Identification of Adolescent Menarche Status Using Biplanar X-ray Images: A Deep Learning-Based Method

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Adolescent Health

Background:

• Accurate assessment of menarche status is crucial for adolescent health monitoring and reproductive health management.
• Current methods for determining menarche status can be invasive or rely on self-reporting, introducing potential inaccuracies.
• Radiographic assessment offers a non-invasive alternative, but manual interpretation can be time-consuming and subjective.

Purpose of the Study:

• To develop and validate a deep-learning-based automated method for identifying adolescent menarche status using EOS (Ensemble of Stars) radiographs.
• To create an algorithm capable of analyzing skeletal maturity indicators from EOS images to predict menarcheal status.
• To establish a computationally efficient and accurate tool for clinical use in assessing female adolescent skeletal development.

Main Methods:

• Development of a deep-learning algorithm comprising a region of interest detection network and a classification network.
• Training and validation using a retrospective dataset of 738 adolescent EOS cases with a five-fold cross-validation strategy.
• Testing on an independent clinical validation set of 259 adolescent EOS cases to assess real-world performance.

Main Results:

• The algorithm achieved high performance on the clinical validation set, with an accuracy of 0.942, macro precision of 0.933, macro recall of 0.938, and macro F1-score of 0.935.
• Excellent performance in distinguishing between males and females was observed.
• For females, the algorithm demonstrated accuracy of 0.910, sensitivity of 0.943, and specificity of 0.855 in estimating menarche status (AUC=0.959), with a kappa value of 0.806 indicating strong agreement.

Conclusions:

• The developed deep-learning algorithm provides an efficient and accurate automated method for identifying adolescent menarche status from EOS radiographs.
• The algorithm shows strong potential for clinical application, assisting healthcare providers in decision-making regarding adolescent skeletal maturity and reproductive health.
• This automated approach can streamline the analysis of EOS radiographs, offering a reliable reference for clinical practice.