Consistency between artificial intelligence and expert Greulich-Pyle atlas method for bone age assessment / 中国组织工程研究

ABSTRACT

BACKGROUND:Artificial intelligence-assisted bone age assessment has become a research hotspot.Domestic and foreign studies have shown that the rapid development of artificial intelligence technology in the field of medical imaging provides the possibility of more accurate and rapid assessment of bone age. OBJECTIVE:To investigate the consistency between a domestically developed artificial intelligent Greulich-Pyle(GP)bone age assessment system and an expert manually assessed GP(expert GP),and to provide a basis for the feasibility of applying an artificial intelligent GP in clinical practice or in other fields. METHODS:Wrist radiographs were sampled from children and adolescents aged 6-15 years,of whom 672 were male and 650 were female.Bone age assessment of the same wrist radiograph was performed using the artificial intelligent GP and the expert GP.The accuracy of the artificial intelligent GP reading results was assessed by the absolute value of the difference.The consistency of the bone age results was assessed by Pearson correlation and Bland-Altamn distribution;and the consistency of the assessment was checked by Kappa test. RESULTS AND CONCLUSION:The absolute value of the difference(95%confidence interval)of the difference between artificial intelligent GP and expert GP for male and female was 0.39 years(0.37-0.41 years)and 0.32 years(0.29-0.34 years),respectively.The deviation of Bland-Altamn values for male and female was(-0.096±0.482)years and(0.014±0.415)years,respectively.The correlation results between artificial intelligent GP bone age and expert GP bone age for male and female were r=0.991 and r=0.992,respectively(P<0.000 1).The median difference between all age groups for male and female was within 0.5 years.Kappa test values were greater than 0.4 for both sexes at all ages except for the 9-year age group for male.Overall Kappa values were 0.603 and 0.659 for male and female respectively.To conclude,there is a high degree of consistency between the artificial intelligence and expert evaluation results of GP bone age values in children and adolescents aged 6-15 years.