Vascular Image Registration for Intra- Operative 3D Ultrasound Annotation

1
Vascular Image Registration for Intra-
Operative 3D Ultrasound Annotation
Julien Jomier, Sue Weeks, and Stephen R.
AylwardComputer Aided-Diagnosis and Display Lab
- University of North Carolina at Chapel Hill

• Introduction
• - 20-30 of liver lesions seen on CT/MR are not
  visible under ultrasound.
• Goal Augmenting the intra-operative ultrasound
  with pre-operative findings.

Liver, Vessels and Lesion Modeling From the
pre-operative CT/MR we model 1) Liver using
watershed segmentation algorithm. 1 2) Vessels
using a ridge traversal and width estimation
technique. 2 3) Lesion using connected
components algorithm.

• Application Radio-frequency ablation (RFA).
  Radiologists guide a needle to any pre-defined
  site within cirrhotic livers. thereby treating
  the patients without resorting to more invasive

procedures or waiting until their lesions enlarge.
- Method 3D vascular image registration.

• Ultrasound Probe Tracking
• - Magnetic or optical tracker attached to the
  probe.
• Bring the model and the ultrasound data into
  alignment.

Display techniques - 3D Immersive
(Stereo-vision).

• Vascular Registration
• Register the vasculature modeled from the
  pre-operative data with the 3D ultrasound
  image.3
• The match metric F is defined as

• Require a more precise registration.

• Needle point of view.

• Standard Ultrasound with reconstructed tumor.

R rotation o offset w weight n
samplesk capture range ? blurring factor I
image data

• Results
• Registration of an inhale CT scan with an exhale
  scan takes 2-3 seconds.
• Monte-Carlo experiments registering a model for
  MR with an Ultrasound show sub-voxel consistency.

3) Ultrasound data of a phantom can be annotated
with target locations from a CT scan with an
average error of 2.0 mm and with a maximum error
of 2.8 mm.
1 Vincent, L., and Soille, P., Watersheds in
Digital Spaces An Efficient Algorithm Based on
Immersion Simulations IEEE PAMI,No. 6, June
1991, pp. 583-598. 2 Aylward S, Bullitt E,
"Initialization, Noise, Singularities, and Scale
in Height-Ridge Traversal for Tubular Object
Centerline Extraction ," IEEE Transactions on
Medical Imaging, Feb, 2002, Pages 61-75 3
Aylward S, Jomier J, Weeks S, Bullitt E,
"Registration of Vascular Images," Accepted,
International Journal of Computer Vision,
November 2003, pages 15 4 Insight Software
Consortium, The insight toolkit Segmentation
and Registration toolkit, http//www.itk.org
CARS - June 2004