Abstract: Objective  To compare the efficacy of artificial intelligence (AI) diagnostic group and artificial reading group in the diagnosis for osteoporotic vertebral compression fractures. Methods  From January 2023 to December 2023, 80 patients with osteoporotic vertebral compression fractures and 20 patients without fractures but with nonspecific low back pain were included in the study. According to the patient 's computed tomography(CT) image, the AI software diagnosis and physicians of different seniority (one senior physician, one intermediate physician and one junior physician) diagnosis were performed. The diagnostic efficacy of different detection methods was compared. Results  The sensitivity, specificity, positive predictive value, negative predictive value and area under the receiver operating characteristic(ROC) curve ( AUC )  and Kappa value of each group were as follows: AI image interpretation: 0.975,0.900,0.975,0.900,0.938, 0.875; senior physician: 0.950, 0.900, 0.974, 0.818, 0.925, 0.819; intermediate physician: 0.825, 0.850, 0.957, 0.548， 0.837, 0.560; and junior physician: 0.750, 0.750, 0.923, 0.429, 0.751, 0.390. Conclusion  The diagnostic performance of AI was comparable to that of senior physician, and significantly higher than that of intermediate and primary physicians.

Key words: artificial intelligence, deep learning, osteoporotic vertebral compression fracture, auxiliary diagnosis, diagnostic efficacy, gold standard