(soft) conclusion:
EXTRACT is likely to require human curation. Only in MATLAB for now. Not great at the first try (different dataset).
CITE-On is specific for closed-loop 2p optogenetics exp, very crude (bounding box + 80-95th percentile pixel selection)
STNeuroNet seems good (CNN-based), but only detects "active" neurons. May need to compare with the in-house CNN-based protocol.
Standard methods: Suite2p, CAIMAN, OnACID Recently developed methods: EXTRACT, CITE-On, STNeuroNet
Review slide deck: https://alleninstitute-my.sharepoint.com/:p:/g/personal/jinho_kim_alleninstitute_org/Ebk95v4_NXpInKLYYfwyuaoBfLcpu-PqNh7xVfWFbna5UQ?e=sr9hZm
(soft) conclusion: EXTRACT is likely to require human curation. Only in MATLAB for now. Not great at the first try (different dataset). CITE-On is specific for closed-loop 2p optogenetics exp, very crude (bounding box + 80-95th percentile pixel selection) STNeuroNet seems good (CNN-based), but only detects "active" neurons. May need to compare with the in-house CNN-based protocol.