Merge pull request #28 from ChristianHinge/fix-pet-liver-superior-slice

Fix superior liver slice detection in PET SUV MAE metric

Author: ChristianHinge

README.md

@@ -8,6 +8,7 @@
 
 ---
 ## Updates
+- **July 16, 2026: [BUG FIX]**: The Whole-body SUV MAE metric excluded the wrong end of the liver: PET's z-axis runs in the opposite direction from CT's, so the superior-most liver slice was located incorrectly, shifting the ±4 cm exclusion band off the top of the liver. Please `git pull` for the corrected evaluation logic. Please see [PR #28](https://github.com/bic-mac-challenge/challenge-codebase/pull/28) for details.
 - **July 8, 2026: [CHANGE]**: The fifth metric, TAC-Bias, has been removed from the final evaluation. Final submissions are scored on the four metrics defined in [Evaluation](#evaluation-srcevaluation). Please see [PR #23](https://github.com/bic-mac-challenge/challenge-codebase/pull/23) for details.
 - **July 6, 2026: [BUG FIX]**: DRY-RUN and FINAL submission instructions listed a wrong email. All DRY-RUN and FINAL submissions should be sent to bic-mac-challenge@outlook.com. 
 - **May 15, 2026: [ANNOUNCEMENT]**: NIfTI validation and leaderboards now live - see submission instructions at [https://www.codabench.org/competitions/12555](https://www.codabench.org/competitions/12555)

src/evaluation/metrics/pet_whole_body_mae.py

@@ -48,7 +48,10 @@ def compute_whole_body_suv_mae(
     slice_thickness_mm = nib.load(pred_pet_path).header.get_zooms()[2]
     exclusion_slices = int(round((exclusion_cm * 10.0) / slice_thickness_mm))
 
-    superior_slice = np.max(np.where(liver_mask)[2])
+    # PET's z-axis (STIR-resampled, ring spacing 3.29114 mm) is stored in the
+    # opposite direction from the CT grid: index increases toward inferior,
+    # so the superior-most liver voxel is the *smallest* index here.
+    superior_slice = np.min(np.where(liver_mask)[2])
 
     z_min = max(0, superior_slice - exclusion_slices)
     z_max = min(pred.shape[2], superior_slice + exclusion_slices)