Virtual Meets Reality in
Image Guided Intervention
4:00PM Integrating Models and Imaging to Guide Interventions
David Hawkes, PhD, FREng, FInstP, FIPEM, Director of the Centre
for Medical Image Computing (CMIC), University College London , and
co-Founder, IXICO Ltd.,d.hawkes@ucl.ac.uk
Moderator: Ron Kikinis, MD, Director of the Surgical Planning Laboratory of
the Department of Radiology, Brigham and Women's Hospital and Harvard Medical
School; Professor of Radiology, Harvard Medical School; Co-Program Leader,
Image Guided Therapy, CIMIT, kikinis@bwh.harvard.edu
Many researchers hope that imaging technology
will soon be used to provide valuable, real-time feedback to surgeons, allowing
them to perform procedures more quickly and more precisely. As currently conceived, image-guided surgery
will involve many steps: pre-procedural imaging, surgical planning, the
superimposition of the plan onto the pre-procedural image, and navigational
imaging during surgery.
Navigational imaging
involves mapping two-dimensional images obtained during surgery via onto
three-dimensional images obtained prior to surgery. The pre-procedural images are often obtained
using computed tomography (CT) or magnetic resonance imaging (MRI), and the
images taken during surgery are usually obtained via X-ray imaging or
fluoroscopy. In the process of mapping
2-D images onto a 3-D image, researchers will hopefully be able to compensate
for deformations due to motion of the patient during surgery. This type of compensation would improve the
efficacy of various ablation therapies such as radiofrequency ablation,
photodynamic therapy, high-intensity focused ultrasound (e.g. in the prostate),
and targeted radiotherapy. It is hoped
that image-guided techniques will someday help surgeons maximize the dose of
therapy delivered to a target area while minimizing the damage inflicted upon
surrounding tissue.
The future of surgery and other invasive procedures may be
intertwined with that of imaging.
Functional imaging that accounts for motion will be necessary in
addition to traditional anatomical imaging, and intra-operative navigational
imaging must be improved.
5:00PM Functional
Hierarchy: Representation and Modeling of Spatial Patterns of Activation in
fMRI
Polina Golland, PhD,
Assistant Professor, EECS Department and Computer Science and Artificial
Intelligence Laboratory (CSAIL), Massachusetts Institute of Technology, polina@csail.mit.edu
Moderator: Ferenc Jolesz, MD, Director, National Center for Image guided
Therapy and Director of the Division of MRI, Brigham and Women's Hospital;
Professor of Radiology, Harvard Medical School; Director of the Image Guided
Therapy Program, BWH; Co-Program Leader, Image Guided Therapy, CIMIT,
fjolesz@partners.org
Experiments
involving
functional magnetic resonance imaging (fMRI) are performed frequently, yet
researchers are still searching for representations of fMRI data that connect
directly to anatomy and make the data easy to interpret. A variant of traditional MRI imaging, fMRI
measures magnetic differences between oxygenated hemoglobin and deoxygenated
hemoglobin. If a particular area of the
brain works harder than its baseline activity level, then blood rushes into
that area, and the area lights up on the scanner. During an fMRI experiment, a person in a
scanner performs a particular task, such as identifying images. Each collected data point is
four-dimensional, with three spatial coordinates and one time coordinate. The areas of the brain in which blood flow
increases when a person performs a task are assumed to be involved with that
task, and these areas form a network.
The traditional method of defining a
network begins with choosing a user-selected “seed” region of
interest. The time course of each region
of the brain is compared to that of the seed region, and the correlation of the
two is measured. If the correlation is
above a given threshold, then the region in question is said to coactivate with
the selected seed. A binary map of the
brain is created, showing regions that coactivate with the seed and those that
do not. This method of defining a
network, however, is not the only possibility.
Polina Golland, of MIT, is investigating a new method of defining
networks. Her technique involves identifying
interesting seed regions and simultaneously associating each piece of the brain
with the appropriate seed region, thus defining networks. As the number of possible seed regions
increases, larger networks break into smaller networks. Detected patterns of co-activation seem to be
inherently hierarchical. The brain can
be anatomically represented as a tree of large systems that branch into smaller
systems, and it seems that the brain can be functionally described in a similar
fashion. Golland’s lab has shown that
the brains of different people have similar functional tree structures. In the future, Golland hopes to use fMRI to
see if populations differ in terms of functional brain structure.