Evaluate the annotations produced by the seizure detection algorithm invoked in step 3.
A package named evaluate is provided to do the evaluation. It relies on the timescoring library.
The package provides a simple interface and prints results to the command line:
python -m evaluate rootRefDataset rootHypDatasetThe output provides both sample based and event based scoring. As explained in the SzCORE paper.
# Sample scoring
- Sensitivity : 0.62
- Precision : 0.05
- F1-score : 0.09
- FP/24h : 52487.08
# Event scoring
- Sensitivity : 1.00
- Precision : 0.07
- F1-score : 0.13
- FP/24h : 309.64