This study aims to validate the auxiliary effect of a commercially available AI-powered X-ray bone age analyzer for physicians and to evaluate its impact on the accuracy, consistency, and time efficiency of bone age assessment in Chinese children and adolescents. We conducted a multicenter, prospective study involving 1000 children aged 1 to 18 years across five centers in China. Radiographs were independently assessed for TW3-RUS and TW3-Carpal bone age by five physicians—three juniors and two seniors—both without and with AI assistance. An additional AI group was included, relying solely on the AI analyzer. With AI assistance, the mean absolute error for TW3-RUS ranged from 0.34 to 0.49 for different raters, while for TW3-Carpal, it ranged from 0.32 to 0.46. The mean squared error was between 0.20 and 0.40 for TW3-RUS and between 0.18 and 0.36 for TW3-Carpal. The accuracy varied from 88.20 to 96.80% for TW3-RUS and from 90.60 to 97.20% for TW3-Carpal. Intraclass correlation coefficients exceeded 0.98 for both TW3-RUS and TW3-Carpal. For both the TW3-RUS and TW3-Carpal methods, junior raters exhibited statistically reduced error and increased accuracy with AI assistance (all P < 0.0001). The assistance of AI was also found to significantly benefit senior raters in specific age groups. Evaluation time was also significantly reduced with AI assistance (all P < 0.0001).Conclusion: Incorporating AI into BAA enhances accuracy and consistency, especially for junior raters, while reducing evaluation time. Senior raters also benefit in specific age groups.<Table Float="No" ID="Taba"> <tgroup cols="2"> <colspec align="left" colname="c1" colnum="1" /> <colspec align="left" colname="c2" colnum="2" /> <tbody> <row> <entry nameend="c2" namest="c1"> What is Known: • Bone age assessment (BAA) is crucial for diagnosing and monitoring endocrine and growth disorders in children, traditionally relying on manual interpretation of radiographs, which is time-consuming and subject to variability. What is New: • The incorporation of an AI-powered X-ray bone age analyzer significantly enhances the accuracy and consistency of BAA in Chinese children and adolescents, particularly benefiting junior raters while reducing evaluation time. </entry> </row> </tbody> </tgroup> </Table>

AI-powered automated X-ray bone age analyzer for assisting bone age assessment in Chinese children and adolescents: accuracy, consistency, and time efficiency

期刊论文
发表日期：

2025年4月29日

查看全文

Yan Liang,
Xiumin Wang,
Xiao Bo Chen,
Zhe Su,
Xin Ran Cheng,
Xiao Ping Luo

摘要

This study aims to validate the auxiliary effect of a commercially available AI-powered X-ray bone age analyzer for physicians and to evaluate its impact on the accuracy, consistency, and time efficiency of bone age assessment in Chinese children and adolescents. We conducted a multicenter, prospective study involving 1000 children aged 1 to 18 years across five centers in China. Radiographs were independently assessed for TW3-RUS and TW3-Carpal bone age by five physicians—three juniors and two seniors—both without and with AI assistance. An additional AI group was included, relying solely on the AI analyzer. With AI assistance, the mean absolute error for TW3-RUS ranged from 0.34 to 0.49 for different raters, while for TW3-Carpal, it ranged from 0.32 to 0.46. The mean squared error was between 0.20 and 0.40 for TW3-RUS and between 0.18 and 0.36 for TW3-Carpal. The accuracy varied from 88.20 to 96.80% for TW3-RUS and from 90.60 to 97.20% for TW3-Carpal. Intraclass correlation coefficients exceeded 0.98 for both TW3-RUS and TW3-Carpal. For both the TW3-RUS and TW3-Carpal methods, junior raters exhibited statistically reduced error and increased accuracy with AI assistance (all P < 0.0001). The assistance of AI was also found to significantly benefit senior raters in specific age groups. Evaluation time was also significantly reduced with AI assistance (all P < 0.0001).

Conclusion: Incorporating AI into BAA enhances accuracy and consistency, especially for junior raters, while reducing evaluation time. Senior raters also benefit in specific age groups.

What is Known:

• Bone age assessment (BAA) is crucial for diagnosing and monitoring endocrine and growth disorders in children, traditionally relying on manual interpretation of radiographs, which is time-consuming and subject to variability.

What is New:

• The incorporation of an AI-powered X-ray bone age analyzer significantly enhances the accuracy and consistency of BAA in Chinese children and adolescents, particularly benefiting junior raters while reducing evaluation time.