Comparison of BoneXpert and IB-Lab-PANDA automated Bone Age Evaluation in Children With Growth and Puberty Disorders.

Abstract

Context: Bone age (BA) evaluation in children presenting growth problems is time-consuming. Artificial intelligence (AI) BA assessment programs are increasingly used. However, agreement between different commercially available methods in the same population, or possible age-, puberty- or sex-related differences have not been sufficiently evaluated.

Methods: BA assessment of 521 left hand radiographs of patients aged 2-19 years with IB-lab-PANDA® and BoneXpert® were compared. Of the 521 radiographs, 213 were compared to the Greulich-Pyle (GP) reference. We analyzed gender, age, diagnosis, body mass index (BMI) and puberty categories. Accuracy was calculated as mean-absolute-deviation (MAD) and root-mean-square-error (RMSE).

Results: MAD was 0.61 years and the RMSE 0.83 years between BoneXpert and IB-lab-PANDA, with poor agreement in girls over 14 years (MAD 1.18 years).Compared to the manual rating, both methods showed a positive bias in boys (0.28 years BoneXpert vs 0.51 years IB-lab-PANDA) and in children with pathologies associated with BA delay (BoneXpert 0.18 years vs IB-lab-PANDA 0.35 years). IB-lab-PANDA underestimated BA in girls after 14 years (-0.67 years). IB-lab-PANDA had a MAD of 0.64 years and RMSE of 0.85 years compared to manual assessment, whereas BoneXpert had a MAD of 0.63 years and RMSE of 0.82 years.BoneXpert was significantly more accurate than IB-lab-PANDA in prepubertal children (MAD 0.7 vs 0.83 years; P = .027).

Conclusion: The direct agreement between IB-lab-PANDA® and BoneXpert® falls within human inter-rater variability. Their agreement on manual BA determination is equivalent except in prepubertal children, where BoneXpert seems more accurate. Both are fast, valuable tools for determining BA accurately and efficiently.