Pham Trung Viet

v2.0.1

EN

Dark
Light

Automated Spine segmentation and analysis

A complete pipeline that analyse the spine region using deep learning
Python Tensorflow Deep Learning

About the Project:


The end goal of the project is to develop a complete pipeline that could take in 3D CT scan of the vertebrae region and patient's information to output the stress, strength and displacement of the patient's spine. This process can help doctor to access patients' spine for potential injury that may occur. This procedure ususally take days to complete and our model pipeline could reduce this duration to just within hours.

At the moment, the 3D segmentation was done and I am working on the mapping of spine coordinate to the stress, strength and displacement value.

Spine CT scan raw

Segmented spine

The model used for segmentation was adapted from the top achieved project in the Verse2019 and Verse2020 challenges, Coarse to Fine Vertebrae Localization and Segmentation with SpatialConfiguration-Net and U-Net. This model gave us a good start for our pipeline due to its precised and accurate spine localisation and segmentation. The model process 3D CT scan image in the format .nii.gz