MSSEG-2 individual pipelines results

This zenodo repository contains every metric results for all patients of the testing set and all pipelines evaluated during the MSSEG-2 challenge. Each excel file contains the results for a given pipeline evaluated (see https://portal.fli-iam.irisa.fr/msseg-2/challenge-day/ for a list of evaluated pipelines and teams). Each excel file is split into two parts:

- a table for the 32 patients where new lesions have appeared between time point 1 and time point 2
- a table underneath for the 28 patients where no new lesions have appeared between time point 1 and time point 2

Each table contains several columns, some common to the two tables:

- patients: patient number (same across all excel files)
- NbLesions: number of new lesions in the ground truth

Ans some not: the metrics (described on the challenge website: https://portal.fli-iam.irisa.fr/msseg-2/ and computed using animaSegPerfAnalyzer https://anima.irisa.fr):

- F1_score: F1 score at the lesion level (detection metric)
- Dice: Dice score (segmentation metric)
- Sensitivity: voxelwise sensitivity (segmentation metric)
- Specificity: voxelwise specificity (segmentation metric)
- PPV: voxelwise positive predictive value (segmentation metric)
- NPV: voxelwise negative predictive value (segmentation metric)
- RelativeVolumeError: volume relative difference between ground truth and evaluated segmentation (segmentation metric)
- HausdorffDistance: surface based distance between ground truth and evaluated segmentation (segmentation metric)
- ContourMeanDistance: surface based distance (average  point distance along contour) between ground truth and evaluated segmentation (segmentation metric)
- SurfaceDistance: surface based distance (average point distance along contour) between ground truth and evaluated segmentation (segmentation metric)
- PPVL: positive predictive value at the lesion level (detection metric)
- SensL: sensitivity value at the lesion level (detection metric)
- FN: number of false negatives at the lesion level (detection metric)
- FP: number of false positives at the lesion level (detection metric)
- TP: number of true positives at the lesion level (detection metric)

Metrics for no lesion cases are the following:

- NbTestedLesions: number of lesions found by the algorithm (detection metric)
- VolTestedLesions: volume of lesions detected by the algorithm (segmentation metric)

For any question or remark on the results provided, please contact challenges-iam@inria.fr.