Fig. 2

A schematic depicting the four cohorts, the number of cases used for fine-tuning of the algorithm and the distribution of cases between the three different scanner models. Two cases from medical centers A-C have been used for fine-tuning and therefore excluded from the validation cohort. All of the cases from medical centers A-D have been scanned by the Philips brand scanner. In addition, limited subsets (from the same cohort) of cases from medical center D have also been scanned by the Hamamatsu and 3DHISTECH brand scanners