AAPM Scientific Meeting
Imaging Symposium
Patient Dose in CT: Calculating
Patient Specific Doses in CT
(Joint with Education)
Michael McNitt-Gray, PhD, DABR, FAAPM; UCLA
Peter Caracappa, PhD, CHP; RPI
Ehsan Samei, PhD, DABR, FAAPM, Duke
Acknowledgements
• Funded by NIBIB grant R01EB004898
• Recipient of Research Grant Support
from Siemens Medical Solutions
Patient Dose in CT:
Calculating Patient Specific Doses in CT
1. Limitations of current metrics (e.g. CTDIvol) and methods in
estimating patient dose (McNitt-Gray)
2. Methods to more accurately estimate dose that take into
account scanner, exam and patient factors (McNitt-Gray)
3. The role of Computational Phantoms, Monte Carlo
simulation in developing patient-specific dose estimates
(Caracappa)
4. Estimate patient-specific organ doses, effective doses, and
radiation risk for comparison and optimization purposes
(Samei)
Patient Dose in CT:
Calculating Patient Specific Doses in CT
Questions to keep in mind:
–
–
Should we calculate patient specific dose?
How accurately do we need to calculate this?
•
–
Does it vary from purpose to purpose (fetal dose,
meet legal requirements, etc.)
Do we need to calculate for EACH patient? Or
just for a “class” of patients (Large adult male)
•
•
Should we do online realtime Monte Carlo for each patient?
or precalculate doses somehow?
1
Background
Background
• NCRP 160 and (Mettler et al, Health Physics, Nov 2008)
Estimated US averages
Estimated Average Annual
Radiation Dose
(whole body eff. dose in mSv)
From Medical Radiation
From CT
1980
3.6 mSv/yr
2006
6.0 mSv/yr
• CT procedures
– Estimate 18.3 million in 1993
– Estimate 62.0 million in 2006
– 10% annual growth
• Slightly higher since introduction of MDCT (1994-1998)
– Could be over 100 million by now
0.54 mSv/yr
3.0 mSv/yr
15% of total
50% of total
---
1.5 mSv/yr
25% of total
Current Dose Metrics
 Many Organizations suggesting that CT dose be
tracked (NCI, IAEA, ACR, FDA, etc.)
 California SB 1237 requires reporting on CT dose
by July 1st, 2012 , one of the following:
 “The computed tomography index volume (CTDIvol) and dose
length product (DLP), as defined by the IEC and recognized by
FDA; The dose unit as recommended by the American
Association of Physicists in Medicine”.
Current Dose Metrics
 What is available to be recorded now?
 CTDIvol and DLP
• What could we do in the future?
• Organ Dose
– Could be tracked across scans and across time
– Accumulated organ dose over time
2
Current Dose Metrics
CTDIvol and DLP
 What is available to be recorded now?
• CTDIvol reported on the scanner
 CTDIvol and DLP
• Is Dose to one of two phantoms
• What could we do in the future?
• Organ Dose
– Could be tracked across scans and across time
– Accumulated organ dose over time
• Is NOT dose to the patient
• Does not tell you whether scan was done “correctly” or
“Alara” without other information (such as body region
or patient size)
• MAY be used as an index to patient dose with some
additional information
CTDI and Patient Dose :
They Are Not the Same Thing
• McCollough et al, Radiology, May 2011 ; 259:311–316
Scenario 1: No adjustment in
technical factors for patient size
100 mAs
100 mAs
• CTDI DOES REPRESENT:
– A measure of scanner output (with limitations
being addressed by TG 111 and TG 200)
– Well defined and highly reproducible across CTs
• CTDI DOES NOT REPRESENT
– Patient dose – does not take into account patient
size, shape, composition, scan length
– Dose from scans with no table motion (perfusion)
32 cm phantom
CTDIvol = 20 mGy
32 cm phantom
CTDIvol = 20 mGy
The CTDIvol (dose to phantom) for these two would be the same
3
Scenario 2: Adjustment in technical
factors for patient size
50 mAs
32 cm phantom
CTDIvol = 10 mGy
100 mAs
32 cm phantom
Did Patient Dose Really Increase ?
For same tech. factors, smaller patient absorbs more dose
– Scenario 1: CTDI is same but smaller patient’s dose
is higher
– Scenario 2: CTDI is smaller for smaller patient, but
patient dose is closer to equal for both.
CTDIvol = 20 mGy
The CTDIvol (dose to phantom) indicates larger patient received 2X dose
AAPM TG 204
Size Specific Dose Estimates
Based on Both
Simulations and
Measured Data
CTDIvol
• UNDER estimates dose for small patients
(have to multiply by > 1)
• OVER estimates dose for large patients
• (have to multiply by < 1)
4
CTDIvol
• Not patient Dose
• By itself can be misleading
• CTDIvol should be recorded with:
– Description of phantom size (clarify 16 or 32 cm
diameter)
– Description of patient size (lat. Width, perimeter,
height/weight, BMI)
– Description of anatomic region
Monte Carlo Simulation Methods for
Estimating Radiation Dose
• Monte Carlo methods
– Used in CT for some time
• NRPB report 250 (1990)
• GSF (Zankl)
Tracking/Reporting Dose?
• What should we record/report?
• What do we tell patient?
• What do we tell referring physician?
• CTDIvol? DLP?
• Total CTDIvol? Total DLP?
• Calculate Effective Dose from Total DLP*k?
Background
• These early reports used:
– Detailed Models of Single Detector, Axial Scanners
– Idealized (Nominal) collimation
– Standard Man Phantom
• MIRD V (geometric model) 
• Eva, Adam
5
Monte Carlo for CT Dose - Details
• Monte Carlo Packages
– MCNP (Los Alamos)
– EGS
• Model Transport of Photons from modified (CT) source
• Probabilistic interactions of photons with Tissues
– Photoelectric, Compton Scatter, Coherent Scatter
• Tissues need detailed descriptions
Background
• These form the basis for:
–
–
–
–
CT Dose computer program
CT Expo
ImPACT dose calculator
k factor approach (Effective dose = k* DLP),
which was derived from NRPB simulated data
– Density
– Chemical composition (e.g. from NIST web site)
Current Approaches
Modeling the CT scanner
• Spectra
– Function of beam energy
• Geometry
– Focal spot to isocenter, fan angle
• Beam Collimation
– Nominal or actual
• Filtration
– Bowtie filter (typically proprietary)
– Other add’l filtration (also proprietary)
• Tube Current Modulation Scheme
Photon Fluence Spectra
3.000E+11
Photon Fluence
2.000E+11
80 kVp Spectra
1.500E+11
125 kVp Spectra
150 kVp
1.000E+11
5.000E+10
0.000E+00
0
50
100
150
200
Energy in keV
128 mm
in air at iso
Normalized
1.100
Dose
1.250
1.000
0.900
1.000
0.800
Normalized Dose
Model Scanner (e.g MDCT) in detail
Model Patient (Geometric, Voxelized)
Simulate Scan
Tally Organ Dose
relative dose
•
•
•
•
2.500E+11
0.700
0.600
0.500
0.400
0.300
0.750
128 mm in air at iso
0.500
0.250
0.200
0.100
0.000
0.000
0
10
20
30
40
40
50
60
7060
80
90
100 80 110
120
130 100140
150
160 120
distance in mm Distance (mm)
– x-y only, z-only, x-y-z, etc.
6
Long Axis Modulation
Modeling the CT scanner
Lung
Region
Breast
Tissue
Abdomen
600
500
Tube Current (mA)
• Source Path - dependent on scan parameters:
• Nominal collimation
• Pitch
• Start and Stop Locations (of the source)
Shoulder
Region
400
180 degrees
(LAT)
300
200
90 degrees
(AP)
100
0
0
50
100
150
200
250
300
Table Position (mm)
Validating the CT Scanner Model
• Benchmark MC Model against physical
measurements
– CTDI Phantoms
• Head and Body
• Simulate a tally in a pencil chamber
• Each kVp and beam collimation combination
• Measured vs. Simulated
– Aim for < 5% difference between Simulated
and Measured
Modeling the Patient
• Geometric
– e.g MIRD
– Standard man
– Often androgynous (male/female organs)
– Usually single size
• Size and age variations
– newborn, ages 1, 5, 10, and 15 years
– adult female, and adult male
– Including pregnant patient
7
Modeling the Patient
• All radiosensitive organs identified
– Location
– Size
– Composition and density
Modeling the Patient
• Voxelized Models
– Based on actual patient scans
– Identify radiosensitive organs –
usually manually
– Non-geometric
• Different age and gender
• Different sizes
Modeling the Patient
• GSF models (Petoussi-Henss N, Zankl M et al,
2002)
– Baby, Child, three adult females (shown), two adult
males, Visible Human
– All radiosensitive organs identified manually (ugh!)
Modeling (Parts of) the Patient
• Embryo/Fetus
• Breast
8
Mature Fetus:
7 weeks (embryo not visible)
36 weeks
Gest. Sac
Uterus
Gest. Sac
Uterus
Original Image
Contoured Image
Voxelized Model
Threshold Image
Voxelized Model
Late Gestation
Early Gestation
Original Image
Contoured Image
9
Simulating the Scan
Monte Carlo Methods and Patient Size
• Select Technical Parameters
–
–
–
–
–
Type of scan (helical, axial)
Beam energy
Collimation
Pitch
Tube Current/rotation time (or tube current modulation)
• Select Anatomic Region
– Head/Chest/Abdomen/Pelvis/etc.
• Translate this to:
– Start/stop location -> Source Path
Fetal Dose as a Function of Patient Perimeter
Angel et al, PMB Feb 2009
Normalized Fetal Dose (mGy/100mAs)
16
14
12
10
8
6
y = -0.12x + 23.11
R2 = 0.68
4
2
0
85
90
Angel et al Radiology 2008
95
100
105
110
115
Perimeter of Mother (cm)
120
125
Tube current versus x-axis location of the TCM schema for a
patient model with a perimeter of 125cm. Background is a
sagittal view of the patient.
10
Angel et al, PMB Feb 2009
Breast dose versus patient perimeter for all 30 patient models
in the fixed tube current simulations. Breast dose decreases
linearly with an increase in patient perimeter (R2=0.76).
Angel et al, PMB Feb 2009
Breast dose versus patient perimeter for all 30 patient models
in the TCM simulations. Breast dose increases linearly with an
increase in patient perimeter (R2=0.46).
Angel et al, PMB Feb 2009
Organ Dose Independent of Scanner
Dose Savings
Dose Increase
Percent dose reduction for the TCM simulations as compared
to the fixed tube current simulations. Dose reduction decreases
linearly with an increase in patient perimeter (R2=0.81).
11
Organ dose (in mGy/mAs) and effective dose (in
mSv/mAs) for GSF model Irene resulting from a whole
body scan with similar parameters for each scanner
Organ dose and effective dose normalized by measured
CTDIvol for GSF model Irene resulting from a whole
body scan.
Turner et al Med Phys 2010
Turner et al Med Phys 2010
Mean organ dose/CTDIvol across scanners
Normalized Organ Dose as function of Pt. Size
(Abdomen Scans for each Patient)
3.0
Future of Dosimetry?
Patient
Size info
2.5
Stomach
Baby
Liver
Adrenals
2.0
Gall Bladder
y = 3.780e-0.011x
R² = 0.970
Child
Kidney
1.5
Pancreas
Helga
Irene
Spleen
Expon. (Stomach)
Golem
Donna
1.0
Visible
Human
0.5
Size
Coefficients
Patient Organ Dose
•Accounting for patient size
•Accounting for scanner
•Accounting for anatomic region
Frank
CTDIvol
(or TG 111)
0.0
25
50
75
100
125
150
Patient Perimeter (cm)
Turner et al Med Phys 2011
12
Radiation Dose : Organ Dose
•
BEIR VII report (2005)
–
•
ICRP 103 (2007)
–
•
Risk based on radiation dose to organ, age,
gender, etc.
Calculates “effective dose” based on weighted
sum of organ dose
Use dose to radiosensitive organs as a basis
for estimating metrics that relate to risk
Summary - Estimating Organ Doses
• Demonstrate feasibility of NOT having to do
detailed analysis on each Patient
• Not quite ready for implementation
• A path to estimate organ doses that takes into
account:
–
–
–
–
Scanner
Acquisition parameters (including TCM)
Anatomic Region
Patient Size
Summary - Estimating Organ Doses
• Organ Doses are meaningful indicators of Dose
• More informative than CTDI, DLP, E alone
–
–
–
–
Take into account differences in scanner
Take into account differences in patient size
Take into account differences in body region
Take into account dose reduction methods (TCM)
• Will be a better indicator as to when we truly
reach sub mSv exam
Acknowledgements
• Funded by NIBIB grant R01EB004898
• Technical Support from:
– Siemens Medical Solutions
– GE Healthcare
– Toshiba Medical Systems
– Philips Healthcare
13