Automatic and robust estimation of sex and chronological age from panoramic radiographs using a multi-task deep learning network: a study on a South Korean population

Abstract

Sex and chronological age estimation are crucial in forensic investigations and research on individual identification. Although manual methods for sex and age estimation have been proposed, these processes are labor-intensive, time-consuming, and error-prone. The purpose of this study was to estimate sex and chronological age from panoramic radiographs automatically and robustly using a multi-task deep learning network (ForensicNet). ForensicNet consists of a backbone and both sex and age attention branches to learn anatomical context features of sex and chronological age from panoramic radiographs and enables the multi-task estimation of sex and chronological age in an end-to-end manner. To mitigate bias in the data distribution, our dataset was built using 13,200 images with 100 images for each sex and age range of 15-80 years. The ForensicNet with EfficientNet-B3 exhibited superior estimation performance with mean absolute errors of 2.93 ± 2.61 years and a coefficient of determination of 0.957 for chronological age, and achieved accuracy, specificity, and sensitivity values of 0.992, 0.993, and 0.990, respectively, for sex prediction. The network demonstrated that the proposed sex and age attention branches with a convolutional block attention module significantly improved the estimation performance for both sex and chronological age from panoramic radiographs of elderly patients. Consequently, we expect that ForensicNet will contribute to the automatic and accurate estimation of both sex and chronological age from panoramic radiographs.

Keywords: Age estimation; Deep learning; Multi-task learning; Panoramic radiographs; Sex estimation.

Fig. 1

Examples of panoramic radiographs of males or females aged 15–80 years

Fig. 2

Overview of the proposed multi-task deep learning network (ForensicNet). ForensicNet consists of a backbone with age and sex attention branches. Each attention branch has a convolutional block attention module (CBAM) composed of channel and spatial attention modules. ForensicNet takes panoramic radiographs as inputs and simultaneously estimates sex and chronological age by each attention branch

Fig. 3

Confusion matrices for sex estimation from different backbones. (a)–(h) Results of VGG16, MobileNet v2, ResNet101, DenseNet121, Vision Transformer, Swin Transformer, TransNet, and EfficientNet-B3, respectively

Fig. 4

Box plots for estimation performance of chronological age from different backbones on each age group. Each blue box contains the first and third quartiles of accuracy. Medians are located inside the blue boxes as black lines, with the minimum and maximum values visualized as vertical lines. Black circles are outliers. (a)–(h) Results of VGG16, MobileNet v2, ResNet101, DenseNet121, Vision Transformer, Swin Transformer, TransNet, and EfficientNet-B3, respectively

Fig. 5

Representative estimation results and corresponding Grad-CAM generated by EfficientNet-B3. GT and PR are the ground truth and estimation results, respectively

Fig. 6

Linear regression plots for estimation performance of chronological age from different backbones. Blue dots are observations between ground truth and estimated ages, and the red line denotes a linear regression line. R² is a measure of the goodness of fit of a backbone. (a)–(h) Results of VGG16, MobileNet v2, ResNet101, DenseNet121, Vision Transformer, Swin Transformer, TransNet, and EfficientNet-B3, respectively

Fig. 7

Bland–Altman plots for estimation performance of chronological age by backbones. Blue dots denote the differences between ground truth and estimated ages, the red line presents a mean difference, and black dash lines are 95% limits of agreement. (a)–(h) Results of VGG16, MobileNet v2, ResNet101, DenseNet121, Vision Transformer, Swin Transformer, TransNet, and EfficientNet-B3, respectively

Fig. 8

Representative estimation errors and corresponding Grad-CAM generated by EfficientNet-B3. GT and PR are the ground truth and estimation results, respectively