Prostate seed brachytherapy

1
Motivation

• Prostate seed brachytherapy
• Intra-operative 3D dosimetry based on C-arm
  imaging TRUS
• 60 of practitioners have C-arm
• No one can use fluoro today for quantitative
  dosimetry

2
Our Suggested Approach
3
Role of Fluoroscopy Seed locations
4
Why fluoroscopy is so difficult?

• Calibration
• C-arm Pose Recovery
• Seed reconstruction
• Registration of X-Ray to US

angle?
5
Prior Works Pose Recovery by Tracking Hardware

• ?
• Extra Hardware
• Intrusive
• Expensive
• Cumbersome

6
Our Approach Stationary X-ray Fiducial
?
Source1
Source2

• Non-intrusive
• No expensive tracker
• Also solves
• C-arm calibration
• US?X-ray registration if mounted on TRUS

Image1
Image2
Register each image to the fiducial
7
The Correspondence Problem and Triangulation
8
Software package for Quality Control of C-arm
distortion using the FTRAC fiducial Project
description Most C-arms suffer from image
distortion due to various reasons. For each pose,
the image distortion is modeled using a high
degree curve (CIS-1), evaluated using a
distortion correction grid. This solution does
not assume any underlying pattern in the
distortion and hence explicitly corrects at every
3D pose. The question is can we understand
these properties to simplify the distortion
correction procedure. Skills required CIS-1,
mathematical ability, linear algebra Programming
language Matlab. Resources C-arm, calibration
phantom and software in Matlab. Deliverable A
software for clinical use. Extension Can be
extended to a masters project, thesis or a
paper. Mentor Ameet Jain, can meet once every
7-10 days. Number of students 2
9
Virtual Fluoroscopy using mis-calibrated
C-arms Project description Most C-arms suffer
from pose dependant calibration, needing them to
be calibrated at every pose. We suspect that some
clinical procedures do not need such extensive
calibration (great news ?) and have some
theoretical and experimental results supporting
the same. We now need to evaluate the following
if we poke needles into tumors under X-ray
guidance, how accurate does calibration need to
be. Theoretical bounds are already available and
will be compared to the experimental ones. Skills
required Computer vision/X-ray imaging, CIS-1,
math, linear algebra, calibration, registration,
basic system integration of many available
components. Programming language Matlab, C for
optical tracking of needle. Resources C-arm,
calibration phantom and various softwares in
Matlab. Deliverable Some phantom experimental
results. Extension Can be extended to a masters
project, thesis or a paper. Mentor Ameet Jain,
can meet once every 7-10 days. Number of
students 2-3
10
2D-3D registration of pre-operative seed
locations to intra-operative X-ray
images Project description An alternate way to
track the C-arm is to use the inserted-seeds as a
phantom for tracking. To do this, a package is
currently being developed to both track and
reconstruct seeds at the same time. To
_initialize_ the package, an initial pose needs
to be estimated. This can be done using the
already known intended 3D locations of the seeds
(from preoperative plan) and the X-ray image
currently taken. A 2D-3D registration needs to be
done. Skills required Computer vision/X-ray
imaging, CIS-1, math, registration. Programming
language Matlab. Resources Clinical and
phantom data. Various softwares in
Matlab. Deliverable A Matlab function and some
results. Extension Can be extended to a masters
project, thesis or a paper. Mentor Ameet Jain,
can meet once every 7-10 days. Number of
students 2
11
Intensity based pose estimation using the FTRAC
fiducial Project description We have validated
the use of our fluoroscope tracking and
calibration (FTRAC) fiducial. Presently geometric
features of the fiducial are segmented from the
X-ray image and an optimization done. To do away
with segmentation part of it, the optimization
can be conducted directly on the intensity
values. Skills required Computer vision/X-ray
imaging, math, basic optimization Programming
language Matlab Resources Lab computers, FTRAC
pose estimation code (geometric), both simulation
and processed phantom data with ground truth is
available Deliverable Extend the geometric code
to an intensity based pose estimation by adding a
few Matlab functions. Extension Can be extended
to a masters project, thesis or a paper. Mentor
Ameet Jain, can meet once a week. Number of
students 2
12
Spurious seed correction and orientation
reconstruction in MARSHAL Project description
The correspondence/seed matching problem has been
solved by us by proposing the MARSHAL algorithm
and extending it to solve for hidden seeds.
MARSHAL has also been extending to incorporate
seed orientation. We need to integrate these two
together. Also try and correct for some spurious
seeds. Skills required Computer vision/X-ray
imaging, math, some combinatorial
intuition Programming language Matlab Resources
Matlab software functions, data. Deliverable
Extend and integrate the code. Some extended
Matlab functions. Mentor Ameet Jain, can meet
once a week. Number of students 2-3