Training Dataset
Patient-level splitting was used to prevent leakage between training and held-out testing. Of the 7,600 mIF slices, 3,554 had co-registered H&E and clinical metadata and 4,046 were mIF-only. The paired pretraining set included 26,669,005 t…
1 sources - 4 claims
Patient-level splitting was used to prevent leakage between training and held-out testing. Of the 7,600 mIF slices, 3,554 had co-registered H&E and clinical metadata and 4,046 were mIF-only. The paired pretraining set included 26,669,005 tri-modal patches from 3,218 tissue sections. The dataset came from a multi-center, multi-disease Enable Medicine cohort with 7,600 mIF tissue slices.