Abstract:

Objective To explore a deep learning-based automatic bone age estimation model for elbow joint X-ray images of Chinese Han children and adolescents and evaluate its performance. Methods A total of 943 （517 males and 426 females） elbow joint X-ray images of Chinese Han children and adolescents aged 6.00 to <16.00 years were collected from East, South, Central, and Northwest China. Three experimental schemes were adopted for bone age estimation. Scheme 1: Directly input preprocessed images into a regression model; Scheme 2: Train a segmentation network using “key elbow bone annotations” as labels, then input segmented images into the regression model; Scheme 3: Train a segmentation network using “full elbow bone annotations” as labels, then input segmented images into the regression model. For segmentation, the optimal model was selected from U-Net, UNet++ and TransUNet. For regression, VGG16, VGG19, InceptionV2, InceptionV3, ResNet34, ResNet50, ResNet101, and DenseNet121 were selected for bone age estimation. The dataset was randomly split into 80% （754 images） for training/validation for model fitting and hyperparameter tuning and 20% （189 images） as an internal test set to test the performance of the trained model. An additional 104 elbow X-ray images from the same demographic and age group were collected and used as an external test set. Model performance was evaluated by comparing the mean absolute error （MAE）, root mean square error （RMSE）, accuracy within ±0.7 years （P_{±0.7 years}） and ±1.0 years （P_{±1.0 years}） between the estimated age and the actual age, and by drawing radar charts, scatter plots, and heatmaps. Results When segmented with “full elbow bone annotations,” the UNet++ model achieved a segmentation loss of 0.0004 and an accuracy of 93.8% at a learning rate of 0.0001. In the internal test set, DenseNet121 model with Scheme 3 yielded the best results with MAE, P_{±0.7 years} and P_{±1.0 years} being 0.83 years, 70.03%, and 84.30%, respectively. In the external test set, the DenseNet121 with Scheme 3 also performed best, with an average MAE of 0.89 years and an average RMSE of 1.00 years. Conclusion When performing automatic bone age estimation using elbow X-rays in Chinese Han children and adolescents, it is recommended to use the UNet for segmentation. DenseNet121 with Scheme 3 achieves optimal performance. Using a segmentation network, especially one that includes the full elbow joint including the distal humerus, proximal radius, and proximal ulna as the annotation area, can improve the accuracy of bone age estimation based on elbow X-ray images.

Key words: forensic anthropology, age estimation, X-ray image, elbow joint, deep convolutional neural network, segmentation network, Han nationality, adolescents, children