Forensic age estimation for pelvic X-ray images using deep learning
To develop a deep learning bone age assessment model based on pelvic radiographs for forensic age estimation and compare its performance to that of the existing cubic regression model.
Materials and method
A retrospective collection data of 1875 clinical pelvic radiographs between 10 and 25 years of age was obtained to develop the model. Model performance was assessed by comparing the testing results to estimated ages calculated directly using the existing cubic regression model based on ossification staging methods. The mean absolute error (MAE) and root-mean-squared error (RMSE) between the estimated ages and chronological age were calculated for both models.
For all test samples (between 10 and 25 years old), the mean MAE and RMSE between the automatic estimates using the proposed deep learning model and the reference standard were 0.94 and 1.30 years, respectively. For the test samples comparable to those of the existing cubic regression model (between 14 and 22 years old), the mean MAE and RMSE for the deep learning model were 0.89 and 1.21 years, respectively. For the existing cubic regression model, the mean MAE and RMSE were 1.05 and 1.61 years, respectively.
The deep learning convolutional neural network model achieves performance on par with the existing cubic regression model, demonstrating predictive ability capable of automated skeletal bone assessment based on pelvic radiographic images.
• The pelvis has considerable value in determining the bone age.
• Deep learning can be used to create an automated bone age assessment model based on pelvic radiographs.
• The deep learning convolutional neural network model achieves performance on par with the existing cubic regression model.
KeywordsAge determination by skeleton Forensic anthropology Pelvis Radiography Machine learning
Convolutional neural network
Bone age estimated by the CNN
Bone age calculated by the cubic regression model
Ossification centre of the iliac crest
Kreitner and Kellinghaus ossification staging methods
Mean absolute difference
Receiver operating characteristic
The authors state that this work has not received any funding.
Compliance with ethical standards
The scientific guarantor of this publication is Zhen-hua Deng.
Conflict of interest
The authors declare that they have no conflict of interest.
Statistics and biometry
No complex statistical methods were necessary for this paper.
Informed consent was waived.
This study was performed with the approval of the ethics committee of the West China Hospital of Sichuan University.
Study subjects or cohorts overlap
Some study subjects or cohorts have been previously reported in Zhang K, Dong XA, Fan F, Deng ZH (2016) Age estimation based on pelvic ossification using regression models from conventional radiography. International Journal of Legal Medicine 130:1143–1148.
• Diagnostic or prognostic study
• Performed at one institution
- 6.Lottering N, Reynolds MS, Macgregor DM et al (2016) Apophyseal ossification of the iliac crest in forensic age estimation: computed tomography standards for modern Australian subadults. J Forensic Sci 62:292–307Google Scholar
- 14.Mansourvar M, Ismail MA, Herawan T, Raj RG, Kareem SA, Nasaruddin FH (2013) Automated bone age assessment: motivation, taxonomies, and challenges. Comput Math Methods Med 2013:391626–391626Google Scholar
- 17.Larson DB, Chen MC, Lungren MP, Halabi SS, Stence NV, Langlotz CP (2017) Performance of a deep-learning neural network model in assessing skeletal maturity on pediatric hand radiographs. Radiology 287:170236Google Scholar
- 18.Kim JR, Shim WH, Yoon HM et al (2017) Computerized bone age estimation using deep learning based program: evaluation of the accuracy and efficiency. AJR Am J Roentgenol 209:1Google Scholar
- 21.Mutasa S, Chang PD, Ruzal-Shapiro C, Ayyala R (2018) MABAL: a novel deep-learning architecture for machine-assisted bone age labeling. J Digit Imaging 9:1–7Google Scholar
- 24.Diedrichs V, Wagner UA, Seiler W, Schmitt O (1998) Reference values for development of the iliac crest apophysis (Risser sign). Z Orthop Ihre Grenzgeb 136:226Google Scholar
- 26.Krizhevsky A, Sutskever I, Hinton GE (2012) ImageNet classification with deep convolutional neural networks. International Conference on Neural Information Processing Systems, pp 1097–1105Google Scholar
- 28.Yosinski J, Clune J, Bengio Y, Lipson H (2014) How transferable are features in deep neural networks?. International Conference on Neural Information Processing Systems, pp 3320–3328Google Scholar
- 31.Pinchi V, Norelli GA, Pradella F, Vitale G, Rugo D, Nieri M (2012) Comparison of the applicability of four odontological methods for age estimation of the 14 years legal threshold in a sample of Italian adolescents. J Forensic Odontostomatol 2:17–25Google Scholar
- 32.Pinchi V, Luca FD, Focardi M et al (2016) Combining dental and skeletal evidence in age classification: pilot study in a sample of Italian sub-adults. Leg Med (Tokyo) 20:75–79Google Scholar