Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Probabilistic Finite Element Model of Functional Spine Units from Deep Learning Segmented Images
Our project at the Orthopedic Biomechanics group in Balgrist Campus in collaboration with Uniklinik Balgrist aims to augment the preoperative imaging data on which a surgeon can rely. This problem is addressed with the development of an adaptable pipeline that can automatically extract features on clinical data of healthy and pathological vertebrae, based on machine learning techniques. These features will be used in different simulation environments like Finite Element models and Multibody Dynamic models to provide information about different surgical procedures.
Keywords: Statistical Deformable Models
Biomedical Engineering
Finite Element Models
Deep Learning
We provide a first implementation of a pipeline that prepares and runs a FE model from segmented vertebrae using the FEBio solver, which is implemented in Matlab. The aim of this Master or Semester Project is to improve this pipeline starting with the implementation of different mechanical properties in the FE model. Different models should be created automatically and tested for different patients. The project involves the following milestones:
1. Literature research. Familiarize with the FEBio FE solver and the Matlab scripts used to create the model.
2. Improve the model by implementing ligaments’ properties, contact definition between vertebrae, and modelling the Intravertebral Disc.
3. Employ the pipeline for many different patient-specific cases to test the robustness of the implementation. Test the sensitivity of the pipeline by changing attachment points of ligaments, intravertebral disc properties, and ligaments’ properties.
4. If time and progress allow, a clinical study with real pathological cases could be performed. This will be planned during the project.
Hypothesis:
- An automatically generated FE model achieves similar results as a manually prepared FE model.
We Provide:
- An existing implementation on which you can rely, this will help you in understanding how the FEBio FE solver works and how such a model can be prepared using Matlab.
- An applied research project addressing a clinical need and aiming at the improvement of patient satisfaction.
- A friendly working environment in close contact with the real clinical world of the Balgrist Clinic.
We provide a first implementation of a pipeline that prepares and runs a FE model from segmented vertebrae using the FEBio solver, which is implemented in Matlab. The aim of this Master or Semester Project is to improve this pipeline starting with the implementation of different mechanical properties in the FE model. Different models should be created automatically and tested for different patients. The project involves the following milestones:
1. Literature research. Familiarize with the FEBio FE solver and the Matlab scripts used to create the model.
2. Improve the model by implementing ligaments’ properties, contact definition between vertebrae, and modelling the Intravertebral Disc.
3. Employ the pipeline for many different patient-specific cases to test the robustness of the implementation. Test the sensitivity of the pipeline by changing attachment points of ligaments, intravertebral disc properties, and ligaments’ properties.
4. If time and progress allow, a clinical study with real pathological cases could be performed. This will be planned during the project.
Hypothesis:
- An automatically generated FE model achieves similar results as a manually prepared FE model.
We Provide:
- An existing implementation on which you can rely, this will help you in understanding how the FEBio FE solver works and how such a model can be prepared using Matlab. - An applied research project addressing a clinical need and aiming at the improvement of patient satisfaction. - A friendly working environment in close contact with the real clinical world of the Balgrist Clinic.
• First goal will be to develop a robust patient-specific Lumbar Spine model generator.
• Second goal will be the analysis of different properties of the model as explained in milestone 3. A possible approach could be to treat model properties as variables that can be changed given the input geometries (parametric variation), resulting in a probabilistic approach in the model creation.
• First goal will be to develop a robust patient-specific Lumbar Spine model generator.
• Second goal will be the analysis of different properties of the model as explained in milestone 3. A possible approach could be to treat model properties as variables that can be changed given the input geometries (parametric variation), resulting in a probabilistic approach in the model creation.
Sebastiano Caprara: sebastiano.caprara@hest.ethz.ch
Marco Senteler, PhD: msenteler@ethz.ch
Department of Orthopaedics, University of Zürich
Institute for Biomechanics, ETH Zürich
Balgrist Campus, Lengghalde 5
Tel: +41 (44) 510 7340