We now have a pretty good idea of the metrics we need when running the ray segmentation pipeline for scientific experiments. In particular:
run same watershed on train and test volumes
train classifier (collect training error and feature importance if possible)
segment test volume (collect VOI and boundary precision-recall metrics)
We now have a pretty good idea of the metrics we need when running the ray segmentation pipeline for scientific experiments. In particular:
run same watershed on train and test volumes train classifier (collect training error and feature importance if possible) segment test volume (collect VOI and boundary precision-recall metrics)