Deep Learning Based Quantitative Cervical Vertebral Maturation Analysis

Abstract Objectives: This study aims to enhance clinical diagnostics for quantitative cervical vertebral maturation (QCVM) staging with precise landmark localization. Existing methods are often subjective and time-consuming, while deep learning alternatives withstand the complex anatomical variations. To address these challenges, we introduce an advanced two-stage convolutional neural network customize for improved accuracy in cervical vertebrae analysis. Methods: This study analyzed 2100 cephalometric images. The data distribution to an 8:1:1 for training, validation, and testing. The CVnet system is designed a two-step method with comprehensive evaluation of various ROI sizes was undertaken to locate 19 cervical vertebral landmarks and precision classifying maturation stages. The accuracy of the pinpointing landmarks was assessed by success detection rate and one way ANOVA test. The diagnostic accuracy test was conducted to evaluate system's performance and agreements with six examiners. Results: Upon precise calibration with the optimal region of interest (ROI) size, the landmark localization, registering an average error of just 0.66±0.45 mm and a success detection rate of 98.51% within 2 mm. Additionally, in 210 test samples, the accuracy rate is 69.52% in identifying cervical vertebral maturation stages. Conclusions: This study launched a two-stage neural network that effectively and reliably identifies landmarks and automates the calculation of cervical vertebral maturation stages. Through this approach, the neural network achieved an accuracy rate of approximately 69.52%, resulting in an enhancement of about 10.95% in the accuracy of primary orthodontists' staging. Clinical relevance:The high accuracy and speed of this method in pinpointing cervical vertebrae landmarks are significant for automating skeletal age estimation with CVM techniques. This method could add a valuable information to clinician’s underdeveloped areas or inexperienced to make reliable treatment decision.