Draft Status
Draft - team will hold off on page creation
Category
Other
Key Investigators
Alexandre Buisson (University of North Carolina at Chapel Hill, USA)
Paul Dumont (University of North Carolina at Chapel Hill, USA)
Juan Carlos Prieto (University of North Carolina at Chapel Hill, USA)
Lucia Cevidanes (University of North Carolina at Chapel Hill, USA)
Steve Pieper (Isomics, USA)
Project Description
Predicting surgical movements and bone displacement vectors in virtual surgical planning software remains an expert-intensive task, requiring surgeons to simulate osteotomies and manually adjust bone segments. Although statistical shape models and deep learning regression networks have been explored to automate this phase, they output dense deformation fields that lack the geometric interpretability needed to guide clinical or surgical decisions.
This project introduces a dedicated 3D Slicer module driven by a Machine Learning Stacking model, trained on a robust dataset of 1,496 patients. The module simplifies the clinical workflow by allowing users to upload an input file (e.g., Excel/CSV containing clinical parameters) and instantly receive accurate, data-driven predictions of the required maxillofacial bone movements.
Objective
-Embed a pre-trained Stacking ML model into the 3D Slicer environment to leverage multi-patient surgical data.
-Enable seamless parsing of input files (Excel/CSV) to extract patient-specific features.
-Generate and display precise bone movement recommendations directly within the Slicer interface based on the model's inference.
Approach and Plan
-Develop the input file parser (Excel/CSV) to clean and prepare patient data for the model.
-Integrate the pre-trained Stacking ML model into a Python-based 3D Slicer scripted module.
-Design a user-friendly Slicer UI that allows clinicians to load patient files, trigger the prediction, and visualize the recommended bone displacements.
Progress and Next Steps
The core Stacking ML model has been successfully trained and validated using a dataset of 1,496 patient cases.
During the project week we'll build an interactive UI and backend pipeline within 3D Slicer to handle file inputs and run the model's prediction pipeline.
Also, we'll verify the accuracy of the outputs within the Slicer environment and explore intuitive ways to display the predicted movements to the user.
Illustrations
No response
Background and References
No response
Draft Status
Draft - team will hold off on page creation
Category
Other
Key Investigators
Alexandre Buisson (University of North Carolina at Chapel Hill, USA)
Paul Dumont (University of North Carolina at Chapel Hill, USA)
Juan Carlos Prieto (University of North Carolina at Chapel Hill, USA)
Lucia Cevidanes (University of North Carolina at Chapel Hill, USA)
Steve Pieper (Isomics, USA)
Project Description
Predicting surgical movements and bone displacement vectors in virtual surgical planning software remains an expert-intensive task, requiring surgeons to simulate osteotomies and manually adjust bone segments. Although statistical shape models and deep learning regression networks have been explored to automate this phase, they output dense deformation fields that lack the geometric interpretability needed to guide clinical or surgical decisions.
This project introduces a dedicated 3D Slicer module driven by a Machine Learning Stacking model, trained on a robust dataset of 1,496 patients. The module simplifies the clinical workflow by allowing users to upload an input file (e.g., Excel/CSV containing clinical parameters) and instantly receive accurate, data-driven predictions of the required maxillofacial bone movements.
Objective
-Embed a pre-trained Stacking ML model into the 3D Slicer environment to leverage multi-patient surgical data.
-Enable seamless parsing of input files (Excel/CSV) to extract patient-specific features.
-Generate and display precise bone movement recommendations directly within the Slicer interface based on the model's inference.
Approach and Plan
-Develop the input file parser (Excel/CSV) to clean and prepare patient data for the model.
-Integrate the pre-trained Stacking ML model into a Python-based 3D Slicer scripted module.
-Design a user-friendly Slicer UI that allows clinicians to load patient files, trigger the prediction, and visualize the recommended bone displacements.
Progress and Next Steps
The core Stacking ML model has been successfully trained and validated using a dataset of 1,496 patient cases.
During the project week we'll build an interactive UI and backend pipeline within 3D Slicer to handle file inputs and run the model's prediction pipeline.
Also, we'll verify the accuracy of the outputs within the Slicer environment and explore intuitive ways to display the predicted movements to the user.
Illustrations
No response
Background and References
No response