External validation of the DREAM Digital Mammography Competition Ensemble Model
- Download Docker images of the Ensemble Models
docker login docker.synapse.org docker pull docker.synapse.org/syn7887972/9646823/scoring docker pull docker.synapse.org/syn7887972/9648211/scoring docker pull docker.synapse.org/syn7887972/9648881/scoring docker pull docker.synapse.org/syn7887972/9648888/scoring docker pull docker.synapse.org/syn7887972/9648889/scoring docker pull docker.synapse.org/syn7887972/9648917/scoring docker pull docker.synapse.org/syn7887972/9650054/scoring docker pull docker.synapse.org/syn7887972/9650055/scoring docker pull docker.synapse.org/syn7887972/9650232/scoring docker pull docker.synapse.org/syn7887972/9650290/scoring docker pull docker.synapse.org/syn7887972/9650300/scoring
If Docker images cannot be downloaded, one needs to contact DREAM Challenge team for download permission.
-
Prepare metadata and imaging files by running the Python program, sc2_data_prep.py. Output: ./images: DICOM image files ./metadata/exams_metadata.tsv ./metadata/images_crosswalk.tsv
When facing a large number of exams, the software can divide exams into sub-collections. User can specify a max number of subjects to be in a collection through a command-line argument.
-
Launch SC2 DREAM challenge ensemble models place both 'metadata' and 'images' in a same parent directory. In the parent directory, copy and run the following shell program.
run_sc2_models.sh When running the models for sub-collections, copy 'run_sc2_models.sh' over and run it in each sub-collection.