Image Processing for Interventional MRI

Download Report

Transcript Image Processing for Interventional MRI

Image Processing for
Interventional MRI
Derek Hill
Professor of Medical Imaging
Sciences
King’s College London
Image Processing for
Interventional MRI
Derek Hill
Professor of Medical Imaging
Sciences
University College London
The team
•
•
•
•
•
•
Kawal Rhode
Marc Miquel
Redha Berboutkah
David Atkinson
Maxime Sermesant
Rado
Andriantsimiavona
• Kate McLeish
• Sebastian Kozerke
•
•
•
•
•
•
•
•
Reza Razavi
Vivek Muthurangu
Sanjeet Hegde
Jas Gill
Pier Lambraise
Cliff Bucknall
Eric Rosenthal
Shaqueel Qureshi
Context
• Interventional MRI provides particular
opportunities and challenges for image
analysis.
– Hostile environment for computers
– “real time” requirements
– Link between acquisition and analysis
Overview
•
•
•
•
Background to XMR guided interventions
Integrating x-ray and MRI
Automatic cathether tracking
Integration of image analysis in acquisition
XMR
• X-ray + cylindrical bore MRI in the same
room
• Becoming main platform for MR guided
interventions
– Resection control in neurosurgery
– Endovascular procedures
• Not ideal for percutaneous procedures
XMR suite at Guy’s
(funded of JREI, Philips Medical Systems and Charitable
Foundation of Guy’s & St Thomas’)
Staff
Patient
XMR System at Guy’s Hospital
 XMR = hybrid X-ray/MR imaging
 Common sliding patient table
 Provides path to MR-guided intervention
XMR suite at Guy’s
Catheter manipulation
Visualizing catheters
• Fast imaging (70 msec per frame)
– TE = 1.3, TR = 2.6
– SSFP sequence (balanced TFE)
– Acquisition: 78 x 96, 80% FOV, 80% acq, SENSE factor 2
(ie: only 25 phase encodes!)
• Carbon dioxide filled balloon as contrast agent
Catheter Manipulation
Images acquired with standard Philips real time or interactive
sequences
Catheter Manipulation
Miquel et al. Visualization and tracking of an inflatable balloon catheter using SSFP in a flow
phantom and in the heart and great vessels of patients. Magn Reson. Med. 51(5):988-95 2004
Integrating x-ray and MRI
• XMR provide rapid transfer between
modalities
• No capability to integrate the images
• X-ray and MRI provide complementary
information
• Combined x-ray and MR has value in
complex interventions eg: electrophysiology
Registration Matrix Calculation
 Overall registration transform is composed of a series of stages
 Calibration + tracking during intervention
M1
Scanner Space
3D Image Space
X-ray Table Space
T
M2
R*P
X-ray C-arm
Space
M3
2D Image Space
XMR Registration:
Software Overview
XMR Registration: Calibration




Acrylic calibration object with 14 markers
Interchangeable caps for MR and X-ray imaging
Determine geometric relationship between MR and X-ray system
Determine X-ray projection geometry
MR
X-ray
Calibration
(1) Fixing flange for sliding table.
(2) Saline-filled acrylic half cylinder
with 20 divot cap markers in a
helical arrangement.
(3) Slot in acrylic base plate to allow
sliding of half cylinder.
(4) & (5) End stops.
(6) Fixing to allow MR surface coil
attachment
XMR Registration:
MR Overlay on X-Ray
XMR Registration:
3D Reconstruction
XMR Registration:
Phantom Validation
T1-weighted MR volume + 5 pairs of
tracked x-ray images using calibration
object as a phantom
2D RMS Error = 4.2mm (n=35), Range =
1.4 to 8.0 mm
3D RMS Error = 4.6mm (n=17), Range =
1.7 to 9.0 mm
Clinical Example
• Patient undergoing electrophysiology study
prior to RF ablation of heart rhythm
abnormality
MR Imaging - Anatomy
 SSFP threedimensional
multiphase sequence
 5 phases
 256x256 matrix
 152 slices
 resolution=1.33 x
1.33 x 1.4 mm
 TR=3.0 ms
 TE=1.4 ms
 flip angle=45
MR Imaging - Motion
 SPAMM tagged imaging
sequence
 59 phases SA & 50 phases
LA
 256x256 matrix
 11 slices SA & 4 slices LA
 resolution=1.33 x 1.33 x
8.0 mm
 TR=11.0 ms
 TE=3.5 ms
 flip angle=13
 tag spacing=8 mm
X-ray Imaging + Electrical Mapping
Contact electrical mapping system
Constellation catheter (Boston Scientific)
LAO View
AP View
MR Anatomy Overlay
Catheter Reconstruction
Refining the Registration
 Errors due to limitations
of registration technique
and patient motion
 Basket point cloud
meshed
 Rigid surface-to-image
registration used to
realign the basket mesh
Visualising the Electrical Data
 Cycle 1 - normal
 Cycle 2 - ectopic
Instantiation of model
Simulation results
LV volume
Catheters re-visited
• Essential properties of
catheters
– Clearly visible
– Safe
• mechanically
• electrically
• Magnetically
• Desirable properties
– Automatic localization
– Tip and length visible
• CO2 filled balloon catheters are
safe
• Tip location ambiguous
• Length not visible
• Cannot be localized automatically
Is there an image analysis
solution?
• Find catheter automatically in modulus
image?
• Is it easier to find in a phase image?
Better solution: change nucleus
• Fluorine is not present
in body
• High NMR sensitivity
• Safe blood subsitutes
available (eg: PFOB)
Catheter tracking
(b)
(a)
+
(c)
Phantom setup
SSFP proton image plus
fluorine projections
Catheter tracking
(b)
(a)
(c)
Phantom setup
Automatic superposition
Of catheter tip on proton image
Lumen visible
Dynamic scan
Catheter Tracking and Visualization
Using
19F Nuclear Magnetic Resonance
•
Sebastian Kozerke1,2, Sanjeet Hegde3, Tobias
Schaeffter4, Rolf Lamerichs5,
Reza Razavi3, Derek L. Hill2
Magn. Reson. Med. 2004 (in press)
Image analysis combined with
acquisition
• Real time MRI can provide high temporal
resolution, but low quality
• Can we subsequently combine real time
images to generate high image quality?
Real time MRI with slice
tracking
• Real time undersampled radial acquisitions
Navigator
Slice tracking
Registration to compensate for
motion
Rigid body then non-rigid registration to correct motion
During scanning
Demonstration on gated
volunteer heart images (n=4)
• Undersampled images
Demonstration on gated
volunteer heart images (n=4)
• Combined with no registration
Demonstration on gated
volunteer heart images (n=4)
• Combined with rigid registration
Demonstration on gated
volunteer heart images (n=4)
• Combined with rigid then non-rigid
registration
Conclusions
• Interventional MRI is fertile area for image
analysis
•
•
•
•
Real time requirements
New applications (eg: RF ablation)
Improving guidance
Novel acquisition and reconstruction
incorporating image analysis