Registration Foundations

Preview:

DESCRIPTION

Registration Foundations. Bring multiple image data sets into anatomical agreement. The Registration Problem. T init. T k. T final. Provided by Lilla Zollei. Applications. multi-modality fusion (same patient?) time-series processing e.g.: MS, f MRI experiments, cardiac ultrasound - PowerPoint PPT Presentation

Citation preview

National Alliance for Medical Image Computing http://na-mic.org

Registration Foundations

• Bring multiple image data sets into anatomical agreement

National Alliance for Medical Image Computing http://na-mic.org

The Registration Problem

Tinit

Tk

Tfinal

.

.

.

.

.

.

Provided by Lilla Zollei

National Alliance for Medical Image Computing http://na-mic.org

• multi-modality fusion (same patient?)• time-series processing

– e.g.: MS, fMRI experiments, cardiac ultrasound

• warping across patients to atlas for labeling• accommodate tissue deformations in image-

guided surgery• image-guided surgery of organs other than head

Applications

National Alliance for Medical Image Computing http://na-mic.org

Manual Registration

• Not too bad with a few data sets

• Re-Position one data set for visual agreement

National Alliance for Medical Image Computing http://na-mic.org

Medical image data sets

Transform (move around)

Compare with objective function

Optimization algorithminitial value

motion parameters

score

Automated Medical Image Registration

Provided by Lilla Zollei

National Alliance for Medical Image Computing http://na-mic.org

Estimate Relationship Among two Signals

• U : a signal

• V : another signal, transformed by

National Alliance for Medical Image Computing http://na-mic.org

Estimate Relationship Among two Signals

• If p(U,V) is Gaussian– Then best f is correlation (or

squared difference)

National Alliance for Medical Image Computing http://na-mic.org

Estimate Relationship Among two Signals

• If p(U,V) is UNKNOWN– Look for strongest statistical

relationship among the signals

I : Mutual Information

National Alliance for Medical Image Computing http://na-mic.org

Mutual Information (MI)

• H : entropy– measures information content

• I : Mutual Information - a statistic that measures lack of statistical independence

National Alliance for Medical Image Computing http://na-mic.org

MI Registration

• Default Method for Multi-Modal Medical Image Registration

• Viola Wells et al. circa 96– Collignon, and Hill & Hawkes

• Pluim et al. Survey, 2003: More than 160 published applications

National Alliance for Medical Image Computing http://na-mic.org

Example MRT Rigid Registration

Pre-operative SPGR MRI Intra-operative T2-weighted MRI

Provided by D. Gering

National Alliance for Medical Image Computing http://na-mic.org

Before Registration After Registration

Provided by D. Gering

Example MRT Rigid Registration

National Alliance for Medical Image Computing http://na-mic.org

Real 3D CT data

National Alliance for Medical Image Computing http://na-mic.org

3D MR data

National Alliance for Medical Image Computing http://na-mic.org

“Real” CT-MR registration:3D starting position

National Alliance for Medical Image Computing http://na-mic.org

CT-MR registration final result

National Alliance for Medical Image Computing http://na-mic.org

•The end.

National Alliance for Medical Image Computing http://na-mic.org

3D Slicer Design

• Cross-platform• Built on VTK

– Open source platform for visualization– GE, industrial strength– C++, Tck/TK GUI

• Open GL – Library interface to graphics hardware

• Easily extended• Open source• Available free: www.slicer.org

National Alliance for Medical Image Computing http://na-mic.org

Estimate intensity correctionusing residuals based on current posteriors.

Compute tissue posteriors using current intensity correction.

M-Step

E-Step

EM-Segmentation

Provided by T Kapur

National Alliance for Medical Image Computing http://na-mic.org

EM Segmentation…

PD, T2 Data

Seg Resultw/o EM

Seg ResultWith EM

National Alliance for Medical Image Computing http://na-mic.org

EM Segmentation: MS Example

Data provided by Charles Guttmann

PD T2

National Alliance for Medical Image Computing http://na-mic.org

EM Segmentation: MS Example

Seg w/o EM Seg with EM

Recommended