Cone Beam CT & Imaging Biomarkers

Cone Beam
Computed Tomography
and Imaging Biomarkers
Michael W
W. Vannier
Vannier, M
M.D.
D
University of Chicago
mvannier@uchicago.edu
Outline
• Introduction
• Cone Beam CT in 2013
• New CT scanner technologies and
applications
• Imaging biomarkers & informatics
Introduction of Spiral CT
Cone Beam CT scanners
• Rotational fluoroscopic flat panel
neuroangiography
• Linac treatment verification - Radiotherapy
• 3D breast CT imaging
• Cardiovascular
C di
l CT
• Homeland Security CT
• Industrial
I d ti lN
Nondestructive
d t ti T
Testing
ti
• Dentomaxillofacial CT
Rotational Neuroangioraphy CBCT
Radiotherapy CBCT
3D Breast CBCT
Security CBCT
Industrial CBCT
BOEING 787 DREAMLINER
Past & Future:
Multiplex source-detector pairs
Multi-energy CNT Sources
• Carbon Nanotube (CNT) Sources
•
•
•
•
Spatial distribution
Temporal modulation
Multi-energy imaging
Bright & stable flux
Anode
e-beam
X-ray
Focus 2
Focus 1
2 m
G t
Gate
CNT
Vg
CNT
Carbon
Nano Tube
X-ray
Source
X-ray
Microtomography
CNT x-ray source
2011
2011
2011
2011
2010
2010-15
Medical Imaging
g g Market
Medical CT Market
• US – largest country market;
• Europe – largest regional market;
• Asia – largest market growth
•
•
•
•
CT is central to inpatient and outpatient care
Replacement lifecycle; added capacity
Obsolescence; cost of maintenance
D
Dose
reduction
d ti
• Four major vendors: GE, Siemens, Toshiba, Philips
Future
• A 2010 study on top trends in medical imaging
– budgets are tight
– reimbursement is shrinking
– competition is fierce
– need to deliver more, faster is endemic
• Increased R&D investment in hybrid modalities and
add-on technologies (including software
applications) continues to drive market forward
China Medical Imaging
13 yr old Female- Scanned w/256-slice CT
• 4.8 sec scan
• 2D AntiAnti-Scatter
detector grid improves
contrast resolution
• Smart Focal Spot for
artifact elimination
http://www.pedrad.org/associations/5364/ig/
Tissue Adaptation
• Automatically adapts to the tissue.
• Decrease noise in the soft tissue and increase the contrast in the lung.
original
processed
processed
original
o
g a
Pediatric - Liver
processed
original
Thin slice 0.6 mm
5 y.o. (different patients)
2.6 mGy
100 kVp, 65 mAs
Today
12.2 mGy
120 kVp, 120mAs
2009
Iterative
Reconstruction
(2011)
Significant dose saving
2011
Philips DoseWise tools
Optimizes the entire imaging chain to save dose
Iterative reconstruction
Many
innovations (e.g., tools)
iDose
for CT
dose reduction
are available,
available unrelated
to the reconstruction algorithm
Spectral Imaging
• Energy Discriminating Detectors & Contrast Agents
•
•
•
Multi-contrast perfusion
K-edge imaging with nanoparticles
Cellular & molecular imaging
Spectral
p
and Dual Energy
gy CT
Simple Analogy
Traditional
CT
Spectral
CT
Limited Spectral
CT
Dual Energy
CT
Yesterday & Today
Future
Future
Today
54
Dual-energy
gy Material Separation
p
HU of E1
H
Separation line
Iodine
X-ray Detector
signals
Iodine > Calcium
Calcium
Iodine < Calcium
H 2O
Atttenuation [[a.u.]
HU of E2
0
10
Calcium
-1
10
X-ray tube
X
t b
spectrum
-2
10
20
40
Iodine
solution
E1
60
E2
kv
80
100
120
140
Siemens
Spectral
p
vs. Dual Energy
gy CT
Techniques That are Possible on Commercial Systems
Dual Source
Dual kV Switch
Dual Spin
Not Spectral CT
57
Spectral
p
vs. Dual Energy
gy CT
Techniques That are Available on Research Systems
Dual-layer (“Double Decker”) Detector*
Photon Counting*
*Works-in-Progress: Pending commercial availability and regulatory clearance
58
Spectral
p
CT Scanners
Hybrid true color micro-CT system
Hybrid true color micro-CT system
Cylindrical
y
phantom with 7 materials
p
Soft tissue
Ca 12%
% /Water
Ca 6% / Water
Iodine / Water
Barium / Water
Gadolinium / Water
Gold / Water
Experimental
p
Results
MARS-CT
New Zealand
Energy
gy discrimination in CT
Multi-Energy CT
Photon Counting
Energy integrating detector
Variance (y)
V Flux
Vs.
Fl (x)
( )
Electronic
Noise
Pile-up
Technique: 80 kV, 5 mAs, 32*0.625 mm, 0.5 sec axial, 10 mA,
window width-1600 HU, window level-160 HU.
Photon Counting
g Prototype
yp
Clinical Study
Z-map images that are color coded according to tissue atomic number.
Efficient energy separation allows for true mono-energetic images.
Nature Physics
2, 258 - 261 (2006)
Phase retrieval and differential phase-contrast
phase contrast imaging
with low-brilliance X-ray sources
Franz Pfeiffer, et al.
Paul Scherrer Institut, 5232 Villigen PSI, Switzerland
Phase Contrast X-ray Imaging
Imaging of a rat paw.
Zhu P et al. PNAS 2010;107:13576-13581
©2010 by National Academy of Sciences
ESRF
European
Synchrotron
Research
Facility
145 million years old Cretaceous mammalian tooth from Cherves-de-Cognac
(Charente, France). This minute fossil tooth was imaged using sub-micrometre
resolution holotomography on ID19.
Computed tomography was feasible due to existence of a key
technology.
Mathematical Method:
IMAGE RECONSTRUCTION
FROM PROJECTIONS
Johann K. A. Radon, Ph.D. (1887 – 1956)
The Radon
t
transform.
f
In integral
geometry based
geometry,
on integration
over hyperplanes
— convert line
integrals to an
interior measure,
with application
pp
to tomography.
Born in Bohemia,
educated in Austria
and served as
mathematics
professor
f
in
i
Germany &
Austria.
Radon Transform (1917)
The n-dimensional Radon transform as a
J . Radon,
Radon “Uber
Uber die Bestimmung von Funktionen
durch ihre Integralwerte langs gewisser
Mannigfaltigkeiten,” Berichte Sachsische Acadamie der
Wissenschaften, Leipzig, Math.-Phys. Kl., vol. 69,
pp. 262-267, 1917.
Radon Transform
R(x,y)
R-1(s,ԕ)
Forward
Inverse
Image reconstructed
from projections
Shepp Logan CT
Shepp-Logan
phantom object
Sinogram
Picture of the measurements
that CT scanner acquires… “Raw data”
Cone Beam CT
Reconstruction
from Projections
FDK Image
g Reconstruction
• Feldkamp-Kress-Davis (also referred to as FDK)
algorithm
– [Feldkamp L A, Davis L C and Kress J W (1984)
Practical cone-beam algorithm, J Opt Soc Am,
A6, 612-619]
• A filtered back projection technique
– for the reconstruction of slices from projections
t k with
taken
ith cone b
beam geometry
t
In 1984, Feldkamp started a collaboration with
investigators at Henry Ford Hospital….
The direct examination of three-dimensional
three dimensional bone
architecture in vitro by computed tomography
Lee A. Feldkamp Ph.D., Steven A. Goldstein Ph.D., et al.
Journal of Bone and Mineral Research
Volume 4, Issue 1, pages 3–11, February 1989
Orthopedic Surgery
University of Michigan Medical Center
Open Source Cone Beam
CT Reconstruction
OSCAR – Open
p Source CT Reconstruction
• http://www.cs.toronto.edu/
http://www cs toronto edu/~nrezvani/OSCaR
nrezvani/OSCaR.html
html
• An Open-Source Cone-Beam CT Reconstruction Tool for Imaging
Research
• This work was supported by the AAPM Imaging Research
Subcommittee, MITACS and Princess Margaret Hospital.
Nargol Rezvani
Department of Computer Science,
U i
University
it off T
Toronto
t
Cone Beam CT Artifacts
Shepp Logan Phantom
2D
• Used for axial CT
reconstruction algorithm
testing
3D
• Generalization for whole
head facsimile, used to test
cone beam
b
algorithms
l ith
Cone Beam CT Artifacts
Original
2D A
Axial
i l
CBCT
Original
CBCT
2D C
Coronall
Depends
p
on algorithm
g
& fan angle
g
A=FDK
B=RB
C=SVF
D=SART
Cone Angle
Algorithm
International Journal of Biomedical Imaging
Volume 2006, Article ID 80421, Pages 1–8
Where did iterative image
reconstruction come from?
Image-based noise reduction
techniques
Overcoming the limitations of conventional reconstruction
Prior to introduction of IR techniques, targeted at overcoming some of
th lilimitations
the
it ti
off conventional
ti
l reconstruction
t ti
Addresses to some extent the statistical random noise.
Severely limited in being able to address photon starvation artifacts
such as streaks and image bias.
Conventional reconstruction
Image-based correction
Iterative (raw & image)
Iterative Reconstruction: How it works
Projection space
Image space
Optimizing image quality &
artifact prevention
Model based noise removal &
resolution improvement
Data dependant noise and structural models
used iteratively to eliminate the quantum image
noise while preserving the underlying edges
associated with changes in the anatomic
structure.
Noise power spectrum maintained through
dynamic frequency noise removal
removal.
Data
variation
analysis
Model
selection
Multifrequency
Model
Based
noise
removal
Structure
(Anatomy)
Model
Acquisition
Noise Model
Update
projections
p
j
n
Model y
Optimized
?
Noise
N
optim
mization
•
• Each projection examined for points likely to
result from noisy measurements
• Iterative diffusion process where noisy data and
edges are differentiated - noisy data is penalized
and edges are preserved
•
• Prevents low signal streaks and bias errors.
Images
Iterative Reconstruction: How it works
Projection space
update
Projection
space
original
i i l
RAW data
Projection
Signal
Image domain
update
Image Noise
(in carotid)
Conventional Reconstruction: How it works
Projection space
update
Projection
space
original
i i l
RAW data
Collect
data
((projections
j
are
the data)
Projection
Signal
Image domain
update
Image
Requires ONE
pass through
the data (fast!)
Modify projection
(put it through a filter)
Add the
modified
projection into
the image
Any No
more
data ?
Yes
D
O
N
E
Iterative Reconstruction: How it works
Projection space
update
Projection
Signal
Image domain
update
Image Noise
(in carotid)
Projection
space
original
i i l
RAW data
Collect data
( j
(projections
are
the data)
Modify projection
(put it through a filter)
Yes
Add the
modified
projection into
the image
Any No
more
Data ?
Iterative reconstruction: How it works
Projection space
update
Projection
Signal
Projection
space
original
RAW data
Variation analysis of projection data
1st Update
of RAW/
projection
data
Image domain
update
Image Noise
(in carotid)
Iterative reconstruction: How it works
Projection space
update
Projection
Signal
Image domain
update
Image Noise
(in carotid)
Projection
space
original
RAW data
Variation analysis of projection data
nth Update
of RAW/
projection
data
Courtesy: Cleveland Clinic, USA
Iterative reconstruction: How it
works
k Projection space Projection
iDose
Image domain
4
process
update
Signal
update
Image Noise
(in carotid)
Projection
space
original
RAW data
Parameter optimization and noise modeling
1st Update
of image
(subtraction of
noise while
validating against
structure model)
Courtesy: Cleveland Clinic, USA
Iterative reconstruction: How it
works
k Projection space Projection
iDose
Image domain
4
process
Projection
space
original
RAW data
update
Signal
update
Requires
R
i
MANY passes
through the
Analysis of model update
data (slower!)
nthh Update
of image
(subtraction of
noise while
validating against
structure model)
Note: The total number of iterations is greater than demonstrated in the simplified schematic above
Image Noise
(in carotid)
Iterative
We are accustomed to a certain type of noise in images.
Metal Artifact Reduction
Where did iterative image
reconstruction come from?
Where did iterative image
reconstruction come from?
Where did iterative image
reconstruction come from?
Where did iterative image
reconstruction come from?
Where did iterative image
reconstruction come from?
Where did iterative image
reconstruction come from?
IC Technology
Gordon Moore
•Gordon Moore worked for Fairchild Semiconductors
•He noticed a trend in IC manufacture
•Every 2 years the number of components on an area of
silicon doubled*
doubled
•He published this work in 1965 – known as Moore’s Law
•His predictions were for 10 years into the future
•His work predicted personal computers and fast
telecommunication
te
eco
u cat o networks
et o s
* Sources vary regarding time period
Graph of Moore’s
Moore s Law
What can you do with
a Billion transistors?
Answer:
Iterative CT
Reconstruction
!!
Spatial resolution
improvement: Body
Conventional CT
Iterative Reconstruction
Ultra low-dose acquisitions
Conventional
CT
Conventional
CT
Iterative
Reconstruction
Full Dose (188mAs)
Ultra Low-Dose (14mAs)
Ultra Low-Dose (14mAs)
5.5mSv
0.4mSv
0.4mSv
Extending low-dose for Bariatric Imaging
120kV, 360 mAs, 5 mSv, Prospective Gated Cardiac CTA (Step & Shoot Cardiac)
Conventional
Iterative
80 kVp, 200 mAs
Conventional
Iterative
3 mSv – minimal dose
~ 1 year of normal background radiation
18 mo. old female
with heart murmur
Iterative Reconstruction
1. Iterative reconstruction allows reduced
dose and improved image quality
2 There
2.
Th
are many ‘it
‘iterative’
ti ’ methods:
th d
– Operating in projection space -> reduced
photon
h t starvation
t
ti streaks
t k
– Preserving noise power spectrum ->
familiar ‘image
image texture
texture’
L1
Single pixel camera
(compressed sensing)
Geometric Truncation
Current Architectural Limitation
Detector
Array
Source
Array
Spiral
p
Cone-beam CT
Wang, G, Lin, TH, Cheng PC, Shinozaki DM, Kim HG: Proc. SPIE 1556:99-112, July 1991
Wang G,
G Lin TH,
TH Cheng PC
PC, Shinozaki DM: IEEE Trans
Trans. on Med
Med. Imaging 12:486
12:486-496,
496 1993
Zhao, J, Jin YN, Lv Y, Wang G: IEEE Trans. Med. Imaging 28:384-393, 2008
Lv Y, Katsevich A, Zhao J, Yu HY, Wang G: IEEE Trans. Med. Imaging 29:756-770, 2010
Professor
Mathematics Department
University of Central Florida (UCF)
Interior Tomography
X-rays
http://arxiv.org/abs/1304.7823
50 cm FOV - Thorax
15 cm FOV - Heart
Imaging the Vulnerable Plaque
Temporal Resolution: 83ms → 50ms
Spatial Resolution:
1mm → 0.4mm
Contrast Resolution: Spectral imaging & novel agents
Radiation Dose: Functional, interventional, pediatric
Target:
Heart
Reagent:
Nanoparticle
Technology:
Scanner &
Image Analysis
Future – Omni-tomography
g p y
PET Ring
Magnet
Slip Ring
Motor
Patient Table
Scanner
Stage
40cm
References:
G. Wang,
g F. Liu, F. Liu, G. Cao, H. Gao, M. W. Vannier, Top-level
p
design
g of the first CT-MRI
scanner, 2013. Paper accepted at the Int'l. Meeting on Fully Three-Dimensional Image
Reconstruction in Radiology and Nuclear Med., Lake Tahoe, CA, 17 June 2013.
IEEE Spectrum
http://spectrum ieee org/biomedical/imaging/path found to a combined mri and ct scanner
http://spectrum.ieee.org/biomedical/imaging/path-found-to-a-combined-mri-and-ct-scanner
SPIE News - http://spie.org/x94063.xml?pf=true&ArticleID=x94063
Design proposed for a combined MRI/computed-tomography scanner
Ge Wang, Feng Liu, Fenglin Liu, Guohua Cao, Hao Gao and Michael W. Vannier.
First CT
CT--MRI System Design
References:
Imaging Biomarkers
>20,000 Citations
In MedLine
•Definition
D fi iti
•Biological
Biological Marker (Biomarker) –
characteristic that is objectively
measured and evaluated as an indicator
of normal biologic processes
processes, pathogenic
processes, or pharmacologic responses
to a therapeutic intervention
intervention.
•Source: Biomarker Definitions Working Group - 1998
•Definition
•Clinical Endpoint - A characteristic or
variable that reflects how a patient feels
feels,
functions or survives.
•Surrogate
Surrogate Endpoint - Biomarker intended
to substitute for a clinical endpoint. A
surrogate
g
endpoint
p
is expected
p
to p
predict
clinical benefit (or harm, or lack of benefit or
harm) based on epidemiologic, therapeutic,
pathophysiologic or other scientific
evidence.
•Source: Biomarkers Definition Working Group -1998
USES
USES OF BIOLOGICAL MARKERS
CLINICAL TRIALS
Stratifying study populations
Conducting interim analysis of efficacy/safety
 Applied toward regulatory approval
CLINICAL PRACTICE
Establish diagnosis, prognosis
Monitor treatment response
Use
U as prognostic
ti measure
Imaging
Biomarker
Ontology
Top Level Classes
Imaging
Bi
Biomarker
k
Ontology
gy
http://biomarkers.org
From MGH – 2013:
Many biomarkers, few real uses – low adoption
•Biomarker Validity
•“Although validation, qualification, or evaluation has been used
interchangeably in the literature, the distinction should be made to
properly describe the particular phase the biomarker is transitioning
through in the drug development process.”
•A known valid biomarker is defined as -
•‘‘a
ab
biomarker
o a e that
a is
s measured
easu ed in a
an a
analytic
a y c test
es sys
system
e with welle
established performance characteristics and for which there is
widespread agreement in the medical or scientific community about
the physiologic, toxicologic, pharmacologic, or clinical significance of
the test results.’’
AAPS Journal. 2007; 9(1): E105E108. DOI: 10.1208/aapsj0901010
Biomarker Qualification Pilot
Process at the US Food and Drug
Administration
Federico Goodsaid1 and Felix Frueh1
1Genomics
Group, Office of Clinical
Pharmacology, Office of Translational Science,
Center for Drug Evaluation and Research, US
Food and Drug
g Administration, 10903 New
Hampshire Avenue, Building 21, Room 3663,
Silver Spring, MD 20903-0002
How Does an
Exploratory Biomarker
Become Probable or
Known Valid?
Conclusion
• Cone beam CT technology is used in a wide
variety of applications
• Improvements in x-ray sources, detectors and
reconstruction methods have been invented
• Technology developments promise to deliver
new capabilities for future CT scanners
• Most promising among these are dual
energy/spectral CT, iterative reconstruction,
compressed sensing and CNT x-ray sources
with interior reconstruction to realize
omnitomography (e
(e.g.,
g CT-MRI
CT MRI, …))
Acknowledgments
•
•
•
•
Ge Wang, Rensselaer Polytechnic Institute
Philips Medical Systems
Siemens AG
General Electric Corp.