eDE: a metric of quality and dose for digital radiography

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Effective dose efficiency: an application-specific metric of quality and dose for digital radiography

This content has been downloaded from IOPscience. Please scroll down to see the full text. 2011 Phys. Med. Biol. 56 5099 (http://iopscience.iop.org/0031-9155/56/16/002) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 122.96.59.106 This content was downloaded on 17/10/2013 at 10:45

Please note that terms and conditions apply.

IOP PUBLISHING

PHYSICS IN MEDICINE AND BIOLOGY

Phys. Med. Biol. 56 (2011) 5099–5118

doi:10.1088/0031-9155/56/16/002

Effective dose efficiency: an application-specific metric of quality and dose for digital radiography Ehsan Samei1,2,3 , Nicole T Ranger1 , James T Dobbins III1,2 and Carl E Ravin1 1

Carl E Ravin Advanced Imaging Laboratories, Department of Radiology Medical Physics Graduate Program, Departments of Biomedical Engineering and Physics 3 Clinical Imaging Physics Group, Department of Electrical and Computer Engineering, Duke University, Durham, NC 27710, USA 2

E-mail: [email protected]

Received 10 March 2011, in final form 22 June 2011 Published 20 July 2011 Online at stacks.iop.org/PMB/56/5099 Abstract The detective quantum efficiency (DQE) and the effective DQE (eDQE) are relevant metrics of image quality for digital radiography detectors and systems, respectively. The current study further extends the eDQE methodology to technique optimization using a new metric of the effective dose efficiency (eDE), reflecting both the image quality as well as the effective dose (ED) attributes of the imaging system. Using phantoms representing pediatric, adult and large adult body habitus, image quality measurements were made at 80, 100, 120 and 140 kVp using the standard eDQE protocol and exposures. ED was computed using Monte Carlo methods. The eDE was then computed as a ratio of image quality to ED for each of the phantom/spectral conditions. The eDQE and eDE results showed the same trends across tube potential with 80 kVp yielding the highest values and 120 kVp yielding the lowest. The eDE results for the pediatric phantom were markedly lower than the results for the adult phantom at spatial frequencies lower than 1.2–1.7 mm−1, primarily due to a correspondingly higher value of ED per entrance exposure. The relative performance for the adult and large adult phantoms was generally comparable but affected by kVps. The eDE results for the large adult configuration were lower than the eDE results for the adult phantom, across all spatial frequencies (120 and 140 kVp) and at spatial frequencies greater than 1.0 mm−1 (80 and 100 kVp). Demonstrated for chest radiography, the eDE shows promise as an application-specific metric of imaging performance, reflective of body habitus

0031-9155/11/165099+20$33.00

© 2011 Institute of Physics and Engineering in Medicine

Printed in the UK

5099

5100

E Samei et al

and radiographic technique, with utility for radiography protocol assessment and optimization. (Some figures in this article are in colour only in the electronic version)

1. Introduction As digital radiography imaging system technology has matured, it is important to establish performance metrics that are both meaningful and reflective of actual system performance under clinical imaging conditions. The detective quantum efficiency (DQE) has been the standard method used to characterize the signal-to-noise transfer properties of these imaging systems (IEC 2003, Illers et al 2005, Samei et al 2006, Dobbins et al 2006, Ranger et al 2007). However, as a detector-centric performance metric, the DQE falls short of capturing the performance of the system under clinical conditions in that it does not reflect the effects of focal spot blurring, magnification, grid response and scatter which are commonplace in routine clinical imaging (Samei 2003, Boyce and Samei 2006). Several investigators have proposed theoretical frameworks for characterizing digital radiography and angiography systems using a generalized DQE that would incorporate most system performance aspects (Kyprianou et al 2004, 2005a, 2005b, Yadava et al 2005, Shaw et al 2000). We have previously proposed a conceptually similar, but empirically derived metric, the effective DQE (eDQE), designed to capture the overall performance of an imaging system in routine clinical chest radiography using an application-specific geometric phantom to simulate the patient (Samei et al 2008, 2009). Intuitively, the eDQE can be interpreted as ‘the actual output signal to noise ratio squared that an imaging system provides relative to the fluence of radiation incident on the patient, without penalizing the system for primary attenuation through the patient’ (Ranger et al 2009). Compared to the DQE metric, the eDQE more closely reflects the performance of digital chest radiography systems under clinical operating conditions. However, it does not reflect the dose or radiation risk to the patient associated with the imaging procedure. For that purpose others have proposed a doseto-information-conversion efficiency or similar figures of merit (FOM) based on the DQE normalized by incident air KERMA or absorbed dose (Tapiovaara 1993, Dobbins et al 2001, Ullman et al 2006, Lee et al 2007). Since the DQE is limited only to the detector performance, such a metric will still fall short of accurately reflecting the performance of the imaging system for the clinical application of interest (Ranger et al 2009, Ertan et al 2009). In the current study, we propose and measure a normalization of the eDQE by effective dose (ED) to derive an effective dose efficiency (eDE) metric for radiography. We evaluated the utility of eDE in the assessment and optimization of radiographic technique factors in digital chest radiography using geometric chest phantom configurations representing pediatric, average adult and large adult patient populations. 2. Methods 2.1. Theoretical framework As previously derived, the eDQE is an extension of the conventional DQE metric (supplemented with the use of a geometric phantom) given by (Samei et al 2008, 2009) eDQE(fm )=

eMTF(fm )2 · (1 − SF)2 , eNNPS(fm ) · TFnb · E0 · q

(1)

Effective dose efficiency

5101

where m is the magnification in the object plane of interest, fm is the magnification-corrected spatial frequency (f × m) in the object plane, effective modulation transfer function (eMTF) and eNNPS are respectively the effective MTF and effective NNPS measured at the surface of the phantom, SF is the measured scatter fraction, TFnb is the narrow beam transmission fraction through the phantom, and E0 is the air KERMA measurement extrapolated to the receptor plane. q is the squared ideal signal-to-noise ratio (SNR) estimated from spectral simulation models, using the known tube potential, estimated inherent filtration of the x-ray tube, and known thickness and composition of the geometric phantom. Because the eDQE does not reflect the ‘cost’ or detriment associated with increasing absorbed dose, used alone, it is of limited utility for clinical technique optimization, where the tradeoffs between image quality and dose must be assessed. However, the eDQE approach can incorporate ED to address this limitation by normalizing the effective noise equivalent quanta (eNEQ) by ED. Defined as effective dose efficiency, eDE can be computed from eDE(fm ) =

SNR2out eNEQ(fm ) SNR2in = , × ED ED SNR2in

(2)

where SNR2in and SNR2out are, respectively, the input and output squared SNR, and ED is the effective dose. By substituting eDQE for the ratio of SNR2out to SNR2in and expressing SNR2in in terms of the entrance air KERMA (ES), the x-ray source-to-surface distance (SSD) and x-ray source-to-image receptor distance (SID), the expression becomes   SSD 2 eNEQ(fm ) ES = eDQE(fm ) × q × × eDE(fm ) = . (3) ED ED SID The ED is computed from the weighted sum of absorbed dose to individual organs across all organs, using pre-determined published organ weighting factors (ICRP 1990, 2007). The quantity ED/ES in equation (3) can be computed from estimates of entrance air KERMA, published organ dose tables, and weighting factors for a reference standardized human body (Hart et al 1994a, Huda and Gkanatsios 1997, McCullough and Schueler 2000, Liu et al 2008). Alternatively, ED/ES can be obtained from Monte Carlo simulations where a specific body habitus can be modeled (Tapiovaara et al 1997, Hart et al 1994b). The eDE definition given above uses air KERMA as the normalizing factor to be consistent with the conventional definition of the DQE. However, an equally applicable alternative definition can be based on air KERMA–area product—a metric that is not explored in the current study. 2.2. Phantoms The framework described above for eDQE and eDE measurements requires the use of an application-specific phantom that emulates the scatter and magnification aspects of the imaging application. This study focused on the measurement of the eDQE and eDE for chest radiography. For that purpose, the investigation employed a pediatric and an adult geometric chest phantom designed by the Food and Drug Administration (FDA) for use in the Nationwide Evaluation of X-ray Trends (NEXTTM ) program (Conway et al 1984, Food and Drug Administration and Conference of Radiation Control Program Directors 2003). These phantoms were used in this study as they are designed and recognized in the United States to emulate the attenuation and scatter conditions encountered in normal chest radiography. They were composed of 25.4 cm × 25.4 cm slabs of acrylic and aluminum, separated by an air gap. The phantoms were imaged in three configurations (figure 1, table 1) to reflect a range of body habitus types encountered in digital chest radiography; two configurations exactly matched those targeted by the FDA for pediatric and adult imaging. A third phantom configuration,

5102

E Samei et al

(a)

(b)

(c)

Figure 1. Pediatric (a) and adult (b) geometric chest phantoms utilized by the FDA/CDRH Nationwide Evaluation of X-ray Trends (NEXTTM ) program and a composite of both phantoms (c) used to represent a large adult body habitus. Phantoms were comprised of aluminum and acrylic layers with an air gap to simulate the effects of lung. The pediatric phantom, 11.5 cm thick, comprising a total of 66.7 mm acrylic and 0.4 mm Al with a 4.8 cm air gap, was designed to be representative of the chest attenuation and thickness of an average 15 month old infant with a weight of 10.89 kg and a height of 78.5 cm. The adult phantom, 26.3 cm thick, comprising a total of 73.0 mm acrylic and 4.1 mm Al with a 19.0 cm air gap, is representative of a 30 year old adult male with a weight of 74.84 kg and a height of 172.2 cm. The large adult phantom, 32.7 cm thick, comprising a total of 130.2 mm acrylic and 4.3 mm Al with a 19.0 cm air gap, is intended to be representative of a 30 year old adult male with a weight of 136.1 kg and a height of 172.2 cm.

Effective dose efficiency

5103

Table 1. Geometric chest phantom attributes.

Representative demographic Phantom

Age

NEXTTM pediatric

15 months

NEXTTM adult

Large adult

Composition materials (mm)

Ht (cm)

Wt (kg)

78.5

10.89

66.7 0.4

30 years

172.2

74.84

30 years

172.2

73.0 4.1 54.0 130.2 4.3 54.0

136.1

Acryllic Aluminum Air gap Acryllic Aluminum Air gap Acryllic Aluminum Air gap

Chest thicknessa (cm) 11.5

26.3

32.7

a

Measured along the beam axis. Represents the distance from the posterior surface of the phantom adjacent to the detector cover plate to the anterior surface of the phantom facing the x-ray tube.

constructed to represent the upper range of normal adult chest size, consisted of the adult NEXTTM phantom with the addition of a 0.2 mm thick sheet of aluminum and a 57.2 mm thick slab of acrylic affixed to the exterior of the anterior (tube-side) surface of the phantom (figure 1, table 1). The spine–heart shadow insert and ionization detector support holder normally associated with these phantoms were not employed for this study. 2.3. Imaging system All imaging studies were performed using a digital radiographic imaging system (Revolution XQ/i, GE Healthcare, Milwaukee, WI) employing an aSi:CsI flat-panel detector with a 0.2 mm pitch and a stationary grid with a 13:1 grid ratio and a grid density of 78 lines per centimeter (Mitaya Manufacturing Co. Ltd, Tokyo, Japan). An x-ray SID of 180 cm was utilized throughout the study. The system’s default calibration tables, acquired without a grid as per the manufacturer’s protocol, were used. All subsequent images were acquired with the grid in place. Image data were obtained from the system after correction but prior to the application of image processing, i.e. the DICOM Presentation State was ‘For Processing’. 2.4. eDQE measurement 2.4.1. Air KERMA measurements. Imaging studies were performed at each of four spectral conditions using 80, 100, 120 and 140 kVp, with no additional filtration. At each spectral condition, a phototimed exposure was acquired for each phantom configuration (figure 1, table 1) to establish the fixed mAs target exposure condition for the noise power and related exposure estimates. Using the mAs setting closest to the phototimed mAs, the air KERMA was measured for each spectral condition and phantom configuration using a calibrated ionization detector (MDH Model 1015, 10X5-6 ionization chamber, Radcal, Monrovia, CA) placed at 104 cm from the x-ray tube and centered on the beam axis. Air KERMA measurements were extrapolated to the imaging receptor (i.e. E0) and phantom surface (i.e. ES) planes, respectively, using the inverse square law.

5104

E Samei et al

2.4.2. Transmission fraction. The narrow beam transmission fraction through the phantom, TF, was measured at each spectral condition from the ratio of the average of five exposures with the phantom present, to the average of five exposures without the phantom, using the same calibrated ionization chamber (MDH Model 1015 and 2025, 10X5-6 ionization chamber, Radcal, Monrovia, CA). To avoid scattered radiation, this measurement was obtained using a narrowly collimated beam and with the phantom centered on the beam axis and placed near the x-ray source at approximately one third of the distance between the source and image receptor. The ionization chamber was centered on the beam axis and positioned halfway between the image receptor and the posterior surface of the phantom. For air KERMA measurements without the phantom present, the phantom was removed without repositioning the ionization chamber. 2.4.3. Scatter fraction. The scatter fraction (SF) was measured at each spectral condition and for each phantom configuration using the beam stop technique of Floyd et al (1991). Each phantom configuration was in turn centered on the beam axis and positioned against the detector cover plate (figure 2) with a beam stop device affixed to the anterior (tube-side) surface. The beam stop device was comprised of a 6 mm thick sheet of polystyrene into which was embedded in a 14 × 16 array of 224 lead cylinders spaced 25 mm apart, each 6 mm in thickness and 3 mm in diameter. Three images of the device were captured at an air KERMA of 2E0. A 10 pixel × 10 pixel region of interest (ROI) was positioned over each beam stop and on either side of the beam stop. The SF was then computed from the ratio of attenuated mean count (behind the beam stop) to the average of the two mean background counts, averaged across 20 beam stops in the central portion of the image. 2.4.4. Input signal-to-noise. The number of quanta incident on the detector per unit area per unit air KERMA (q value) was estimated for each phantom configuration and beam quality using an x-ray spectral simulation algorithm (xSpect, Henry Ford Hospital, Detroit, MI), (Samei and Flynn 2003) which employed the known x-ray tube kVp, inherent tube filtration and specific phantom composition based on the assumption that the imaging system employed is an ideal photon-counting detector. The metrology of this paper is applicable to both energyintegrating and photon-counting detectors, but the latter was used for defining the ideal detector for the DQE measurements in compliance with the IEC standards. 2.4.5. Noise. For each phantom configuration, the noise performance of the system, as reflected by the effective noise power spectrum (eNNPS), was evaluated with the phantom placed against the detector cover plate using an air KERMA level of E0, where E0 is the air KERMA measurement extrapolated to the image receptor, measured using the target mAs determined from a phototimed exposure. For each phantom and each spectral condition, six images of the phantom were acquired in order to obtain a total of 4000 000 independent pixels in an 18 cm × 18 cm analysis area within the central portion of the phantom. The image data were then processed to compute the noise power spectrum using a previously described technique (Dobbins et al 2006). The central 18 cm × 18 cm analysis area of each image was divided into a 7 × 7 array of 128 pixel × 128 pixel analysis ROIs. Data were then detrended using a second-order polynomial background subtraction and converted to fractional values. The eNNPS for each ROI was obtained by Fourier transformation, followed by averaging of the NNPS estimates obtained from each of the 49 ROIs across all six images for that condition. Finally, the directional NNPS was computed by band averaging near the frequency axis (excluding the orthogonal on-axis data) and rebinning of the NNPS values

Effective dose efficiency

5105

into spatial frequency sampling intervals of 0.05 mm−1 (IEC 2003). This measurement was intended to capture the noise performance of the overall system as in routine clinical use; hence, these results capture the fixed pattern (high-frequency) noise resulting from the use of the anti-scatter grid and the scattered radiation, while the low frequency non-uniformities not generally considered noise are minimized. 2.4.6. Sharpness. For each phantom and kVp condition, the eMTF in the object plane was measured with the phantom placed against the detector cover plate, using an opaque edge test device placed at the nominal surface of the phantom. The edge test device (TX5 W Edge Device, Scanditronix-Wellh¨offer, Schwarzenbruck, Germany) was oriented vertically and positioned on the x-ray beam axis at 5 cm from the surface of the phantom facing the x-ray source, to measure the horizontal MTF orthogonal to the anode–cathode axis. (The vertical MTF was also initially measured but was found to be very similar to the horizontal MTF and thus was not included in the full study.) This setup permitted the estimation of the eMTF in the object plane corresponding to maximum magnification, or the worst-case scenario for typical clinical imaging with an approximate chest thickness corresponding to the phantom of interest. To assess the impact of phantom configuration and kVp on the MTF, independent of geometric magnification and focal spot blurring effects, measurements were also taken at a common reference point along the beam axis corresponding to the nominal surface of the large adult phantom. Three images were acquired at an air KERMA of 3E0 at 80, 100, 120 and 140 kVp for each phantom configuration, at the common reference point and at the nominal phantom surface. Only one measurement position was employed for the large adult phantom since the reference point corresponded to the nominal surface for this phantom. The MTF was then computed using a previously described method (Samei et al 1998, 2006), which included the determination of the edge and line spread functions, from which the MTF was derived by Fourier methods. The MTF results for the three images at each condition were subsequently averaged and rebinned into spatial frequency sampling intervals of 0.05 mm−1 (IEC 2003). This methodology permitted the assessment of the sharpness performance in the presence of the anti-scatter grid while including the effects of focal spot blur, geometric magnification, and scattered radiation, to capture the overall system performance as in routine clinical chest radiography. 2.4.7. Computation of the eDQE. The eDQE in the object plane of interest (nominal surface of phantom) was computed for each phantom configuration at each spectral condition using equation (1) and the previously determined factors, eMTF, eNNPS, T, SF, E0 and q, to reflect the typical chest radiography conditions encountered clinically. Computations were performed in the horizontal direction as the MTF was only measured in that direction. 2.5. Computation of the ED The ED per unit entrance skin exposure (air KERMA) was calculated from organ doses computed from simulations and organ weighting factors obtained from ICRP 103 (ICRP 2007) using a commercially available Monte Carlo simulation package for calculating patient doses in diagnostic x-ray imaging (v2.0.1, PCXMC, STUK, Helsinki, Finland) (Tapiovaara et al 1997). This package employs pre-defined models of the human body at six different ages (Christy and Eckerman 1987), i.e. newborn, 1, 5, 10, 15 years of age and adult, assuming hermaphrodite anatomy using the average height and weight for the specific population (age) of interest. User-defined specifications of height and weight are also supported, in which case

5106

E Samei et al

the age-appropriate phantom model is geometrically scaled by increasing the length and/or thickness of the phantom model; that mode was employed for the present study. To compute the ED for patient body habitus equivalent to that represented by each of the phantom configurations, the height and weight parameters tabulated in table 1 were employed. The NEXTTM pediatric phantom is based on a 15 month old infant with a mass (weight) of 10.9 kg (Food and Drug Administration and Conference of Radiation Control Program Directors 2003). The average height associated with a male infant of this weight and age was estimated from the 50% point on published length versus weight growth charts reported by the Centers for Disease Control and Prevention (CDC 2000). (The male metrics were used in this study as that was the basis of the design of the NEXTTM adult phantom employed in the study.) The NEXTTM adult phantom is based on a representative average adult male with a height of 172.2 cm and a mass (weight) of 74.8 kg (Food and Drug Administration and Conference of Radiation Control Program Directors 2003). The large adult phantom comprising the NEXTTM adult phantom with an additional 57.2 mm of acrylic and 0.2 mm of Al was estimated to be representative of a larger adult with a height of 172.2 cm and a mass (weight) of 136.1 kg. The ED per unit entrance air KERMA (ES) for the patient-equivalent of each phantom was computed further assuming an SID of 180 cm, an x-ray field centered between the breasts, and an entrance beam dimension (width × height) of 16 cm × 19 cm for the pediatric phantom (SSD = 161.2 cm), 36 cm × 33 cm for the average adult (SSD = 154.4 cm) and 39 cm × 33 cm for the large adult (SSD = 147.2 cm), consistent with imaging protocols in use at our institution. In all cases, the arms were assumed to be raised and outside of the field of view. The effect of entrance beam size was evaluated for the pediatric condition only using a square x-ray beam with diameters ranging from 16 cm through 36 cm, in increments of 2 cm. The Monte Carlo computation employed a million photon histories per energy level for each of 15 energy levels from 10 to 150 keV, in 10 keV increments. The x-ray spectrum simulation assumed an anode angle of 14◦ , 1.48 mm Pyrex, 3.0 mm of oil and 1.9 mm added Al filtration. The organs included in the computation of ED included the following: skeleton, brain, heart, testes, spleen, lungs, ovaries, kidneys, thymus, stomach, salivary glands, oral mucosa, pancreas, uterus, liver, lower intestine, upper intestine, small intestine, thyroid, bladder, gall bladder, esophagus, prostate and pharynx/trachea/sinus. 2.6. Computation of the eDE The eDE was computed using equation (3) employing the previously determined estimates of phantom entrance air KERMA (ES), ED, and eDQE for each phantom configuration at each spectral condition in the horizontal direction, together with the known x-ray tube SID and x-ray tube source-to-phantom surface distances (SSD), and computed squared input SNR, q, for each phantom/spectral condition. 3. Results 3.1. Intermediary results The measured air KERMA extrapolated to the phantom surface, ES (table 2), analogous to the entrance skin exposure during clinical examinations, increased with increasing phantom effective thickness (table 1) at each spectral condition (kVp). This resulted from the differential attenuation of the x-ray beam with different phantom configurations and use of the phototimer to establish the target mAs for each phantom/kVp condition. For a given phantom configuration, the ES decreased going from 80 to 120 kVp, but remained generally

Effective dose efficiency

5107

(a)

(b)

(c)

Figure 2. The system noise performance under clinical chest radiography conditions are reflected by the eNNPS results obtained at the surface of the pediatric ((a), (d)), adult ((b), (e)) and large adult ((c), (f)) phantoms at tube potentials of 80, 100, 120 and 140 kVp, for both the horizontal ((a)–(c)) and vertical ((d)–(f)) orientations.

unchanged from 120 to 140 kVp. Table 2 further tabulates the intermediary results of the phantom transmission fraction, SF, and ideal SNR2. 3.2. Noise Figure 2 shows the effective NNPS results (in the presence of phantom and anti-scatter grid) in both the horizontal and vertical orientations for the pediatric, adult and large adult

5108

E Samei et al

(e)

(d)

(f)

Figure 2. (Continued.)

phantom configurations at 80, 100, 120 and 140 kVp. When comparing the noise result for a given phantom configuration and orientation, the small differences in noise magnitude with increasing kVp are consistent with normal phototimer response and spectral sensitivity variation. A comparison of the horizontal and vertical orientation results for each phantom configuration demonstrates the same order of magnitude of noise in both orientations, except for the horizontal orientation at frequencies approaching the Nyquist cutoff, where the presence of the grid is associated with increased fixed pattern noise (Samei et al 2009) resulting in a high frequency noise spike at spatial frequencies near the Nyquist frequency for the object plane,

Effective dose efficiency

5109

Table 2. Factors employed in the computation of eDQE.

X-ray tube potential (kVp)

80

100

120

140

NEXTTM pediatric NEXTTM adult Large adult

Air KERMA at image receptor (E0) (μGy) 58.8 43.0 37.7 41.2 98.2 70.2 59.6 62.3 322.7 192.1 166.6 164.9

NEXT pediatric NEXTTM adult Large adult

Entrance air KERMA (ES) (μGy) 67.5 49.1 43.9 47.4 134.2 96.5 81.6 85.1 481.5 286.8 248.2 246.4

NEXTTM pediatric NEXTTM adult Large adult

Narrow beam transmission fraction (%) 15.1 17.3 19.2 20.5 6.9 8.9 10.5 11.7 1.7 2.3 3.0 3.5

NEXTTM pediatric NEXTTM adult Large adult

25.6 27.2 37.6

NEXTTM pediatric NEXTTM adult Large adult

23 727 26 723 28 381

TM

Scatter fraction (%) 30.2 33.3 35.7 31.3 34.2 36.5 42.6 46.0 48.0 SNRin2 (μGy−1 mm−2) 26 045 27 053 27 268 28 640 29 088 28 758 29 979 29 979 29 190

fNm, given by the product of the Nyquist frequency and the magnification in the object plane. When comparing the pediatric, adult and large adult horizontal results, this noise spike is seen at higher spatial frequencies with increasing physical thickness of the phantom, corresponding to increased magnification in the object plane, i.e. the surface of the phantom closest to the x-ray source.

3.3. Sharpness Figure 3(a) demonstrates a very small degradation in the MTF with increasing kVp for the adult phantom configuration, which reflects the same trend with increasing kVp for the large adult phantom, across spatial frequencies above 0.2 mm−1. The pediatric phantom demonstrated this trend for spatial frequencies from 0.2 to 2.0 mm−1, beyond which the differences in MTF results were indistinguishable. The data are reported with respect to the nominal surface of the adult phantom. Changing the reference plane to the detector plane would lower the MTF, as included as a reference line in figure 3(a). The difference will have a corresponding effect (proportional to MTF2) on the subsequent quantities of eDQE and eDE. The MTF results for all three phantoms are almost indistinguishable, when measured at a common reference point corresponding to the nominal surface of the large adult phantom (figure 3(b)). However, a comparison of the MTF results obtained at the nominal surface of

5110

E Samei et al

(a)

(b)

(c)

Figure 3. The system sharpness performance under clinical chest radiography conditions are reflected by the horizontal eMTF results obtained at the nominal surface of the adult phantom (at 153.7 cm from focal spot) at tube potentials of 80, 100, 120 and 140 kVp (a). The eMTF results are reported for all phantoms at a common reference point corresponding to a location along the beam axis at the nominal surface of the large adult phantom (at 147.3 cm from focal spot), evaluated at tube potentials of 80, 100, 120 and 140 kVp (b). The effects of focal spot blurring enhanced by geometric magnification are reflected in the eMTF results for the object plane corresponding to the nominal surface of each phantom (at 168.5, 153.7 and 147.3 cm from focal spot for the pediatric, adult and large adult phantoms, respectively), evaluated at tube potentials of 80, 100, 120 and 140 kVp (c).

Effective dose efficiency

5111

Table 3. ED computed from Monte Carlo simulation.

X-ray tube potential (kVp)

NEXTTM pediatric NEXTTM adult Large adult NEXTTM pediatric NEXTTM adult Large adult

80

100

120

140

Effective dose per entrance surface air KERMA (ES) (μSv μGy–1) 0.36 0.41 0.44 0.47 0.24 0.30 0.34 0.36 0.19 0.23 0.27 0.30 Effective dose (μSv) 24.4 19.9 19.2 22.0 32.9 28.6 27.4 31.3 90.8 66.7 66.6 73.1

each phantom (figure 3(c)) shows increasing degradation in the MTF measured in the object plane, with increasing physical thickness of the phantom. 3.4. Effective dose As reported in table 3, estimated ED per unit entrance air KERMA (ES) is highest for the pediatric phantom and lowest with the large adult phantom with the average adult phantom falling intermediately between the two. For all three conditions the ED/ES increases linearly with kVp. The total ED reported in table 3 is highest at 80 kVp for all three phantom conditions, lower at 100 and 120 kVp with a slight increase at 140 kVp. For the large adult condition, the ED is 36% higher at 80 kV compared to 120 kVp. Similarly, for the adult and pediatric conditions, the ED at 80 kVp are, respectively, 20% and 27% higher than the result at 120 kVp. The ED at 140 kVp is 16%, 14% and 9% higher than the result obtained at 120 kVp for the pediatric, adult and large adult phantoms, respectively. 3.5. Effective DQE and eDE The effective DQE results are shown in figure 4 and table 4. As expected, the eDQE results are almost an order of magnitude smaller than typical conventional DQE results reported for this system (Ranger et al 2007). For all three phantom configurations, the eDQE was highest at 80 kVp, decreased with increasing kVp, and was lowest at 140 kVp. In comparison with the adult and large adult eDQE results, the pediatric results showed a sharp drop at high spatial frequencies approaching the Nyquist limit. The eDE results are shown in figure 5 and table 5. For all phantom configurations, the magnitude of eDE was highest at 80 kVp, decreased with increasing kVp, and was lowest at 140 kVp, similar to the results for the eDQE. The advantage of incorporating the ED into an image quality metric predicated on the measured system DQE (eDQE) is best illustrated by comparing paired phantom configurations, i.e. the results of pediatric versus adult and adult versus large adult phantom configurations. The eDQE results for the adult and pediatric phantom conditions are consistent at the lowest spatial frequencies but there is a relative drop off in the eDQE result for the adult phantom in comparison to the pediatric condition at mid- to high-spatial frequencies, as well as a sharp drop off in eDQE at the highest spatial frequencies approaching the Nyquist limit, as previously

5112

E Samei et al

(a)

(b)

(c)

Figure 4. The eDQE results for the pediatric (a), adult (b) and large adult (c) phantom configurations at tube potentials of 80, 100, 120 and 140 kVp. For all phantom configurations the highest eDQE was at 80 kVp (black lines) and the lowest at 140 kVp (pale gray lines). For each plot, the spatial frequencies correspond to the object plane at the surface of the corresponding phantom (at 168.5, 153.7 and 147.3 cm from focal spot for the pediatric, adult and large adult phantoms, respectively).

noted. The eDE results show a marked difference between the results for the pediatric and adult phantom conditions at low frequencies. The eDE results for the adult phantom show the same drop off relative to the pediatric phantom but this relative drop off is shifted to the mid- to high-frequency range with the same sharp drop off in the pediatric results at spatial frequencies approaching the Nyquist limit. At each kVp, the eDQE results for the adult configuration are consistently higher than the results for the large adult configuration across all spatial frequencies. The 80 and 100 kVp eDE results for the adult and large adult agree well at low spatial frequencies but the large adult results fall off with respect to the adult results at mid- to high-spatial frequencies. For the 120 and 140 kVp results, there is a consistent relative increase in the adult phantom eDQE results relative to the large adult phantom results, across all spatial frequencies. 4. Discussion The broad utilization of digital radiographic imaging systems for chest imaging provides a strong motivation for establishing reliable image quality metrics to characterize these systems

Effective dose efficiency

5113

Table 4. Measured eDQE.

X-ray tube potential (kVp)

80

100

120

140

Effective detective quantum efficiency NEXTTM pediatric (spat. freq. mm−1) 0.5 ± 0.1 1.0 ± 0.1 1.5 ± 0.1 2.0 ± 0.1 2.5 ± 0.1 Low–mid frequency (0.5 to 1.0) Mid frequency (1.0 to 1.5) Mid–high frequency (1.5 to 2.0) NEXTTM adult (spat. freq. mm−1) 0.5 ± 0.1 1.0 ± 0.1 1.5 ± 0.1 2.0 ± 0.1 2.5 ± 0.1 Low frequency (0.5 to 1.0) Mid frequency (1.0 to 1.5) High frequency (1.5 to 2.0) Large adult (spat. freq. mm−1) 0.5 ± 0.1 1.0 ± 0.1 1.5 ± 0.1 2.0 ± 0.1 2.5 ± 0.1 Low frequency (0.5 to 1.0) Mid frequency (1.0 to 1.5) High frequency (1.5 to 2.0)

0.098 0.087 0.068 0.044 0.015 0.093 0.078 0.056

0.091 0.078 0.060 0.039 0.013 0.086 0.070 0.050

0.081 0.066 0.052 0.033 0.012 0.073 0.060 0.043

0.071 0.061 0.046 0.030 0.010 0.066 0.054 0.038

0.100 0.077 0.048 0.021 0.006 0.092 0.062 0.033

0.088 0.065 0.039 0.016 0.004 0.078 0.052 0.027

0.078 0.059 0.034 0.014 0.003 0.071 0.047 0.023

0.071 0.052 0.029 0.011 0.003 0.063 0.041 0.020

0.084 0.057 0.029 0.009 0.001 0.071 0.042 0.018

0.076 0.050 0.024 0.007 0.001 0.064 0.037 0.015

0.061 0.042 0.019 0.006 0.0004 0.052 0.030 0.012

0.055 0.036 0.017 0.005 0.000 0.046 0.026 0.010

and optimize their utilization in the clinic. Historically, the DQE has been used to assess these systems but there is a growing recognition of the limitations of the DQE in reflecting the performance clinically. Recently, some investigators have introduced extensions of the DQE methodology that are intended to capture the performance in a clinically realistic context (Kyprianou et al 2004, 2005a, 2005b, Yadava et al 2005, Shaw et al 2000, Samei et al 2008, 2009, Ranger et al 2009). However, system DQE methods, even when targeted to characterize the performance of the system under conditions that reflect clinical usage, do not capture the relative performance of these systems in a full context from the perspective of radiation detriment, i.e. absorbed dose to the patient. Therefore, if used alone, they would have limited use for technique optimization applications. Incorporation of the system DQE metric into a broader metric that also factors in the ED to the patient can address that need. This dose to information conversion efficiency metric concept was introduced by Tapiovaara (1993), utilizing conventional detector-centric measures that do not reflect the application-specific sharpness and noise performance. Further incorporating radiation dose in that approach, in this study we proposed a new metric, the effective dose efficiency or eDE, which utilizes

5114

E Samei et al

(a)

(b)

(c)

Figure 5. The eDE results for the pediatric (a), adult (b) and large adult (c) phantom configurations at tube potentials of 80, 100, 120 and 140 kVp. For all phantom configurations the highest eDE was at 80 kVp (black lines) and the lowest at 140 kVp (pale gray lines). For each plot, the spatial frequencies correspond to the object plane at the surface of the corresponding phantom (at 168.5, 153.7 and 147.3 cm from focal spot for the pediatric, adult and large adult phantoms, respectively).

the output SNR squared normalized by the ED. The eDE was evaluated for its utility as a performance metric with applications for technique optimization in chest radiography in terms of image quality performance and dose, using phantom configurations representing the pediatric, average adult, and large adult populations. Our findings show that the eDQE is an order of magnitude smaller in comparison to the typical DQE results. Assuming an ideal imaging system would have an efficiency of 1, this implies that there is substantial room for improvement in terms of system design and performance efficiency. In contrast, the eDE metric which incorporates ED has a magnitude that is related to the employed units of dose providing means to balance image quality and dose. One of the most noteworthy findings of the study is that for each phantom tested, the lowest kVp provided the best performance in terms of the eDQE and eDE. This finding is consistent with earlier studies reporting similar results when using lesion detection as a FOM (Dobbins et al 2001). Specifically, the study of Dobbins et al shows that the best lesion SNR can be obtained at a low kVp. That study concluded a higher kVp as optimum primarily to

Effective dose efficiency

5115

Table 5. Measured eDE.

X-ray tube potential (kVp)

80

100

120

140

Effective dose efficiency (mSv−1) NEXTTM pediatric (spat. freq. mm−1) 0.5 ± 0.1 1.0 ± 0.1 1.5 ± 0.1 2.0 ± 0.1 2.5 ± 0.1 Low–mid frequency (0.5 to 1.0) Mid frequency (1.0 to 1.5) Mid–high frequency (1.5 to 2.0) NEXTTM adult (spat. freq. mm−1) 0.5 ± 0.1 1.0 ± 0.1 1.5 ± 0.1 2.0 ± 0.1 2.5 ± 0.1 Low frequency (0.5 to 1.0) Mid frequency (1.0 to 1.5) High frequency (1.5 to 2.0) Large adult (spat. freq. mm−1) 0.5 ± 0.1 1.0 ± 0.1 1.5 ± 0.1 2.0 ± 0.1 2.5 ± 0.1 Low frequency (0.5 to 1.0) Mid frequency (1.0 to 1.5) High frequency (1.5 to 2.0)

5.1 4.5 3.5 2.3 0.8 4.9 4.1 2.9

4.7 4.0 3.0 2.0 0.7 4.4 3.6 2.6

4.0 3.2 2.5 1.6 0.6 3.6 2.9 2.1

3.3 2.8 2.2 1.4 0.5 3.1 2.5 1.8

8.0 6.2 3.9 1.7 0.4 7.4 5.0 2.7

6.3 4.6 2.8 1.2 0.3 5.5 3.7 1.9

5.0 3.8 2.1 0.9 0.2 4.5 3.0 1.5

4.1 3.0 1.7 0.6 0.2 3.6 2.3 1.1

8.5 5.8 2.9 0.9 0.1 7.2 4.3 1.8

6.6 4.3 2.1 0.6 0.01 5.5 3.2 1.3

4.5 3.1 1.4 0.4 0.03 3.9 2.2 0.9

3.6 2.4 1.1 0.3 0.02 3.0 1.7 0.7

balance the lesion contrast with respect to the rib contrast. Nonetheless, if lesion conspicuity due to the overlying rib contrast is assumed to be of secondary importance relative to lesion conspicuity due to the SNR alone, the study clearly pointed to a need to reduce kVp for digital chest radiography exams, presently performed with 120 kVp and an anti-scatter grid. However, it should be noted that a reduction in kVp, if implemented, would be associated with an increase in mAs and exposure times, increasing the likelihood of patient motion blur. Furthermore, such implementation would require a follow-up study to ascertain the need for the utilization of an anti-scatter grid and an optimization of image processing to ensure proper contrast rendition. As expected for the pediatric phantom, the eDQE performance was relatively preserved at increasing spatial frequencies relative to the other phantom configurations as a result of the decreased focal spot blurring due to the smaller thickness of the phantom. However, this improved performance is independent of the associated ED. When the ED is incorporated into the performance metric, the performance under conditions similar to the pediatric condition is markedly inferior to that for the average and large adult configurations, as reflected by a

5116

E Samei et al

lower eDE. This is primarily due to the higher value of ED per ES for that phantom/population, possibly due to less self-shielding in children compared to adults. The relative performance for the adult and large adult phantoms was generally comparable with some differential spectral sensitivity when comparing results obtained using different kVps. The eDE was demonstrated to be sensitive to body habitus and x-ray spectrum. As such, it may be useful for technique optimization where the objective is to establish the radiographic technique (kVp, filtration, SID, field size, etc) that yields an optimal level of image quality per unit dose. The implementation of such technique guidelines will still heavily rely on the optimization of the system exposure factors, ideally obtained from the use of a well-calibrated and stable automatic exposure control system during clinical operation. Independent verification of such systems should be undertaken when establishing optimized clinical imaging protocols, to ensure that appropriate diagnostic reference levels and/or regulatory limits are adhered to. While eDE can be viewed as a natural extension of the current concepts of dose (i.e. ED) and image quality (i.e. DQE), and appropriately captures their current metrics into one, it should not be interpreted as a final metric for quantification of imaging performance. We should recall that ED by itself is a definitional construct (as opposed to a specific measurable physical quantity) to provide a metric of radiation risk. As our understanding of radiation risk expands, newer and better metrologies may be incorporated into the eDE concept. For example, perhaps ED may be replaced by risk index (Li et al 2011), recently proposed as a more advanced metric of radiation risk taking into account newest organ-based risk coefficients as well as their age and gender dependences. In terms of image quality as well, a frequencyintegrated form of the eDE taking into account frequency weighting functions corresponding to specific clinical tasks of interest may be implemented to extend the reach of the eDE concept to optimizations in terms of targeted clinical tasks (Boyce and Samei 2006, Richard and Samei 2010, International Commission on Radiation Units and Measurements (ICRU) 1996). The current paper did not include those additional extensions as it aimed to set the ground work in definition a task generic metric of performance (similar to DQE) within a manageable length, but they certainly deserve follow-up investigations and implementations. The results reported here must be interpreted in the context of the limitations of this study. This study employed specific geometrical phantoms. While based on standard FDA designs, the results are dependent on the specific phantom used. Future studies should deploy more realistic anthropomorphic models which include the complications associated with anatomical noise. Secondly, the geometric magnification and the impact on the focal spot blurring was maximized in this study as the measurements were obtained at the phantom surface; anatomical structures positioned closer to the detector will be less impacted. This strategy reflects the worst-case scenario for these effects and that should be noted in the interpretation of the results. Thirdly, while the eDE methodology captures the output SNR squared per unit dose, it does not reflect the differential contrast encountered in clinical chest imaging (e.g. contrast between mediastinum and lung areas), the impact of motion unsharpness that can be enhanced with the use of lower kVp/higher mAs acquisition settings, and the effect and the required optimization of image post-processing when the acquisition technique is optimized. Those factors may be incorporated in future studies. These limitations notwithstanding, the eDE methodology offers a significant extension of DQE to capture the imaging system performance in the context of the absorbed dose penalty required to attain a given quantum signal to noise squared. Lastly, the current study demonstrated the utility of eDE to assess the performance of a single digital radiographic system from a single vendor. The metrology can readily be applied to other digital systems from other vendors, a worthy goal for future studies.

Effective dose efficiency

5117

5. Conclusions We have introduced a new metric, the effective dose efficiency or eDE, which has been demonstrated to be useful in evaluating the performance of digital radiographic imaging systems for chest radiography applications. The eDE has been shown to be sensitive to body habitus and radiographic technique, demonstrating its usefulness in clinical technique optimizations. Acknowledgments The authors wish to thank Thomas Boehringer and Cynthia Browning of Duke University for their technical assistance and Dr David C Spelic of the FDA for the loan of phantoms used in this study. References Boyce S J and Samei E 2006 Imaging properties of digital magnification radiography Med. Phys. 33 984–96 CDC 2000 http://www.cdc.gov/growthcharts Christy M and Eckerman K F 1987 Specific absorbed dose fractions of energy at various ages from internal photon sources—appendix A: description of the mathematical phantoms ORNL Publication TM-8381/VI (Oak Ridge, TN: Oak Ridge National Laboratory) Conway B J, Butler P F, Duff J E, Fewell T R, Gross R E, Jennings R J, Koustenis G H, McCrohan J L, Rueter F G and Showalter C K 1984 Beam quality independent attenuation phantom for estimating patient exposure from x-ray automatic exposure controlled chest examinations Med. Phys. 11 827–32 Dobbins J T III, Samei E, Chotas H G, Warp R J, Baydush A, Floyd C E and Ravin C E 2001 Chest radiography: optimization of x-ray spectrum for cesiumn iodide–amorphous silicon flat-panel detector Radiology 226 221–30 Dobbins J T III, Samei E, Ranger N T and Chen Y 2006 Intercomparison of methods for image quality characterization: II. Noise power spectrum Med. Phys. 33 1466–75 Ertan F, Mackenzie A, Urbanczyk H J, Ranger N T and Samei E 2009 Use of effective detective quantum efficiency to optimise radiographic exposures for chest imaging with computed radiography Proc. SPIE 7258 72585O Floyd C E, Lo J Y, Chotas H G and Ravin C E 1991 Quantitative scatter measurement in digital radiography using a photostimulable phosphor imaging system Med. Phys. 18 408–13 Food and Drug Administration and Conference of Radiation Control Program Directors 2003 Nationwide Evaluation of X-ray Trends: Twenty-five Years of NEXT Brochure pub. Hart D, Jones D G and Wall B F 1994a Estimation of effective dose in diagnostic radiology from entrance surface dose and dose-area product measurements Report NRPB-R262 (London: HMSO) Hart D, Jones D G and Wall B F 1994b Normalised organ doses for medical x-ray examinations calculated using Monte Carlo techniques Report NRPB-SR262 (Chilton: NRPB) Huda W and Gkanatsios N A 1997 Effective dose and energy imparted in diagnostic radiology Med. Phys. 24 1311–6 ICRP 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60 Ann. ICRP 21 1–3 ICRP 2007 Recommendations of the International Commission on Radiological Protection, ICRP Publication 103 Ann. ICRP 21 2–4 IEC 2003 Medical electrical equipment—characteristics of digital x-ray imaging devices: part 1. Determination of the detective quantum efficiency IEC 62220-1 (Geneva: International Electrotechnical Commission) Illers H, Buhr E and Hoeschen C 2005 Measurement of the detective quantum efficiency (DQE) of digital x-ray detectors according to the novel standard IEC 62220-1 Radiat. Prot. Dosim. 114 39–44 International Commission on Radiation Units and Measurements (ICRU) 1996 Medical imaging—the assessment of image quality Report 54 (Bethesda, MD: ICRU) Kyprianou I S, Ganguly A, Rudin S, Bednarek D R, Gallas B D and Myers K J 2005a Efficiency of the human observer compared to an ideal observer based on a generalized NEQ which incorporates scatter and geometric unsharpness: evaluation with a 2AFC experiment Proc. SPIE 5749 251–62 Kyprianou I S, Rudin S, Bednarek D R and Hoffmann K R 2004 Study of the generalized MTF and DQE for a new microangiographic system Proc. SPIE 5368 349–60

5118

E Samei et al

Kyprianou I S, Rudin S, Bednarek D R and Hoffmann K R 2005b Generalizing the MTF and DQE to include x-ray scatter and focal spot unsharpness: application to a new microangiographic system Med. Phys. 32 613–26 Lee S C, Wang J N, Liu S C and Jiang S H 2007 Evaluation of dose-image-quality optimization in digital chest radiography Nucl. Instrum. Methods A 580 544–7 Li X, Samei E, Segars W, Sturgeon G, Colsher J and Frush D P 2011 Patient-specific dose and risk estimation in pediatric chest CT Radiology 259 862–74 Liu H, Zhuo W, Chen B, Yanling Y and Dehong L 2008 Patient doses in different projections of conventional diagnostic x-ray examinations Radiat. Prot. Dosim. 132 334–8 McCullough C H and Schueler B A 2000 Calculation of effective dose Med. Phys. 27 828–37 Ranger N T, Mackenzie A, Honey I D, Dobbins J T III, Ravin C E and Samei E 2009 Extension of DQE to include scatter, grid, magnification, and focal spot blur: a new experimental technique and metric Proc. SPIE 7258 7258A Ranger N T, Samei E, Dobbins J T III and Ravin C E 2007 Assessment of detective quantum efficiency: intercomparison of a recently introduced international standard with prior methods Radiology 243 785–95 Richard S and Samei E 2010 Quantitative imaging in breast tomosynthesis and CT: comparison of detection and estimation task performance Med. Phys. 37 2627–37 Samei E 2003 Performance of digital radiography detectors: factors affecting sharpness and noise Advances in Digital Radiography (Oak Brook, IL: Radiological Society of North America (RSNA)) pp 49–61 Samei E and Flynn M J 2003 An experimental comparison of detector performance for direct and indirect digital radiography systems Med. Phys. 30 608–22 Samei E, Flynn M J and Reimann D A 1998 A method for measuring the presampled MTF of digital radiographic systems using an edge test device Med. Phys. 25 102–13 Samei E, Ranger N T, Dobbins J T III and Chen Y 2006 Intercomparison of methods for image quality characterization: I. Modulation transfer function Med. Phys. 33 1454–65 Samei E, Ranger N T, Mackenzie A, Honey I D, Dobbins J T III and Ravin C E 2008 Detector or system? Extending the concept of detective quantum efficiency to characterize the performance of digital radiographic imaging systems Radiology 249 926–37 Samei E, Ranger N T, Mackenzie A, Honey I D, Dobbins J T III and Ravin C E 2009 Effective DQE (eDQE) and speed of digital radiographic systems: an experimental methodology Med. Phys. 36 3806–17 Shaw C C, Liu X, Lemacks M, Rong J X and Whitman G J 2000 Optimization of MTF and DQE in magnification radiography—a theoretical analysis Proc. SPIE 3977 467–75 Tapiovaara M 1993 SNR and noise measurements for medical imaging: II. Applications to fluoroscopic x-ray equipment Phys. Med. Biol. 38 1761–88 Tapiovaara M, Lakkisto M and Servomaa A 1997 PCXMC: a PC-based Monte Carlo program for calculating patient doses in medical x-ray examinations Report STUK-A139 (Helsinki: Finnish Centre for Radiation and Nuclear Safety) Ullman G, Sandborg M, Dance D R, Hunt R A and Carlsson G A 2006 Towards optimization in digital chest radiography using Monte Carlo modelling Phys. Med. Biol. 51 2729–43 Yadava G K, Kyprianou I S, Rudin S, Bednarek D R and Hoffman K R 2005 Generalized performance evaluation of x-ray image intensifier compared with a microangiographic system Proc. SPIE 5745 419–29