Bone Age Assessment — Greulich & Pyle

Bone Age Assessment — Greulich & Pyle Method

by Jonathan Colledge

🔒 Patient identifiers stay in your browser — only an anonymised image is sent to the model server for analysis.

🎯 When assessed on the RSNA test set, this online implementation gives an MAE of 3.51 months (beating current state of the art), MSE 20.12, and RMSE 4.485 months. Best MAE achieved with these models was 3.46 months, but the model selection for the ensemble was aimed at minimising MSE.

Patient & Image

Radiograph file

⇧

Click to upload or drag & drop

JPG · PNG · TIFF · DICOM (.dcm) · or a ZIP of multiple files

Patient sex (optional — overridden by a DICOM tag or ZIP mapping when present)

Male Female

Date of birth (optional)

Date of radiograph (optional)

Reference dataset

Stuart Brush

Centre normal range on (percentiles only strictly valid with "Mean bone age")

Mean bone age Chronological age

Skip preprocessing (rotation, crop, segmentation)

Manual Assessment

Manual bone age (months)

Results

◻

Upload a radiograph and press Analyse

Loading Pyodide…

About this application

This application uses an ensemble of models to predict bone age according to Greulich and Pyle(1). The models were trained on the RSNA dataset(2), the Deeplasia processed RSNA dataset(3), the DHA dataset(4), and the RHPE dataset(5). Hand segmentation models used the Deeplasia segmentation dataset(3) with a further custom segmentation dataset derived from the DHA and RHPE datasets. Several novel methods were used in training and I hope to publish this at some point. The only input to the models is the image. Seperate models are used for each sex. This application is not CE marked or FDA approved.

References

Greulich WW, Pyle SI. Radiographic atlas of skeletal development of the hand and wrist. 2nd ed. Stanford, CA: Stanford University Press; 1959.
Halabi SS, Prevedello LM, Kalpathy-Cramer J, et al. The RSNA Pediatric Bone Age Machine Learning Challenge. Radiology 2018; 290(2):498-503.
Rassmann, S., Keller, A., Skaf, K., Hustinx, A., Gausche, R., Ibarra-Arrelano, M. A., Hsieh, T.-C., Madajieu, Y. E. D., Nöthen, M. M., Pfäffle, R., & others. (2023). Deeplasia: Deep learning for bone age assessment validated on skeletal dysplasias. Pediatric Radiology, 1–14. Springer.
Gertych, A., Zhang, A., Sayre, J., Pospiech‑Kurkowska, S., & Huang, H. K. (2007). Bone age assessment of children using a digital hand atlas. Computerized Medical Imaging and Graphics, 31(4–5), 322–331. https://doi.org/10.1016/j.compmedimag.2007.02.012
Escobar, M., González, C., Torres, F., Daza, L., Triana, G., & Arbeláez, P. (2019). Hand pose estimation for pediatric bone age assessment. In International Conference on Medical Image Computing and Computer‑Assisted Intervention (pp. 531–539). Springer.