PDCADxFoundation Challenge

Segmentation Scoring Methodology

1. Evaluation Metrics

For each brain region in both hemispheres (left and right), we calculate:

Dice Coefficient (Dice): Measures overlap between prediction and ground truth (weight: 0.7)
Hausdorff Distance (Hausdorff): Measures boundary agreement (weight: 0.3)

2. Weighting Scheme

Category	Weights	Description
Modality Weights	QSM: 0.6 NM: 0.4	Relative importance of each imaging modality
QSM Region Weights	CN: 1.0 Put: 1.0 GP: 1.0 STN: 1.2 SN: 1.5 RN: 1.1 DN: 1.0	Relative importance of each brain region in QSM
NM Region Weights	SN: 2.0	Relative importance of brain region in NM
Metric Weights	Dice: 0.7 Hausdorff: 0.3	Relative importance of each evaluation metric

3. Scoring Formula

The composite score for a team is calculated as:

CompositeScore = 0.6 × QSM_Score + 0.4 × NM_Score

QSM_Score = 1.0×CN + 1.0×Put + 1.0×GP + 1.2×STN + 1.5×SN + 1.1×RN + 1.0×DN
NM_Score = 2.0×SN

Where each region score (e.g., CN) is calculated as:
RegionScore = 0.7×Dice + 0.3×(1 - NormalizedHausdorff)
NormalizedHausdorff = Hausdorff / MaxPossibleHausdorff
(MaxPossibleHausdorff = 10)

4. Ranking Criteria

Teams are ranked by their composite score in descending order. The highest possible score is 9.8 (perfect segmentation in all regions).

Brain Region Segmentation Results (with Dice scores)

* QSM: Quantitative Susceptibility Mapping | NM: Neuromelanin-sensitive MRI

PDCADxFoundation Challenge Leaderboard