Table 1.

Imaging modalities, sample counts and clinical endpoints to predict in the collected cohorts of this study. For clinical and patient characteristics of each cohort, see Supplemental Table S1-S3. Original sample counts refer to the number of cases this study incorporated to its analysis. Harmonized sample counts refer to the number of samples that were mutually-present as valid across all four delineation-specific radiomic databases. Class imbalance ratio refers to the ratio of the minority subclass vs. the number of all samples in the harmonized radiomic datasets

Cohort	Imaging	Sample count (original)	Sample count (harmonized)	Class imbalance ratio (harmonized)	To predict
Glioma	11C-MET PET	105	84	39%	3-years survival
Lung	18F-FDG PET/CT	543	335	39%	2-years survival
Prostate	68GA-PSMA PET/MRI	121	75	40%	Low-vs-high risk