8/18/2011
Hybrid Calibration and Display of CT Images in the
Gram Scale
Can We Ever Give up the HU Unit?
Outline:
* Limitations of water phantom calibrations of HU in MDCT
* Some causes and challenges to making desired corrections
Ben Arnold, MPH, ScD
Image Analysis, Inc
* N-vivo tissue calibrations; specific for patient/scanner/exams
* Hybrid phantom/tissue calibrations; calcium/iodine targets
• Results with Hybrid calibrations and voxels in density units
Conflict-of-Interest Disclosure: Ben Arnold is the principal stockholder
and president of Image Analysis which develops and markets quantitative
CT products including the software here reported.
* A proposed density scale, “ the Gram scale”
“Could we, should we, ever give up the Hounsfield scale??”
BMD Phantom calibration for single slice CT scanner
Unreliability of CT Numbers
225
Phantom Densities mg/cc
200
Phantom Calibration
175
The Unreliability of CT Numbers as Absolute Values
Clifford Levi, Joel E. Gray, Edwin C. McCullough, Robert R. Hattery
1982, AJR; 139:443-47
150
125
100
y = 0.99x + 6.19
R² = 1.00
75
50
25
0
-25
25
75
125
175
225
CT Values HU
1
8/18/2011
200
Phantom Sample (mg/cc
Very Unreliable MDCT Numbers
“Water ranged from 1 to 15 HU; air from -962 to -990”
Groell R et al. Comp Med Image Graphics 24 (2); 53-8
MDCT vs. Single Slice Scanner
64 MDCT
150
Single slice
y = 0.84x + 22.93
R² = 1.00
y = 0.85x - 3.42
R² = 1.00
100
50
“CT number of adipose tissue decreased with distance from center”
All MDCT scanners ,,,,,,, variability in tissue attenuation ( liver tissue 63.4
HU, cyst −15.7 to 23.9 HU)
Bernard A et al 2007 Radiology, 242; 109-19
0
-30
20
70
120
170
220
ROI readings (HU)
Energy Dependent X-ray Imaging
Water phantom Calibration
Two Scanners
Two kVps
225
200
150
125
150
Phantom Densities mg/cc
Phantom Densities mg/cc
175
y = 0.990x + 6.192
R² = 0.999
125
y = 0.860x - 4.129
R² = 1
100
75
Siemens
50
Aquilion
* Idealized circular shape
* Homogeneous material
* Calibrated at center
GE120 kVp
y = 0.838x + 18.09
R² = 1
100
GE80 kVp
75
y = 0.599x + 14.83
R² = 1
50
120 kvp
25
25
80 kvp
0
CT Values HU
0
-25
Problems:
175
25
75
125
-25
175
25
75
125
CT Values HU
175
225
275
225
2
8/18/2011
55
Variability of CT Measured Blood Pool Density
6 Patients, 4 CT Scanners
Blood Pool HU
50
45
40
35
Hybrid and phantom calibrations are computed and vary with the
patient’s anatomy and x-ray attenuation along the z-axis
30
1
2
3
Patients
4
5
6
Automated Blood Pool Density and SD for Three Patient CT Scans
Judy, P. et al SPIE 2009
3
8/18/2011
Scattered Radiation with a 64 MDCT Scanner
Blood as In vivo Tissue Reference
21.5 cm Water phantom, without the bow-tie filter
SPRs after applying inverse software scatter corrections
“There was a small increase in whole blood specific gravity with increasing
hematocrit, but it was not statistically significant over the 40-56 hematocrit
range studied.”
Trudnowski RJ et al. Clin Chem 1974 ;20/5, 615-16
Normal variations of blood chemistry equate to approximately
2-4 HU
A Akbarzadeh et al. Biomed Imaging Interv J 2010; 6(1):e3
0.015
Difference in SG
Difference in Specific Gravity (25°C)of Whole Blood
before and after 1500 ml IV of Normal Plasma (8% solids)
0.01
0.005
0
1.045
1.050
SG of 8 Subjects
1.055
Phillips RA, The Rockefeller Institute for Medical Research, Dec 15,1945
4
8/18/2011
0.9
Development of Automated in vivo
Tissue References
• Full body calibrations
• Fully automated
• Universal standard
0.7
μ/ρ (cm2/g)
Requirements:
0.8
adipose
0.6
0.5
Blood
muscle
0.4
0.3
0.2
0.1
• Better than water phantoms
0
20
30
40
50
Energy (kev)
60
80
100
Automatic Computation of Tissue Measures
Highest Peak
HMF
HMM
STR
5
8/18/2011
Automatic Computation of Tissue Measures
Automatic Computation of Tissue Measures
Highest Peak
Highest Peak
HMM
HMF
HMM
HMF
HU Values vs. Temperature of Water
Body Water CT Number
37°C
Water Density Decreases with Increasing Temperature
- 0.45 HU/°C
Body - Room Temperature = 37° - 20°C = 17°C
HU = 0.45 x 17 = 7.6 HU
Falone, BG et al Med. Phys. 9(5), 1982
Mahnken AH et al, Int J Clin Pract, 65 (S. 171), 2011
6
8/18/2011
1.10
Soft Tissue Reference (STR)
1.10
Muscle/blood = 1.047
Tissue Specific Gravity (g/cc)
1.05
1.00
Body Water = 0.994
y = 0.001x + 1.002
R² = 1.000
0.95
Ref 1
Ref 3
y = 0.001x + 1.000
0.90
R² = 1.000
Ref 2
1.05
STR = -26 mg/cc
1.00
50/50 = 0.974
Body Water = 0.994
0.95
0.90
y = 0.001x + 1.000
R² = 1.000
Adipose lipid = 0.901
PHB
0.85
-120
0.85
-120
-100
-80
-60
-40
-20
0
20
40
60
-100
-80
-60
-40
-20
0
20
40
60
80
Tissue Densities (HU)
80
60
Hybrid Calibration
In-vivo Soft Tissue Reference (STR)
100
Auto Analysis of the Soft Tissue Histogram
60
HMm
40
HU
20
0
2.5
1.25
0.625
STR
-20
-40
-60
HMf
-80
-100
40
Tissue Density referenced to water (mg/cc)
80
20
0
750
800
850
900
950
1000
1050
1100
1150
-20
y = 0.414x - 405.7
y = 0.471x - 463.2
-40
y = 0.557x - 549.1
J80
-60
G80
G120
G80
y = 0.620x - 612.8
J120
J120
J80
G120
-80
Slice Thickness (mm)
-100
Fat/Muscle (HU)
7
8/18/2011
1.05
Tissue Specific Gravity g/cc
Soft Tissue Reference ( STR )
64 MDCT
1.00
single slice
0.95
0.90
0.85
-160
-140
-120
-100
-80
-60
-40
-20
0
20
40
60
80
100
An always in place calibration phantom pad contains reference calcium hydroxyappitte
samples incased within a CT couch pad was present in all scans.
Measured Tissue HU
A
No artifacts were observed
from the small samples
which would interfere with
other exams
B
The calibration pad length
of 125 cm extends to cover
the torso
N-Vivo™ Applications and Calibration Screen
Ben Arnold
Image Analysis, Inc
8
8/18/2011
Patient, CT Technique, and Scanner
Specific Dual Thresholds
The thresholds equations are:
HiThreshold = C1 x BP(z) + C2 x SD (z)+ C3
LoThreshold = C4 x BP(z) + C5 x SD (z)+ C6
Constants Cs, SD, and BP are computed with calibrated
voxels from the muscle/blood along the z-axis.
Automated Calibration in the Gram scale
Ben Arnold
Image Analysis, Inc
CaHA microsphere phantom
EBCT Agatston
GE64 2.5 mm
GE64 .62 mm Auto-calibrated
9
8/18/2011
Example Coronary calcium score
Agatston Score= 0
Auto/Calibrated Mass = 0.7 mg
10
8/18/2011
250
Single Slice Scanner
Standard QCT vs. Hybrid Calibration
250
64 MDCT vs. Single Slice No Hybrid Calibration
225
64 MDCT (mg/cc)
QCT-3D Plus mg/cc
200
y = 1.00x + 0.70
R² = 0.93
150
100
200
175
y = 1.01x + 25.00
R² = 0.81
150
125
74 Vertebra
38 vertebra
100
50
75
75
100
125
150
175
200
225
250
Single Slice (mg/cc)
0
0
50
100
150
200
250
Hybrid Calibration mg/cc
MDCT volume scans create high scatter conditions and poor correlation of BMD
between scanners with Standard phantom calibration
The Gram Scale
The Gram Scale
Why?
Air
Fat
Body
Water
Water Muscle
Bone
1. More consistent images for specific patients
2. Calibration without the use of a water phantom
3. Image voxels in density units for quantitative measurements
HU Scale
Density
-1000
-100
0
0.901
-1000
-100
0
55
3000
0.994
1
1.047
1.85
-6
0
47
1850
4. Potential for better standardization among CT scanners
Gram Scale
11
8/18/2011
Conclusions and Comments:
The Gram Scale
MDCT HU values are not reliable for many applications

HU Scale
-1000
0
3000
Automated Hybrid calibrations appear possible in background
STR

GU Scale
-1000
In vivo tissues may provide improved calib (blood/muscle, fat, air)
-26
A new density scale is proposed, “ The gram Scale”
0
1000
“ Could we, should we, consider converting from the HU scale?”
Density (mg/cc) = 1000 + GU
12