US_CT_VertebraRegistration

Back to Projects List

US-CT Vertebra Registration

Key Investigators

• Houssem Gueziri (Montreal Neurological Institute, Montreal, Canada)
• Tamas Ungi (Queen's University, Kingston, Canada)

Project Description

This project aims at evaluating the feasibility of percutaneous US to CT image registration, on a porcine dataset, for minimally invasive spine surgery. The goal is to combine the registration method for open surgery implemented in IBIS with the segmentation/bone enhancement method in AIGT.

Objective

1. Read/Write US data acquired with IBIS into Slicer.
2. Segment the vertebral surface of US data obtained from porcine cadavers
3. Register segmented images with CT images

Approach and Plan

1. Convert the data from IBIS acquisitions to ultrasound sequences
2. Generate ground truth segmentation from CT images
3. Use AIGT to train model for axial image segmentation - Video tutorial
4. Use segmented data with IBIS registration and evaluate registration

Progress and Next Steps

• Align US images with CT using ground truth transform and export data as Sequences readable in Slicer ✔️
• Segment data using pre-trained model:heavy_check_mark:
• Generate ground truth segmentation and train model [TODO]
• Segment data with fine-tuned model [TODO]

Illustrations

Data processing workflow:

Generate aligned CT-US data:

Segmentation with pre-trained model:

Background and References