Effect of AI-assisted software on inter- and intra-observer variability for the X-ray bone age assessment of preschool children

Abstract

Background With the rapid development of deep learning algorithms and the rapid improvement of computer hardware in the past few years, AI-assisted diagnosis software for bone age has achieved good diagnostic performance. The purpose of this study was to investigate the effect of AI-assisted software on residents’ inter-observer agreement and intra-observer reproducibility for the X-ray bone age assessment of preschool children. Methods This prospective study was approved by the Institutional Ethics Committee. Six board-certified residents interpreted 56 bone age radiographs ranging from 3 to 6 years with structured reporting by the modified TW3 method. The images were interpreted on two separate occasions, once with and once without the assistance of AI. After a washout period of 4 weeks, the radiographs were reevaluated by each resident in the same way. The reference bone age was the average bone age results of the three experts. Both TW3-RUS and TW3-Carpal were evaluated. The root mean squared error (RMSE), mean absolute difference (MAD) and bone age accuracy within 0.5 years and 1 year were used as metrics of accuracy. Interobserver agreement and intraobserver reproducibility were evaluated using intraclass correlation coefficients (ICCs). Results With the assistance of bone age AI software, the accuracy of residents’ results improved significantly. For interobserver agreement comparison, the ICC results with AI assistance among 6 residents were higher than the results without AI assistance on the two separate occasions. For intraobserver reproducibility comparison, the ICC results with AI assistance were higher than results without AI assistance between the 1st reading and 2nd reading for each resident. Conclusions For preschool children X-ray bone age assessment, in addition to improving diagnostic accuracy, bone age AI-assisted software can also increase interobserver agreement and intraobserver reproducibility. AI-assisted software can be an effective diagnostic tool for residents in actual clinical settings.

Artificial Intelligence (AI)-Based Systems for Automatic Skeletal Maturity Assessment through Bone and Teeth Analysis: A Revolution in the Radiological Workflow?

“…HH-boneage.io, on the other hand, is a fully automatic system based on the TW3 method that predicts bone age with an MAE of 0.46 years and an RMSE of 0.62 years when compared to manual determination [17] (Table 2). Zhao et al [69] investigated the positive effects of AI-based software on inter-and intraobserver variability in a recent prospective study, in which six board-certified residents evaluated 56 left-hand wrist radiographs of pre-scholar children aged 3 to 6 years twice: once with and once without the assistance of artificial intelligence. Each resident evaluated the same images in the same way after 4 weeks.…”