Musculoskeletal models help surgeons and researchers to estimate loadings in the spinal segments. Together with finite element simulations, these models form a valuable toolset for surgical planning and spine research. However, patient-specific model creation is a tedious and time-consuming process. Therefore, our research group is developing a completely automatic pipeline that generates individual musculoskeletal models based on bi-planar X-ray images. The radiographs stem from University Hospital Balgrist and were acquired with the EOS System, a low-dose whole-body scanner for simultaneous anteroposterior and lateral X-ray imaging.
The pipeline uses deep convolutional neural networks to identify bony structures on the images. Subsequently, the network output is translated into 3D coordinates for our musculoskeletal model. So far the pipeline processes the parts of the spine, pelvis, and sacrum. But important parts, such as the body shape, head, and arms, are still missing. The goal of the project is to add these parts to the pipeline and to increase its overall robustness.
Our interdisciplinary team of engineers will host the project at the Balgrist Campus in close collaboration with surgeons and medical experts from the University Hospital Balgrist.
Musculoskeletal models help surgeons and researchers to estimate loadings in the spinal segments. Together with finite element simulations, these models form a valuable toolset for surgical planning and spine research. However, patient-specific model creation is a tedious and time-consuming process. Therefore, our research group is developing a completely automatic pipeline that generates individual musculoskeletal models based on bi-planar X-ray images. The radiographs stem from University Hospital Balgrist and were acquired with the EOS System, a low-dose whole-body scanner for simultaneous anteroposterior and lateral X-ray imaging. The pipeline uses deep convolutional neural networks to identify bony structures on the images. Subsequently, the network output is translated into 3D coordinates for our musculoskeletal model. So far the pipeline processes the parts of the spine, pelvis, and sacrum. But important parts, such as the body shape, head, and arms, are still missing. The goal of the project is to add these parts to the pipeline and to increase its overall robustness. Our interdisciplinary team of engineers will host the project at the Balgrist Campus in close collaboration with surgeons and medical experts from the University Hospital Balgrist.
Development of a platform for automatic detection of spinal pathologies and segmentation of disc in 3D.
Tasks:
- Develop and improve deep learning models to detect head and body shape on X-rays (Python & Tensorflow.
- Transform detected shapes into 3D input for the musculoskeletal model (Matlab).
Development of a platform for automatic detection of spinal pathologies and segmentation of disc in 3D.
Tasks:
- Develop and improve deep learning models to detect head and body shape on X-rays (Python & Tensorflow. - Transform detected shapes into 3D input for the musculoskeletal model (Matlab).