Answers are grounded in this paper and cite their sources.
BACKGROUND: An osteoporosis (OP) diagnosis technique based on routine CT examinations, which allows detecting radiological signs of OP, is currently being actively implemented. Given the issue of underdiagnosed compression fractures (CFs) on CT images, radiologists could benefit from artificial intelligence (AI) services. AIM: This study aimed to assess the potential use of AI services for OP diagnosis based on routine CT findings for opportunistic screening. METHODS: The project involved three health facilities (HFs). Chest CT scans obtained in these HFs between October 2022 and October 2023 in patients over 50 years of age were selected, in which AI services detected signs of OP (CFs and/or reduced vertebral bone density). All cases were re-evaluated by radiologists to identify potential errors made by the service. The final list of patients eligible for dual-energy X-ray absorptiometry (DXA) to confirm osteoporosis was provided to attending physicians in each participating HF. RESULTS: Over a 12-month period, AI services analyzed 5394 CT scans. CFs and/or reduced vertebral bone density were identified in 1125 patients. Patients with a previously confirmed OP, as well as those who refused or were unable to undergo further testing, were excluded. A total of 66 patients underwent DXA. Age ranged from 54 to 86 years; the median (Q1–Q3) age was 70 (62–74) years; the male to female ratio was 21% and 79%, respectively. According to DXA findings, bone mineral density (BMD) values consistent with OP, osteopenia, and normal BMD were reported in 26 patients (39.4%), 37 patients (56.1%), and 3 patients (4.5%), respectively. Diagnostic performance metrics were calculated for both DXA and CT-based vertebral bone density assessment, with sensitivity of 0.71 vs. 0.91, specificity of 0.80 vs. 0.55, and accuracy of 0.76 vs. 0.67, respectively. Significant differences were observed between osteoporosis, osteopenia, and normal BMD groups, as well as between age-norm groups and those identified by AI services (p 0.001). CONCLUSION: The results support the use of AI services for diagnosing OP based on routine CT examinations as part of opportunistic screening.
No takes yet. Share an insight, caveat, or question.
Artyukova et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: