Hand Bone Age AI
David Larson MD, Bao Do MD
Hand Bone Age AI predicts pediatric skeletal maturity using a left-hand x-ray and gender. This model was trained on 12,611 radiographs using the 2017 RSNA Challenge dataset from Children's Hospital Colorado and Lucile Packard Children's Hospital at Stanford. On the hold out test set, the ensemble variant of this model achieved an MAD 4.59 months (top 5 best results in the 2017 Challenge were 4.2, 4.4, 4.4, 4.5, and 4.5); this more efficient web deployed variant, which runs on a single ConvNeXt backbone, achieved an MAD 5.09 months (~2 weeks from 4.2). Using the Digital Hand Atlas as an external test, this web model achieved an MAD of 0.48 years or 5.76 months and RMS 0.68.
Requirements: Use LEFT hand x-ray in standard position. Do not rotate. Privacy: Images are not uploaded, remaining strictly on your computer to ensure complete data privacy.
Sample cases: [F13], [F7], [M14]
Hand Bone Age AI is a free resource for education and research and should be used as a supplement with standard clinical references like the Greulich and Pyle atlas. Not for diagnostic use.
Loading
Model loading...