A Critique of the Proposed National Education Policy Reform
MATLAB-based semi automated method for determining animal bone density from CT images
1. MATLAB-based Semi-Automated Method
for Determining Animal Bone Density
from Computed Tomography (CT) Images
Michael C. Oliveira
Department of Bioengineering, University of California Riverside
BPBE 510 – Introduction to Medical Imaging
Feb 24, 2011
2. Presentation at a Glance
•
•
•
•
•
Why do we care about bone density?
What is Computed Tomography?
Image Acquisition and Experimental Setup
MATLAB-based Method for Measurements
CT Images + Density Measurements of Cow,
Pig, Fish and Chicken
• Conclusions and Potential Improvements
3. Why do we care about Bone Density?
• Osteoporosis: the thinning of bone tissue and loss of
bone density over time
– Bone density measurements can help diagnose or
determine if you are at risk for Osteoporosis
– No symptoms in the early stages of the disease
• Tough to catch and diagnose early for preventative treatments
– Late stage symptoms:
• Bone pain
• Fracture without injury
• Loss of height
• Lower back pain – spinal fractures
• Neck pain – spinal fractures
• Stooped posture
– 1 in 5 women over the age of 50 have the disease
• Currently, Dual Energy X-Ray Absorptiometry (DEXA) is
the most widely used clinical test for measuring bone
density
Info from NIH: http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0001400
5. Computed Tomography (CT)
• Source: Hard X-Rays
• Detectors: Scintillators
• Image Reconstruction Algorithms:
– Simple Back-Projection
– Filtered Back-Projection
– And more…!
• Hounsfield Units (HU): normalized value of the X-Ray Attenuation
Coefficient.
– Air = -1000
– Water = 0
– Bone can be up to +3000
• Attenuation Coefficient (μ): quantity that characterizes how a
material affects EM radiation (units length-1)
6. Source-Detector Geometry
Typical CT Scanner
http://www.columbusimaging.com/Brilliance_CT_3.jpg
Resulting CT Image
Image Processing
Algorithms (from
Detector data)
http://radiographics.rsna.org/content/22/4/949/F7.medium.gif
Colormap for CT Images:
Black: low X-Ray Attenuation, lower signal
intensity
White: high X-Ray Attenuation, higher signal
intensity
Signal Intensity
~ Hounsfield Units and
Attenuation Coefficient
8. Image Acquisition and Experimental Setup
• Four animal bones
from Ralph’s
–
–
–
–
Cow ribs
Pork ribs
Fish skeleton
Chicken leg
• Image Acquisition:
– X-Ray Source: 75 kVp
– Tube Current: 1 mA
– Exposure Time: 175 ms
• Scan time: ~10 mins/data
set
• Data sets:
– N = 512 images
– Matrix size: 512 x 512
– Intensity encoded using
unsigned 16-bit integers
[0-(216-1)]
• Images used in analysis:
–
–
–
–
Cow: 340 – 365
Pig: 157 – 182
Fish: 127 – 141
Chicken: 360 – 385
9. MATLAB-based Analysis
Read Images into
MATLAB
Select Images for
ROI selection
Manually select
ROIs from Images
Hounsfield units
converted to
Atten. Coeff.
ROIs scaled to
Hounsfield Units
ROI mask applied
to original images
Atten. Coeff
converted to
Mass Atten. Coeff
Density solved for
each pixel in ROI
Average of all
densities for all
pixels in ROIs
10. ROIs scaled to
Hounsfield Units
Hounsfield units
converted to
Atten. Coeff.
Atten. Coeff
converted to
Mass Atten. Coeff
Density solved for
each pixel in ROI
• Scale the Pixel Intensities to the Hounsfield Scale
(rscValHU min rscValHU )
(max rscValHU min rscValHU )
(CTimage ( x, y, z ) min CTimage )
(max CTimage min CTimage )
CT Scale: [0, 65535]
1HU Scale: [-1000, 3000]
• Conversion from Hounsfield Units to Attenuation Coefficient
HU
1000
pixel
water
2μ
water=
0.1893 cm2/g @ 75 keV
water
• Solve for Density using the relationship between Attenuation
Coefficient and Mass Attenuation Coefficient
pixel
mass,cort.bone
2μ
mass,cort.bone=
0.2526 cm2/g @ 75 keV
pixel
1 Bushberg, Jerrold T. "Computed Tomography" The Essential Physics of Medical Imaging. Philadelphia: Lippincott
Williams & Wilkins, 2002.
2 NIST Physical Measurements Laboratory, http://physics.nist.gov/PhysRefData/XrayMassCoef/ComTab/bone.html
14. Density Measurements
Animal Bone Densities
Bone Density (g/cm3)
3.5
3
Cow
Pig
2.5
Fish
2
Chicken
1.5
1
Animal
1Cow
2.1-2.2
2.547 +/- 0.4844
2.0-2.1
1.910 +/- 0.4729
ND
2.526 +/- 0.3747
1Chicken
Animals
Calc Mean BD*
Fish
0
Ref Mean
BD*
1Pig
0.5
2.1-2.2
2.017 +/- 0.7940
*Density in g cm-3
1 Aerssens et al. “Interspecies Differences in Bone Composition, Density and Quality: Potential Implications for in Vivo
Bone Research.” Endocrinology. 139(2): 663-670. (1998)
15. Conclusions
• Successfully scanned and acquired images
• Wrote a semi-automated software package in
MATLAB for determining bone density
• Results from the software are fairly accurate
compared to literature values
16. Potential Improvements
• Image acquisition geometry
– Set up specimen to avoid the rings or artifact
• OR find robust way to filter out artifacts
– Contributes to improving automation
• Improve the repeatability and throughput
– Improve the automation
• Move towards fully automated processing instead of manual ROI
selection
– Improve processing speed
• Offload some computation to Graphics Processing Units (GPUs)
using Jacket to decrease total processing time
• Use an object with known attributes (density) for
calibration
– Should increase accuracy when converting pixel intensity
to HU
17. Acknowledgments
• Biophysics and Bioengineering, Loma Linda
Univ.
– Non-Invasive Imaging Lab
• Bioengineering, UC Riverside
• Funding Source
– Personal CHASE Acct. Num: #XXXXXXXXX
19. Convert kVp to X-Ray Energy
• Conversation of Energy
• Potential Energy = Kinetic Energy = X-Ray Energy
U
V e
KE
EX
1 2
mv
2
Ray
hc
V = 75 kVp (75,000V)
e = elementary charge = 1.602 x 10-19 C
m = 9.109 x 10-31 kg
h = Planck’s constant = 4.135 x 10-15 eV s
c = Speed of light = 3 x 108 m/s
• EX-Ray = 74.9 keV
• λ = 1.65 x 10-11 m (Hard X-Rays)
20. Linear Interpolation of Mass Attenuation Coefficients
Mass Attenuation Coefficient (cm2/g)
Mass Attenuation Coefficient vs. X-Ray
Energy
1.40
1.20
1.00
0.80
0.60
Cortical Bone
0.40
Water
0.20
0.00
0.00
20.00
40.00
60.00
80.00
100.00
X-Ray Energy (keV)
Reproduced from data at 2 NIST Physical Measurements
Laboratory, http://physics.nist.gov/PhysRefData/XrayMassCoef/tab4.ht
ml
inputPt - lowerBound
InterpValue - lowerBoundVal
=
upperBound - lowerBound upperBoundVal - lowerBoundVal
μwater= 0.1893 cm2/g
μmass,cort.bone= 0.2526 cm2/g