Project: Multimodel registration

Draft Status

Ready - team will start page creating immediately

Presenter Location

Online

Key Investigators

Leroux Gaelle (University of Michigan, USA)
Claret Jeanne (University of Michigan, USA)
Cevidanes Lucia (University of Michigan, USA)
Allemand David (Kitware, USA)
Prieto Juan Carlos (University of North Carolina, USA)

Project Description

The "Multimodal Registration Project" aims to develop a system capable of aligning different medical imaging modalities to enhance diagnostic accuracy and patient outcomes. By focusing on the registration of Cone Beam Computed Tomography (CBCT) with Magnetic Resonance Imaging (MRI), we seek to create a unified imaging model that offers comprehensive insights into patient anatomy and pathology.

Objective

The objective is to register MRI images with CBCT data accurately.

Approach and Plan

Collected a robust dataset comprising MRI and CBCT files.
Compare the performance of three registration approaches : Generative Adversarial Network (GNA), agent-based action learning for non-rigid registration and Elastix's free-form deformation method.
Validate the model's accuracy through rigorous testing against established benchmarks.

Progress and Next Steps

Data collection phase.
Reviewing available papers and code.

Illustrations

No response

Background and References