Hormone Research in Paediatrics has published three papers on BoneXpert online.
The first paper verifies the bone age method in Japanese children and determines the precision to be 0.17 years (this is the SD on a single determination of bone age, including the effect of recording a new X-ray). These longitudinal Japanese data are well suited to this analysis, because they were taken at very short intervals, every 6 months.
The third paper is also based on the first Zurich longitudinal study. It sets up a new framework for validation of a bone age method, by studying its ability to predict the adult height. The Zurich data were originally rated manually according to both the GP and the TW methods, and now the automated BoneXpert method is also applied. The study shows that manual GP rating is superior to manual TW rating, which is somewhat surprising, because previous studies of bone age in these data have always used the TW ratings - in particular, the TW3 method for adult height prediction was based on these images and their manual TW rating. Finally the study shows that BoneXpert's GP rating is as good as manual GP-rating. This is quite a strong validation of BoneXpert, because the GP ratings of the Zurich data were done by the experienced and meticulous raters working in Prader's research team; present-day, routine manual GP ratings are likely to be less precise.
Abstracts of all publications can be accessed from the publication page