Evaluation, Reproducibility, Benchmarks Meeting 7 - Project-MONAI/MONAI GitHub Wiki
Minutes of meeting 7
Date: 27th January 2021
Present: Carole, David, Annika, Jens, Paul, Nicola, Keyvan, Bennett, Michela
TOP 1: Task forces reports
Easy data access task force (Michela)
- Survey and email ready to be sent
- Will be sent to Lena
Metrics task force (Carole):
- Currently on hold
- Many tickets are currently taken care of, especially regarding the reporting
Benchmarking task force (Annika):
- Currently on hold
TOP 2: MICCAI challenges
- Final feedback of acceptance for MICCAI challenges: Beginning of march
- Suggestion: Challenge organizers may implement baselines in MONAI
- Help participants with dataloading, structuring and understanding the challenge
- Decision: Contact challenge organizers to inform them about this possibility and clarify the possible benefit
TOP 3: Delphi process on metrics
- Round 1
- 25 experts replied to questionnaire
- 5 organizers will take care of evaluation
- Extracted multiple scenarios and will assign categories (e.g. Object detection, classification etc.)
- Be aware of biases in the replies (different research fields etc.)
- Round 2
- Keep structure of first questionnaire (only few text fields; easy to fill out)
- Goal: Assign metric short list to categories/applications and collect pitfalls
- Rounds 3+
- Which metrics should (not) be combined?
- Metric with synonyms: Decide which wording to use
- Clinical relevance of individual metrics