VERTE-X Cohort
Data from individuals who underwent lateral spine radiographic examination at Severance Hospital, Seoul, Korea, between January 2007 and December 2018 were reviewed to construct a derivation cohort (the VERTEbral fracture and osteoporosis detection in spine X-ray [VERTE-X] cohort).
This study was approved by the Institutional Review Board of Severance Hospital, Seoul, Korea, with the waiver for written informed consent for medical-record review (IRB no. 4-2021-0937).
The study index date was defined as the date of lateral spine radiographic examination. Among >140,000 spine X-ray scans in 52,466 individuals as potential candidates for the derivation set, the following criteria were applied to create a final cohort with reasonable data size that allow manual labeling process for vertebral fracture within limited resources, while maximizing the value and quality of dataset with the extensibility for subsequent studies. Individuals were excluded if they were under 50years of age based on clinical relevance and lower prevalence of vertebral fracture or osteoporosis (n=19,820). Individuals were of non-Korean ethnicity (n=2), had had any bone metastasis or hematologic malignancy within 1year before the index date (n=648), or had severe scoliosis, kyphosis (n=46), or with missing Digital Imaging and Communications in Medicine (DICOM) files (n=46) were also excluded.
Of the remaining 31,496 individuals, those without any follow-up lateral spine radiographic examination at least 28days after the index date (n=22,220) were also excluded to enable the validation of incident vertebral fracture outcomes for potential follow-up research. A total of 9276 individuals with 26,299 lateral spine radiographic images were included in the final derivation cohort; they were randomized into training (n=5568, 60%), validation (n=1856, 20%), and test sets (a hold-out set, n=1852, 20%), using random split stratified by the age group, sex, prevalence of outcomes, and availability of DXA testing results.
