Work

Deep learning-based algorithms to detect vertebral fractures and osteoporosis using lateral spine X-ray radiograph

Year: 2022

Type: article

Abstract: Introduction: Established scores estimate 10-year fracture risk in osteoporosis to assist with treatment recommendations.This study compares and contrasts the risk probabilities of major osteoporotic (MOF) and hip fractures (HF) calculated by the FRAX tool with those of the 'Dachverband Osteologie' (DVO) score, established in Germanspeaking countries.Methods: This seven-year retrospective study analyzes data of 125 male patients (mean age: 59.2±10.7 years) evaluated for osteoporosis.For the DVO score, the therapy threshold of N30% for vertebral and hip fractures suggested by DVO guidelines was implemented.We calculated fracture risks based on FRAX scores with aBMD and applied a common therapy threshold of ≥3% for hip fracture and subsequently determined the "DVO-equivalent risk level" for FRAX-based assessment that would identify as many male patients as identified by the DVO score.Results: Based on DVO score, 60.0% of patients had a 10-year risk of hip and vertebral fractures N30%.The recommendations for individuals based on FRAX scores for HF with aBMD with risk ≥ 3% overlapped with those based on DVO score in 40.8% of patients.Patients identified for treatment only by DVO score presented a higher percentage of vertebral fractures (65% vs. 29%).The thresholds for this "DVO-equivalent risk level" for 'FRAX with aBMD' would be ≥ 6.7% for MOF and ≥ 2.1% for HF.Conclusion: This study demonstrates that the DVO score was more sensitive than the FRAX score for patients with vertebral fractures.We suggest considering the appropriate score and therapy threshold carefully in the daily care of male patients.

Source: Bone Reports

DOI: 10.1016/j.bonr.2022.101576