synT1CE - Identity Baseline (T1n → T1CE)
This repository contains the results of the Identity baseline for the synthesis of T1CE MRI from T1n (BraTS-MEN dataset).
Tier 1 Metrics (Global)
- SSIM (All): 0.7672
- PSNR (All): 17.97
- MAE (All): 0.5525
Tier 2 Metrics (Expert)
- EVarΔ: 0.0000
- Wasserstein d_W: 0.5688
- Multimodality L1: None
Repository Content
metrics_summary.json: Structured performance summary.splits.csv: List of patients used for training, validation, and testing.visualizations/: Folder containing the final montage and axial slice examples (T1n, GT, Synthesis, Error Map).baseline0_identity.zip: Full archive of the results.
Interpretation
In this baseline, the T1n input is used directly as the prediction for the T1CE. An EVarΔ of 0.0 confirms the total failure to capture the dynamic contrast enhancement of the tumor, serving as a lower bound for subsequent models.
