Fig. 2
From: Detecting the left atrial appendage in CT localizers using deep learning
Patient flowchart with inclusion and exclusion criteria. Three independent cohorts were used. From Hospital A, we used a chronological split to obtain a training cohort and an internal test cohort. It was ensured that there was no overlap between these two cohorts to avoid any bias. In addition, the test sets contained only one scan per patient. In total, 1253 scans from 1213 patients were used for model training and 677 scans from 677 patients were used for model testing.
