Welcome to the FACE: Fully-Automated Condylar Remodeling Evaluation ! ;-)

by Wei Jia, Han Wu, Lanzhuju Mei, Jiamin Woo, Minjiao Wang⁺ and Zhiming Cui⁺

[Project Page]

1. What is condylar remodeling?

• Forces applied during and after orthognathic surgery can induce changes in the condyle's morphology, position, and function, i.e. condylar remodeling.
• Several systematic reviews have emphasized that condylar change should be considered a critical indicator for surgery prognosis.
• This evaluation is critical for understanding the impact of orthognathic surgery on the TMJ, thereby guiding follow-up care.

2. How to perform condylar remodeling evaluation?

• Manual method

  • acquire images during pre- and post-operative phases
  • segmenting and registering condyles manually
    • observer-variant interpretation for condyle
    • experience-demanding
    • time-consuming
  • rough intuitive evaluation

• Our proposed automated method

• acquire images during pre- and post-operative phases
• segmenting and registering condyles from images using V-Net and landmarks.
• overall evaluations
  • qualitative assessment
    • heatmap : [blue, red] → [resorption, hyperplasia]
    • registration mesh : [blue, yellow] → [T0, T1]
  • quantitative assessment
    • Δ D: local pair points distance change with T0 as reference
    • Δ V: volume change with T0 as reference

3. To implement the evaluation of our method

• Getting Started Run the following command to install the required packages:

git clone https://github.com/WeiJiaFiona/JoD_Fully_Automated_Condylar_Remodeling_Evaluation.git
conda env create -f FACE_environment.yaml
cd code

• Data Preparation

  • Put your dataset into the folder 'data'.
  • The data names should be formulated into

    code/data/
    ├── 1
    │   ├── T1_img.nii.gz
    │   ├── T2_img.nii.gz
    ├── 2
    │   ├── T1_img.nii.gz
    │   ├── T2_img.nii.gz
    ├── ...
    └── evaluation
        ├── heatmap
        ├── distance_changes.json
        └──volume_changes.json
    

• Condylar remodeling evaluation

  • Landmark inference

    python two_stage_keypoint_detection.py
    

  • Mandible segemntation

    python two_stage_bone_segmentation.py
    

  • Condylar and ramus segmentation

    python crop.py
    

  • apply ICP registration to transfer mesh

    python ICP regist.py
    

  • Remodeling evaluation

    python evaluation\heapmap_and_distance_change.py
    python evaluation\volume_change.py
    

• Evaluation process and results

  • The data generated during evaluation is saved to code\data\{ID}
  • The evaluation results are in the file code\data\evaluation

3. Some visualization results on the test set