ASTRO poster IBLS

1
MRI Distortion Correction for Submillimeter-Precis
ion in Functional Proton Radiosurgery
Tom S. Lee (1), Keith E. Schubert(1), Dominik
Slusarczyk(2), and R.W. Schulte (3) (1)California
State University, San Bernardino, CA (2) Harvey
Mudd College, Clairmont, CA (3) Loma Linda
University Medical Center, Loma Linda, CA
This research was funded by the Henry L.
Guenther Foundation
Purpose In preparation of a proton radiosurgery
program for functional brain disorders and
epilepsy, we are performing studies to define and
improve the accuracy of MRI-based stereotactic
localization of intracranial targets. MRI
localization is not without problems since it has
a poor geometric fidelity 1. In this report,
we summarize the methods and results of MRI-based
stereotactic localization studies in a phantom
performed in comparison to precision dimensional
inspection by a metrology laboratory.

• Distortion Correction Methods
• Internal Distortion Correction
• Large FOV filter (internal correction of gradient
  nonlinearity).
• Automated 3D shimming with object in place
  (corrects for B0 field inhomogeneities).
• External Distortion Correction 2,3
• Obtain 3D mpr sequences of cubic phantom with
  axial, coronal, and sagittal partitioning.
• Extract inner surfaces using the Canny edge
  detector implemented in Matlab (Mathworks, Inc.)
• Select useful slices and remove unwanted features
  (e.g., air bubbles, holes, drain plug).
• Reconstruct ideal phantom surfaces from fitted
  mid-planes.
• Model relationship between distorted and
  undistorted coordinates as linear combination of
  homogeneous polynomials up to 4th grade.
• Estimate parameters of distortion correction
  functions and apply distortion correction to all
  images.

Distorted faces, produced by the MR scanner, and
the ideal surface planes describing the cubic
phantom that produced them. The axis dimensions
are pixels. Note the different axis scales to
make the distortion visible.

• Objective
• The two aims of this work were
• To quantify the typical errors involved in a
  stereotactic localization procedure based on MRI.
• To find ways to minimize these errors.
• Possible ways to minimize the distortion errors
  inherent in MRI include
• Use of a MRI-compatible stereotactic frame.
• Use of all available internal correction methods
  provided by the manufacturer of the MRI scanner.
• Use of a scanner sequence that is less prone to
  distortion.
• Use a specially developed correction procedure
  for distortions introduced by gradient
  nonlinearities based on the scan of an oil-filled
  cubic phantom.

The Leksell System for stereotactic MRI
localization consisting of the G frame and the
MRI indicator providing 9 fiducials in each image.
Validation Study The Lucy phantom was
dimensionally inspected (Dimensional Metrology
Lab Inc. Riverside, CA) while being mounted to
the Leksell G frame and stereotactic coordinates
of its MR markers were measured with a precision
0.05 mm. The Lucy phantom was stereotactically
scanned on a 1.5 T Sonata MR scanner (Siemens
Med. Solutions) using 3D mpr sequences
partitioned in axial and coronal planes (30 cm
FOV, 512x512 matrix, 2mm partitions). The
MRI-derived coordinates were transformed to
stereotactic coordinates using the method
described by Weaver et al. 4. The
transformation was derived from the position of
the 9 MRI indicator fiducials on three
well-separated axial images. A shift correction
of 1 mm was applied to all diagonal fiducials to
account for susceptibility differences between
fluid-filled marker channels and surrounding
plastic. Both corrected and uncorrected MR images
were used for localization.
Lucy phantom and MR indicator mounted on the
Leksell G frame in preparation for stereotactic
MRI.
The Lucy QA phantom provides MR compatible
markers.
Results and Conclusion
The combination of internal and external
correction methods resulted in a consistent
reduction of the mean absolute stereotactic
localization error to less than 1 mm 0.3 mm. The
next aim in this work is to validate this
accuracy in vivo by targeting the rat pituitary
gland with a high focal proton dose distribution.
Materials The precision 3D Lucy QA phantom
(Fluke Biomedical) provided MRI-compatible target
markers in 20 different locations. The Leksell G
stereotactic system (Elekta Oncology Systems )
was utilized for stereotactic localization. A
custom-made oil-filled cube of (16)3 cm3 was used
to map the nonlinearity of the three MRI
gradients.
Custom-made oil-filled cube to map non-linear
gradient field distortions.

• References
• Paul S. Morgan, Spatial Distortion in MRI with
  Application to Stereotactic Neurosurgery" (PhD
  dissertation, The University of Nottingham, UK,
  1999), 80-85.
• S. Langlois, M. Desvignes, J.M. Constans, and M.
  Revenu, MRI geometric distortion a simple
  approach to correcting the effects of non-linear
  gradient fields," J. Magn. Reson. Imaging, vol.
  9, 821-831, 1999.
• Thomas S. Lee, Software-Based MRI Gradient
  Nonlinearity Distortion Correction (MSc
  dissertation, Calstate University San Bernardino,
  CA, 2006), 37-53.
• K. Weaver, V. Smith, J.D. Lewis, B. Lulu, et al.,
  A CT based computerized treatment planning
  system for I-125 stereotactic brain implants,"
  Int. J. Radiat. Oncol. Biol. Phys., vol. 18, pp.
  445-454, 1990.

The authors would like to thank Dr. Barbara
Holshouser and David Kittle from the Dept. of
Radiology at LLUMC, Jon Miller from Permedics,
Inc., and Dr. John Kirsch from Siemens for
valuable advice and discussions.