Hand Pose Estimation for Bone Age Assessment

Collaboration with Dr. Gustavo Triana, Radiology Department, Fundación Santa Fe de Bogotá
*Both authors contributed equally to this work
We present a new experimental framework for the task of Bone Age Assessment (BAA) based on a local analysis of anatomical Regions Of Interest (ROIs) of hand radiographs. For this purpose, we introduce the Radiological Hand Pose Estimation (RHPE) Dataset, composed of 6,288 hand radiographs from a population that is different from the currently available BAA datasets.
We provide Bone Age groundtruths annotated by two expert radiologists as well as bounding boxes and keypoints denoting anatomical ROIs annotated by multiple trained subjects. In addition to RHPE, we provide bounding boxes and ROIs annotations for the publicly available BAA dataset by the Radiological Society of North America (RSNA) [9].
We propose a new experimental framework with hand detection and hand pose estimation as new tasks to extract local information for BAA methods. Thanks to its fine-grained and precisely localized annotations, our dataset will allow to exploit local information to push forward automated BAA algorithms. Additionally, we conduct experiments with state-of-the-art methods in each of the new tasks. Our proposed model, named BoNet, leverages local information and significantly outperforms state-of-the-art methods in BAA. We provide the RHPE dataset with the corresponding annotations, as well as the trained models, the source code for BoNet and the additional annotations created for the RSNA dataset.

Publications

Hand Pose Estimation for Bone Age Assessment

M. Escobar*, C. González*, F. Torres, L. Daza, G Triana and P. Arbeláez

22nd International Medical Image Computing and Computer Assisted Intervention (MICCAI), 2019

Bone Age Assessment Resources

We created the Bone Age Assessment Resources (BAAR) as a platform for promoting the development of BAA algorithms. In the BAAR you can download the RSNA and RHPE datasets with keypoints, bounding box and boneage annotations for the training and validation sets. Additionally, you can explore an overview of the methods BCV has developed for this task. Finally, the BAAR include an evaluation server for the test set of RHPE and RSNA.

Method

Results

Table 1. BAA results on the RSNA and RHPE test sets..

Experiment


Training on RSNA
  • Baseline (full image)
  • BoNet (full image)
  • Baseline + cropped image
  • BoNet + cropped image

  • Training on RHPE
  • Baseline (full image)
  • BoNet (full image)
  • Baseline + cropped image
  • BoNet + cropped image


  • Training on RSNA + RHPE
  • Baseline (full image)
  • Baseline + cropped image
  • BoNet + cropped image
  • MAD


  • 4.45
  • 4.37
  • 4.20
  • 4.14

  • 8.57
  • 7.78
  • 8.05
  • 7.60

  • RSNA

  • 4.41
  • 4.09
  • 3.85
  • RHPE

  • 8.25
  • 7.99
  • 6.86
  • Downloads

    Citation

    @inproceedings{escobar2019hand,
    title={Hand Pose Estimation for Pediatric Bone Age Assessment},
    author={Escobar, Mar{\'\i}a and Gonz{\'a}lez, Cristina and Torres, Felipe and Daza, Laura and Triana, Gustavo and Arbel{\'a}ez, Pablo},
    booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
    pages={531--539},
    year={2019},
    organization={Springer}
    }
    • Address:
      Cra. 1 E No. 19A - 40. 111711, Bogotá, Colombia - Mario Laserna Building - School of Engineering - Universidad de Los Andes
    • Phone:
      [571] 332 4327, 332 4328, 332 4329