ct)
+        results = evaluate_case(subject_path, pred_pet, pred_ct)
         per_subject[subject_id] = results
 
         print(f"  {subject_id}")
@@ -129,7 +129,7 @@ def main():
     )
     args = parser.parse_args()
 
-    evaluate(args.dataset_path, args.pred_dir, args.subjects)
+    eval_dataset(args.dataset_path, args.pred_dir, args.subjects)
 
 
 if __name__ == "__main__":