11/1/2012

Comparison of Patient-Specific & Reference-Phantom Methods for CT Dose Estimation in the Pediatric Population

Purpose  Compare the characteristics of a) age-based reference phantoms used with Monte Carlo simulation to estimate organ doses with b) patient-specific phantoms based on CT data sets.

 Compare the use of reference-phantom and patient-specific dose distribution maps to estimate organ doses.

DME Bardo, MD1, KA Feinstein, MD2, D Pettersson, MD1, J Wiegert, PhD3, JH Yanof, PhD4

 Describe how differences in organ dose distribution affect the estimation of effective dose.

2

1

Philips Research3, Philips Healthcare4

Content Organization Reference phantom & Patient specific phantoms

Standard CTDI and stylized phantoms

Patient Specific Voxelized models based on patient data sets

Methods for estimating organ dose

Standard From geometrically defined organs in stylized phantoms

Patient Specific Using segmentation of organs in patient-specific dose maps

Effective dose calculations

Standard Regression with DLP & E (k factors)

Patient Specific Weighted sum of dose map organ doses

Limitations of standard dose estimates Volume CTDI & DLP are based on PMMA cylindrical phantoms and are not intended to be estimates of patient dose. They do not account for individual patients body habitus, attenuation characteristics, or specific scanner dosimetry.

DLP conversion factors for Effective dose (reviewed later) – are based stylized phantoms with fixed geometry (modified ORNL set for newborn, 1, 5, and 10 YO as shown). [1,2,3]

PMMA CTDIvol phantoms

Morphology of standard reference phantoms used for CT dose estimation can differ greatly from the anatomy of an individual patient

Striking anatomical differences between reference and the patient can effect the estimation of dose by dosimetry simulation (Monte Carlo).

Background

Major factors that affect CT dose include size/diameter & tissue/material absorption Dose (mGy)

40 30 20 10

k factor – AAPM 95.6 Table 3

A standard reference (stylized, mathematical) phantom (ORNL, Cristy) is compared with CT images of a 5-6 month old patient.

10 mGy CtrLG

20 mGy, PrphryLG

40 mGy CtrSM

40 mGy Prphry,SM

The x-ray beam has more attenuation as it traverses a larger patient (yellow to red to blue) in comparison with a smaller patient (yellow to red). Thus, the average beam intensity and dose, as the tube rotates, tends to increase with decreasing size for the same scan parameters

1

11/1/2012

Background

Average patient dose tends to increase with decreasing patient Anthropomorphic CTDI size/diameter

“Child”

Infant Child Adult Voxelized (GSF)

Cylinders

The Size Specific Dose estimates (SSDE) [6] also show that absorbed dose increases with decreasing size. A regression model (above) relates dose to effective diameter. The model data was based on a range of dosimetry methods (measured and simulated), four sets of phantoms, and four scanner vendors. All phantom sets included pediatric sizes.

Monte Carlo simulations with patient-specific phantoms:

CTDI-Normalized Dose Maps

Standard reference vs. Patient-specific Phantoms and Dosimetry

Patient Data set

Standard reference phantom

Patient specific dose map

Voxelized Phantom

In contrast to a standard reference phantoms (left), a patient’s CT data set can be used to create a patient-specific (virtual) phantom (middle). In addition to patient-specific dosimetry, this enables and individualized dose maps (right).

Monte Carlo simulations with patient-specific phantoms: Individualized dose maps in mGy

Dose to 32 cm CTDI phantom

Average dose map value

15 y/o CTDIvol 32 = 6.3 mGy

13 day old CTDIvol=1.5 mGy

Dose maps can also be displayed in units of CTDI normalized absorbed dose. In this case, dose map pixels are divided by the simulated dose absorbed by the CTDI phantom. The basic trend of CTDI-normalized average dose increasing with decreasing patient size (relative to the CTDI phantom) tends to agree with the SSDE correction factors.

Monte Carlo simulations of patient-specific phantoms can be used to simulate new dose maps

15 year old CTDIvol = 6.3 mGy

A Monte Carlo tool is used to simulate the dose absorbed by the patient specific “virtual phantom” instead of the CTDI phantom. This results in a patient specific dose map (in units of mGy). This example shows that the scan parameters for a 13 day old resulted in less absorbed relative to a 15 year old.

Morphological differences

between patient-specific & standard reference phantoms for dosimetry simulation (Monte Carlo) CT data sets are used to form patient specific virtual phantom (previous slide). The size and shape of organs and tissues can have wide variation.

Organs are represented by patient generic, fixed (stylized) geometric shapes.

Virtual phantoms do not extend beyond reconstructed image volume.

Simulated dose maps

13 day old CTDIvol= 2.5 mGy

Anatomy extends caudal-cranially, enabling dose simulation and scatter beyond scan range.

15 year old CTDIvol = 6.3 mGy

The CTDI of 2.5 mGy for a 13 day old and 6.3 mGy for a 15 year old would yield approximately the same patient specific-dose map.

newborn 1 YO

5 YO

10 YO

Virtual Patient (“Voxelized) Phantoms

10 YO

5 YO

1 YO newborn

Oak Ridge National Lab Phantoms (Cristy)

2

11/1/2012

Background

Flowchart for generation of dose maps: CT data acquisition, creation of voxelized phantom, dose simulation CT image voxels were classified as 1 of 5 tissues based on attenuation

Acquisition of CT images

1

Background Patient-specific organ doses can be estimated with dose maps

Monte Carlo simulations were 3 image sets. performed on voxelized

2

3

5

Dose Maps displaying CTDIvol normalized absorbed dose.

4

Organ doses can be segmented from dose maps

Dose maps

show patient-specific variability in organ doses that cannot be shown in the Oak Ridge National Lab (ORNL) references phantoms age

Lung dose 1.78 (mGy/mGy)

6 days

13 days

1.96

26 days

2.43

Average value of pixels segmented in the organ are used for organ dose estimation (in mGy)

Organ segmentation

Dose Map: Select radiosensitive organ or tissue

In standard reference phantoms such as those used for the DLP conversion factors, organs are defined with fixed geometry using mathematical equations.

Standard Reference Phantoms Can be modified to include additional radiosensitive organs

2 months

2.42

Estimated lung dose (CTDI normalized) from patient specific dose maps varies from 1.78 to 2.43 mGy/mGy. CTDIvol normalized organ doses simulated in the ORNL infant phantom (right) (new born) would not have any variability.

Patient-specific voxelized phantoms have features not included in standard reference phantoms

NRPB reference phantom (center [3]) extended the ORNL phantom concept (lower right) to include gender-based organs. The NRPB phantoms were used by Jessen et al. (with IRCP 60 organ weighing factors) to determine widely used DLP conversion factors [5,11]

Comparison of patient-specific & reference phantom methods used for effective dose calculation Dose Map

Patient Data Set

CTDI, DLPscan

Effective Dose, Patient-Specific Phantom

compare

Monte Carlo Simulation (Scanner Specific)

ODi From Organ Segmentation

Effective Dose, Reference Phantom ED = k x DLPscan

Reference Phantom LEGEND:

Breast dose can be measured (green contours) in dose maps -- this tissue is represented in the NRCP phantoms (previous slide). 14 y/o female (left) and 15 y/o male (right) with approximately the same effective diameter (~27 mm). Bismuth shields (arrows) were used in both exams (not represented in standard phantoms). Average lung dose is higher for the female patient due to relatively larger lung parenchyma (lower beam attenuation).

Monte Carlo Simulation CTDI, DLP

Dose Map

ODi From Organ Compartments

k AAPM Report 96.5

CTDI = Computed Tomography Dose Index DLP = Dose Length Product ED = Effective Dose ODi = Organ Dose wi = tissue weighting factors, ICRP 60

3

11/1/2012

Background

Estimation of effective dose for organ weighting factors

ED = ∑ wi x Odi where OD

w

is the individual organ dose measured from non-patient specific mathematical phantom or patient specific dose map is organ/tissue weighting factor (ICRP 60 or 103) i = 13 (13 most radiosensitive organs)

The effective dose equivalent, therefore, represents a total body dose.

Relative Organ Dose Sensitivities, Wi Patient-specific phantoms Segmenting all the listed organs and tissues for each individual voxelized phantom can pose a challenge.

Relative Organ Sensitivities ICRP (Used by both Pt. Specific and Std. Ref. Phantoms) 0.25

ICRP 60

0.2

ICRP 103

0.15 0.1

Standard reference phantoms The NRPB phantoms do not include all organs and tissues listed. And they would need to be revised if the ICRP adds new organs to the list.

0.05 0

ICRP 60 and NRPB phantoms were used for DLP conversion coefficients (Jessen).

Effective dose using ICRP

Background

weighting factors cannot be patient-specific

Determination of DLP conversion factors Effective Dose

They are based on linear regression analysis of body-region specific effective dose (from the simulations, y-axis) and DLP (x-axis).

Chest newborn as weighted sum of organ doses

Each body and age specific DLP conversion factor (k factor in units of mSv mGy-1 cm-1) was determined by dosimetry simulations with varying DLP.

Chest 1 YO Chest 5 YO Chest 10 YO

Slope = k, for each age

DLP input to Dose Simulation

• •

The organ sensitivity (weighting) factors are based on population data from survivors of the atomic blast, where the sum of the weighting factors is one. A 0.12 value for lung tissue implies that the relative likelihood of developing lung cancer in the population of blast survivors, in comparison with other listed organs, is 12%.

Therefore, any estimation of effective dose that uses these population based weighting factors – patient-specific dose maps or DLP conversion k factors – cannot be patient-specific. Although effective dose is not patient specific, dose maps enable patient specific organ dose estimates (next slide) and these increase the relative patient (as well as scanner) specificity in comparison with DLP conversion factors.

In this method, effective dose is assumed to be linearly proportional to DLP, i.e., E = DLP x k . DLP is linearly proportional to irradiated scan length (includes helical over-ranging). For pediatrics, DLP is based on CTDI 16. Also, K-factors (i.e., DLP conversion) represent an average over scanner types and are not gender specific. The organ dose weighting factors were described on the previous slide.

partial irradiation, ICRP weighting factors Partial irradiation of an organ tends to decrease the organ dose estimate. This is because organ dose is defined as the average over the entire organ. An advantage for the reference phantoms is that the caudalcranial range is not limited to a reconstructed scan volume as with the voxelized phantom. This can help estimate absorbed dose to partially irradiated organs. Basing organ dose on only the fully irradiated voxels will tend of overestimate the estimated organ dose.

Summary comparison dose maps and standard reference phantoms

scan length

Organ doses

partial irradiation of liver

ICRP weighting factors are based on full-body irradiation. Tissues that have wide distribution throughout the body such as red bone marrow are almost certainly partially irradiated in a CT examination.

Dose Map Method

Standard Reference Phantom Method

Representation

Voxelized Phantom Based on Data Set

Four pre-defined geometric representation of organs

Morphology

Patient-specific

Not Patient Specific

Organs

Organs must be segmented.

Organs pre-defined mathematically

Caudal Cranial End-effects

Not modeled (easily)

Extends beyond scan length to model partial organ irradiation

Computation

Computed for each patient and each examination

Material Models

CT Numbers are mapped to ICRU 44

ICRP Publication 89

Effective Dose

Pt. specific organ dose can be used to estimate eff. dose

Generic organ doses are used to determine DLP conversion coefficients.

Can be pre-tabulated for set of examinations and stored for future use

4

11/1/2012

Estimation of CT dose is evolving …

Summary

10 cm CTDI phantom



Patient specific voxelized phantoms can represent complex, patient specific anatomy and materials that are not easily represented in standard reference phantom.



Organ doses estimated from patient specific dose maps ARE patient specific. Patient-specific dose maps demonstrate the variability of organ doses and highlight a key limitation of standard methods for estimating effective dose.



Use of more patient-specific methods to estimate organ and effective doses could lead to better metrics and reporting for CT dose management. Effective doses estimated from ICRP wt. factors and NOT patient specific, but EDose Maps is more patient- and scanner-specific than EDLP.

Dose map sequence (z-axis) based on Monte Carlo simulation with infant CT data set

References

Clinical Relevance 



Patient-specific doses estimated by applying dosimetry simulations to voxelized phantoms may have advantages when patient morphology significantly differs from the reference phantom. Quantitative evaluation of patient-specific dose maps are underway. This will lead to a better understanding how more accurate dose estimate methods will impact CT radiation dose management.

1.

Cristy M . Mathematical phantoms representing children of various ages for use in estimates of internal dose. Report no. ORNL/ NUREG/TM-367. Oak Ridge, Tenn: Oak Ridge National Laboratory, 1980 .

2.

Cristy M , Eckerman KF . Specifi c absorbed fractions of energy at various ages from internal photon sources. I. Methods. Report no. ORNL/TM-8381/V1. Oak Ridge, Tenn: Oak Ridge National Laboratory, 1987 .

3.

A Khursheed, Phd, M C Hillier, P C Shrimpton, Phd And B F Wall, Bsc, Influence of patient age on normalized effective doses calculated for CT examinations

4.

Maria Zankl , Handbook of Anatomical Models for Radiation Dosimetry Edited by Xie George Xu and Keith F Eckerman , 3] Taylor & Francis 2009

5.

American Association of Physicists in Medicine. The measurement, reporting and management of radiation dose in CT. Report 96. AAPM Task Group 23 of the Diagnostic Imaging Council CT Committee. College Park, MD. American Association of Physicists in Medicine, 2008.

6.

American Association of Physicists in Medicine. Size-specific dose estimates (SSDE) in pediatric and adult body CT examinations. Report 204. AAPM Task Group 204. College Park, MD. American Association of Physicists in Medicine, 2011.

7.

McCollough CH, et al., CT Dose Index and Patient Dose: They Are Not the Same Thing. Radiology: Volume 259:(2) 311-316.

8.

Morgan HT., Dose reduction for CT pediatric imaging, Pediatr Radiol. 2002 Oct;32(10):724-8; discussion 751-4. Epub 2002 Aug 29.,

9.

Adam C. Turner1 and Michael McNitt-Gray, The feasibility of patient size-corrected, scanner-independent organ dose estimates for abdominal CT exams, Med Phys. 2011 Feb;38(2):820-9.

10.

Boone JM, Strauss KJ, Cody DD, McCollough CH, McNitt-Gray MF, Toth TL, Goske MJ, Frush DP. Size-specific dose estimates (SSDE) in pediatric and adult body CT examination. Report No. 204. 2011

11.

Cynthia H. McCollough et al. How Effective Is Effective Dose as a Predictor of Radiation Risk?, AJR:194, April 2010

Limitations of CTDIvol

Appendix I

The CTDIvol reports scanner output based on a standard, fixed-sized phantom (32 cm for body), not patient-specific dose. Therefore, dose is over- and underestimated for patients significantly larger or smaller (respectively) than the phantom. (see AAPM report 201)

air adipose tissue lung tissue soft tissue CT image for 6 day old

Virtual patient phantom

cortical bone

A patient-specific (tomographic) virtual phantom (i.e., model) is created by “voxelizing” and automatically segmenting patients CT dataset. Each voxel is assigned one of five material types based on an a priori, global HU classification intervals (ICRU 44). These material types are also assigned mass density to compute absorbed dose. The resulting virtual patient phantoms are used for dose simulation (Monte Carlo) and the results are referred to as Dose Maps (next slide).

Estimated CTDIvol [mGy] (120 kV)

Creation of a Patient-specific Voxelized Models 15 10 5

Actual dose for 24 cm diameter patient

Reported dose Actual dose for 50 cm diameter patient

0 Patient diameter [cm]

24

32

50

5

11/1/2012

DLP and Effective Dose Dose Length Product

CTDI (mGy)

CTDIvol Equation or dose calculation method

CTDIvol is presently measured with 16 and 32 cm phantoms

(mGy *cm)

DLP =

Dose map reconstructions Effective Dose (mSv)

Dose maps show representations of:

Irradiated Scan Length x CTDIvol

Helical scan length: the reconstructed scan length plus helical over-ranging Axial scan length: the reconstructed scan length for one “axial shot” * number of axial “shots”. (The CTDIvol accounts for “overbeaming”).

k = conversion coefficient for the DLP method of estimation

Effective dose (ED) parameter shown here is also based on the plastic CTDI phantoms. It is a risk-related quantity used to indicate equivalent whole body exposure that includes DLP as well as other factors such as the radiation sensitivities of the various organs in the body, age, and gender.

Energy imparted map [Joules]

Absorbed dose map [mGy]

Absorbed dose map/CTDIvol [mGy / mGy]

Typical range: 10-5 J/pixel

Typical range: 0-20 mGy

Typical range: 0-2.5 mGy/mGy

Notes: 1) Effective dose using DLP conversion coefficients are estimated with averages over gender and age and therefore do not estimate risk for an individual patient. 2) Reference for ED are based on estimates for annual background radiation (3 mSv). 3) Another method to compute ED is based on the summation of organ dose estimates.

6