This Article Abstract Figures Only Full Text (PDF) Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by McVey, G Articles by Alm Carlsson, G Search for Related Content PubMed PubMed Citation Articles by McVey, G Articles by Alm Carlsson, G

A study and optimization of lumbar spine X-ray imaging systems

¹Joint Department of Physics, The Royal Marsden NHS Trust, Fulham Road, London SW3 6JJ, UK and ²Department of Radiation Physics, Faculty of Health Sciences, Linköping University, SE581 85 Linköping, Sweden

Correspondence: G McVey: North Wales Medical Physics, Glan Clwyd Hospital, Bodelwyddan, Denbighshire LL18 5UJ, UK

	Abstract

Top Abstract Introduction Methods and materials Results Discussion Conclusions References

 
A Monte Carlo program has been developed that incorporates a voxel phantom of an adult patient in a model of the complete X-ray imaging system, including the anti-scatter grid and screen–film receptor. This allows the realistic estimation of patient dose and the corresponding image (optical density map) for a wide range of equipment configurations. This paper focuses on the application of the program to lumbar spine anteroposterior and lateral screen–film examinations. The program has been applied to study the variation of physical image quality measures and effective dose for changing system parameters such as tube voltage, grid design and screen–film system speed. These variations form the basis for optimization of these system parameters. In our approach to optimization, the best systems are those that can match (or come close to) the calculated image quality measure of systems preferred in a recent European clinical trial, but with lower patient dose. The largest dose savings found were 21% for a 400 speed class system with a grid having a strip density of 40 cm^-1 and a grid ratio of 16. A further dose saving of 13% was possible when a 600 speed class system was employed. The best systems found from the optimization correspond to those recommended by the European Commission guidelines on image quality criteria for diagnostic radiographic images.

Key Words: A7g • B1h

	Introduction

Top Abstract Introduction Methods and materials Results Discussion Conclusions References

 
Lumbar spine radiographs allow clinicians to judge the configuration and alignment of bones with a high degree of accuracy. Malalignment or other changes in the shape of the vertebrae can then be identified and may imply the presence of a tumour, fracture or infection. Lumbar spine radiography is a routine examination for lower back pain, which is very common; 27 patients per 1000 inhabitants in the UK undergo plain radiography of the lumbar spine each year [1]. These examinations contribute 4.3% of the annual collective effective dose for all medical and dental exposures compared with 0.9% for chest examinations in the UK [2].

Optimization is necessary to balance the requirement for good image quality with low patient dose. The Commission of the European Communities (CEC) image quality criteria [3] describe the presentation of the normal anatomy in a lumbar spine radiograph. Almén et al [4] have evaluated the image quality of lumbar spine radiographs using the CEC criteria [3]. These studies showed that systems using a low tube voltage (70 kV) and a medium speed class (400) for the screen–film receptor fulfilled more of the image criteria for the anteroposterior (AP) projection than those using high tube voltage and high speed class. The systems using high speed class (600) and low tube voltage (77 kV) fulfilled more criteria for the lateral (LAT) projection than those using low speed class and high tube voltage. Vañó et al [5] optimized lumbar spine imaging by varying different technical parameters and found the largest dose saving by decreasing the optical density by changing the settings of the automatic exposure control (AEC). Almén et al did not study the effect of optical density as they did not use AEC.

The assessment by Almén et al [4] of clinical image quality has been complemented by theoretical modelling as part of the same project. A realistic Monte Carlo model of the patient (voxel phantom) and the complete imaging system has been developed [6] for this purpose. The model can be used to calculate physical measures of image quality and patient dose. In Sandborg et al [7], the correlations between our calculated physical measures of image quality and the clinical assessments of image quality are presented for chest and lumbar spine radiographs. For the latter, the signal-to-noise ratio (SNR) of trabecular structures was found to be a good predictor of clinical image quality. This paper presents the application of the Monte Carlo program to the study and optimization of lumbar spine imaging. The optimization approach is similar to that used, with the same model, for chest radiography [8] and involves the use of a reference system known to be of good image quality. Preliminary results for this study are outlined in Dance et al [9]. In this paper, our preliminary study is considerably extended so that the influence of tube voltage, grid design, screen–film speed and operating optical density are all considered for both AP and LAT projections.

	Methods and materials

Top Abstract Introduction Methods and materials Results Discussion Conclusions References

 
Monte Carlo model and voxel phantom
A Monte Carlo computer program has been developed to simulate diagnostic X-ray examinations. It is based on our earlier work, which employed a homogeneous phantom in a model of the complete imaging system [10, 11]. For the present model, the program has been extended by the inclusion of a voxel phantom to provide a more accurate model of the patient. The program transports photons through the patient and the anti-scatter grid to the imaging device. The energy imparted to the voxel phantom allows the patient dose to be calculated and the energy imparted to the screen allows the image quality measures to be calculated. These parameters are discussed in more detail below.

The voxel phantom used is that developed by Zubal et al [12, 13] and was obtained by segmentation of a series of CT slices of an adult male. Female organs (breasts, uterus and ovaries) were added by us to facilitate the calculation of effective dose [14]. An extra layer of voxels was included in the phantom to model the couch top. For the AP view, the dimensions of the voxel phantom were 899 mm from the top of the head to the bottom of the pelvis (236 voxels), 356 mm wide (128 voxels) and 214 mm thick (77 voxels). The phantom's length was adjusted to correspond to the sitting height of the average European male [15]. The phantom's width and thickness were found by the comparison of calculations of entrance air kerma for the voxel phantom with measurements from a patient study [6, 16]. Each voxel in the phantom belongs to 1 of 55 organs and each organ is associated with one of four tissue types: average soft tissue (1030 kg m^-3); lung (260 kg m^-3); average bone (1490 kg m^-3); or bone spongiosa (1180 kg m^-3). Tissue compositions were obtained from the International Commission on Radiation Units and Measurements (ICRU) [17], except for average bone, which was taken from Kramer [18].

Figure 1 shows the model of the voxel phantom and the components of the imaging system. The photon spectrum was obtained from Birch et al [19]. A grid was used as the anti-scatter technique and was specified in terms of strip density N, lead strip width d, grid ratio r and the materials in the interspaces and covers. The model of the image receptor included the cassette front, fluorescent screen and film characteristic (H and D) curve, measured by Dr F Verdun, Lausanne (personal communication). The Monte Carlo code calculates the contrast and SNR of anatomical details at different positions in the image to provide a physical measure of image quality. These parameters were calculated with a large number of photon histories so that the uncertainty of their values is less than ±3% (1 standard deviation).

View larger version (44K): [in this window] [in a new window]   Figure 1. The imaging system included in the Monte Carlo model of the lumbar spine anterior-posterior projection. The bony structures in the voxel phantom have been highlighted.

Small high contrast details were also chosen. These were the trabecular structures on the L1, L3 and L5 vertebrae in the AP projection, referred to subsequently as L1D, L3D and L5D, respectively. For the LAT projection, the trabecular structures were selected to be at an anterior position on the L1 and L5 vertebrae and at a posterior position on the L5 vertebra, referred to subsequently as L1F, L5F and L5B, respectively). All of the trabecular structures were 1 mm thick. This is quite similar to the "important" detail size of 0.3 mm to 0.5 mm given in the CEC image quality criteria document [3]. Trabecular structures were simulated as bone marrow cavities (1030 kg m^-3) and their contrast was calculated against a background of cortical bone. The compositions of the anatomical details and tissue backgrounds were taken from the ICRU [17].

Image quality and patient dose parameters
Contrast
Contrast was calculated in the Monte Carlo program as the difference in optical density (OD) beside and behind the important details superimposed on the voxel phantom. The effects of film gradient and imaging system unsharpness were taken into account in the calculation of OD using:

The H and D curve was measured in accordance with the ISO-standard [20] by Dr F Verdun, Lausanne (Private Communication, 1998). The film gradient () was derived from the H and D curve for the OD beside the detail (OD_det). The quantity cdf_MTF is the reduction in contrast caused by the total system unsharpness (total modulation transfer function, MTF_tot). Image receptor (screen–film) and geometric unsharpness (focal spot size and magnification) are all taken into account in the calculation of the MTF_tot. The MTFs of the screen–film combination were also measured by Dr F Verdun, Lausanne (personal communication). Sandborg et al [8] describes the calculation of cdf_MTF.

The object contrast C was found from Monte Carlo calculations of energy imparted to the fluorescent screen per unit area:

Here, _p1 and _p2 are the energy imparted to the receptor per unit area by primary photons beside and behind the detail, respectively, and _s is the energy imparted to the receptor per unit area by scattered photons. The notation E denotes the expectation value. It was assumed that the detail does not alter the distribution of scattered photons in the imaging plane.

Signal-to-noise ratio
The SNR of the ideal observer, SNR_I [21], of a small detail at an optical density OD_det was obtained using:

The SNR_MC (_det) was calculated by the Monte Carlo program. It was obtained from the energy imparted to an area of the detector A_MC with and without the detail being present assuming that the only noise source is quantum mottle and neglecting image unsharpness. The was scaled from the area of the detector element A_MC=0.25 mm² to the area of the detail A. The SNR_MC due to quantum noise has been shown to give good agreement with experiments [22, 23] for details with diameters larger than or equal to 3 mm. Hence, as many of the details used in our Monte Carlo model were similar to or smaller than 3 mm, the model needs to take into account the effect of image unsharpness. This was implemented using the SNR degradation factor SNRDF, which also accounts for: (i) the different efficiencies with which the signal and quantum noise are transferred through the screen caused by light emitted from different depths in the screen [24]; (ii) the statistical variations in the transport of light from the screen to the film [25]; and (iii) the total system noise including that from the film. These corrections are derived following the methods of Nishikawa and Yaffe [26]. A more detailed description of the implementation is given by Sandborg et al [8].

Calculation of entrance air kerma
The Monte Carlo program calculates air kerma, without backscatter, at the entrance surface of the phantom, air kerma at the surface of the cassette front and energy imparted to the screen per unit area. The entrance air kerma for a fixed OD can be calculated using these quantities combined with the H and D curve measured in terms of the cassette entrance air kerma. The calculation was implemented in two parts.

In the first part, the experimental set-up used to measure the H and D curve was simulated and the air kerma at the surface of the cassette front and the energy imparted to the screen per unit area calculated. In this way, the H and D curve was expressed as the OD for a given value of the energy imparted to the screen per unit area.

In the second part, the voxel phantom in the lumbar spine imaging system under investigation was simulated. Ratios of energy imparted to the screen per unit area to the incident air kerma at the phantom were calculated for approximately 200 evenly spaced points of interest across the whole image and the median ratio found. The calibrated H and D curve was used to convert an OD to be used as a normalization point, for example, the median OD of a radiograph or set of radiographs, to an energy imparted per unit area. The entrance air kerma was then calculated by this value of the energy imparted divided by the median value of the ratio.

Effective dose
Effective dose has been used in this work to quantify the radiation risk. The voxel phantom was segmented into organs each with known mass. The Monte Carlo code calculated the energy imparted to each voxel associated with an organ. The organ dose was obtained by dividing the sum of the energy imparted to all voxels of an organ with the mass of that organ. The effective dose was then found by combining the organ doses with the tissue weighting factors according to the International Commission on Radiological Protection [14]. The Monte Carlo code calculates the ratio of the effective dose to the incident air kerma at the voxel phantom surface. The effective dose for a given situation was found from the product of this ratio and the incident air kerma (see previous section).

Validation of the model
The Monte Carlo program has been validated in two parts. Firstly, Monte Carlo calculations of OD behind polymethyl methacrylate (PMMA) phantoms were compared with measurements carried out under carefully controlled conditions. Good agreement, within 13% was found providing that there was detailed knowledge of the imaging system [6, 16]. Secondly, patient images were collected and the entrance air kerma measured for chest and lumbar spine examinations in both frontal and lateral projections. The images were digitized and analysed. Measurements of contrast were extracted from the digitized radiographs. For the lumbar spine AP projection, it was found that the calculated entrance air kerma was slightly lower than the minimum value in the range of measured entrance air kermas. This was due to the voxel phantom being slightly thinner than required. However, as the calculated entrance air kermas were within the range of measured values for the other projections, it was decided not to increase the thickness of the voxel phantom for the lumbar spine AP projection as the calculated value was still reasonably representative of the range of calculated values. The program was also successfully validated by comparing the calculated contrast of important anatomical details and the calculated dynamic range of the image with the range of measured values [6, 16]. The voxel phantom was thus found to be sufficiently representative of a patient undergoing both chest and lumbar spine X-ray examinations.

Reference system and optimization
In order to optimize the parameters used in X-ray imaging systems, one system had to be identified that provided good image quality, and this was designated as the reference system. Thus, we determined a suitable reference system to be the imaging system that produced images with the highest image quality as judged by an expert panel of European radiologists in a recent clinical trial [4]. These preferred images were thus the reference images. Table 1 shows the characteristics of the reference imaging systems for the AP and LAT views. The reference systems used 72 kV with a 400 speed class screen–film system for the AP view and 77 kV with a 600 speed class screen–film system for the LAT view. The preferred system from the clinical trial corresponds to the good radiographic practice outlined in the CEC image criteria document [3], except that a lower tube voltage was used than suggested by the guidelines.

A good quality image may be one that fulfils its diagnostic purpose but may not always be an image with the highest possible contrast or SNR [27]. In our optimization scheme, it was decided to use the best systems from the clinical trials as the reference systems and the images they produce as the reference images. It was assumed that an image for which the contrast or SNR were 10% lower than those in the reference image would still fulfil its diagnostic purpose. Values of SNR_I and OD were calculated for each detail for a specified scatter-rejection technique, speed class, OD and film type for tube voltages between 60 kV and 110 kV in the AP view and between 70 kV and 110 kV in the LAT view. The tube voltages required to give 0.9 of the appropriate SNR_I and OD value for each detail were then deduced. The detail requiring the lowest tube voltage is referred to as the limiting detail. This tube voltage is the highest employable that ensures all details fulfil the criterion of the associated image quality measure being greater than or equal to 0.9 of that for the reference system. The effective dose is calculated for this limiting tube voltage and compared with the values for the reference system. The procedure is then repeated for different imaging systems and the system resulting in the lowest effective dose is the optimum.

	Results

Top Abstract Introduction Methods and materials Results Discussion Conclusions References

 
Effect of varying image system parameters on patient dose and image quality
Tube voltage
Figure 2 shows the results for the AP projection of varying the tube voltage between 60 kV and 110 kV on (a) the effective dose, (b) the contrast of the L5 transverse process and (c) the SNR of a trabecular structure on the L1 vertebra. The reference system gives an incident air kerma without backscatter of 0.88 mGy and an effective dose of 0.12 mSv. The calculated incident air kerma is within the range of entrance surface doses given in Hart et al [28]. The calculated effective dose is lower than would be expected for example, from the effective doses given in Hughes [29], owing to the voxel phantom thickness being slightly thinner than is required. However, this will not affect the results as they are quoted relative to the reference system values in this paper. The effective dose decreases by 73% between 60 kV and 110 kV. The three transverse processes show approximately the same variation of contrast with tube voltage. The same applies to the SNR for the three trabecular structures. For example, the contrast of the L5 transverse process decreases by 54% between 60 kV and 110 kV, with a similar decrease in SNR of 58% for the L5 trabecular structure.

View larger version (22K): [in this window] [in a new window]   Figure 2. The effect of tube voltage on (a) effective dose, (b) optical density (OD) difference of the L5 transverse process and (c) signal-to-noise ratio (SNR) of a trabecular structure on the L1 vertebra for the anteroposterior projection. The effect of tube voltage on (d) effective dose, (e) OD difference of the L3 spinous process and (f) SNR of a trabecular structure on the front of the L1 vertebra for the lateral projection.

Grid design
Figure 3 shows the results for the AP projection of increasing the grid ratio (r=8–16) for three grids: (1) strip density N=40 cm^-1, strip width d=40 µm, aluminium covers and interspaces; (2) the same parameters except with carbon fibre covers and interspaces; and (3) N=70 cm^-1, d=20 µm, carbon fibre covers and interspaces. The figure shows the variation of (a) effective dose, (b) contrast of the L3 transverse process and (c) the SNR of the trabecular structure on the L3 vertebra. The results are shown relative to the reference system, which has a grid constructed with N=52 cm^-1, r=10, d=36 µm with aluminium covers and carbon fibre interspaces. The effective dose increases for increasing grid ratio for all grids, for example, increasing by 34% for the aluminium grid between r=8 and r=16. The carbon fibre grids give the lowest effective dose. For the carbon fibre grid with N=40 cm^-1, the effective dose is lower by 11% (r=8) compared with the mixed material grid, and lower by 13% (r=8) compared with the aluminium grid. There is a further dose reduction by increasing the strip density and decreasing the strip width. The effective dose for the N=70 cm^-1, d=20 µm grid is lower by 19% (r=8) than for the N=40 cm^-1 carbon fibre grid.

View larger version (26K): [in this window] [in a new window]   Figure 3. The effect of two grids with strip density N=40 cm-1 with aluminium and carbon fibre covers and interspaces and a third grid with a strip density N=70 cm-1, strip width d=20 µm with carbon fibre covers and interspaces on (a) effective dose, (b) optical density (OD) difference of the L3 transverse process and (c) signal-to-noise ratio (SNR) of the trabecular structure on the L3 vertebra for the anteroposterior projection. The effect of the same grids on (d) effective dose, (e) OD difference of the L3 spinous process and (f) SNR of the trabecular structure at an anterior position on the L5 vertebra for the lateral projection.

Figure 3 also shows the results for the LAT projection for increasing grid ratio for the three grids mentioned above. The results show the variation of (d) effective dose, (e) the contrast of the L3 spinous process and (f) the SNR of the trabecular structure at an anterior position on the L5 vertebra. The dose reductions obtained with a carbon fibre grid are less for the LAT view than the AP view owing to the higher tube voltage. The effective dose for the N=40 cm^-1 grid (r=8) is 6% less than the reference system. By increasing the strip density and decreasing the strip width the dose is decreased by a further 14%.

The contrast and SNR advantages from using carbon fibre grids in the LAT view are generally the same or smaller than for the AP view. The contrasts obtained using the N=40 cm^-1, r=16 and N=70 cm^-1, d=20 µm, r=16 grids are 8% and 5% greater than for the reference system. The corresponding increases in SNR for these grids compared with the reference system are 14% and 3%, respectively. For grids with low grid ratios where there is less contrast or SNR than the reference system, the tube voltage does not need to be decreased significantly, especially if carbon fibre grids are used since the loss of contrast and SNR is small. For carbon fibre grids with high grid ratios, the tube voltage may be increased without losing contrast or SNR and therefore, such a system may have a significantly reduced dose.

Screen–film speed
Figure 4 shows the results for the AP projection of varying the speed class between 320 and 600 (all using TML film) on (a) the effective dose, (b) the contrast of the L3 transverse process and (c) the SNR of the trabecular structure on the L3 vertebra. The results are shown at both 72 kV and 90 kV. The effective dose decreases by 42% as the speed class increases from 320 to 600 for both 72 kV and 90 kV X-rays. At 72 kV, the contrast of the L3 transverse process is near its maximum value for the 400 speed class system. The contrast decreases by 10% and 3% when the 400 speed class system is replaced by a 320 and 600 speed class systems, respectively. This is due to differences in the shape of the H and D curves for the different screen–film combinations. At 72 kV, the SNR of the trabecular structure varies by a greater amount than the contrast. The SNR decreases by 19% for increasing the speed class from 320 to 600. Similar variations of contrast and SNR are observed at 90 kV. If a 600 speed class system is used instead of a 400 speed class system, the tube voltage has to be decreased slightly to regain the loss of contrast but significantly more to regain the loss of SNR. However, large dose reductions are still possible due to the greater sensitivity of the system, despite significantly lower tube voltages being required to maintain contrast and SNR.

View larger version (24K): [in this window] [in a new window]   Figure 4. The effect of screen–film speed class on (a) effective dose, (b) optical density (OD) difference of the L3 tranverse process and (c) signal-to-noise ratio (SNR) of the trabecular structure on the L3 vertebra for the anteroposterior projection. The effect of screen–film sensitivity class on (d) effective dose, (e) OD difference of the L3 spinous process and (f) SNR of the trabecular structure at an anterior position on the L5 vertebra for the lateral projection.

Optical density
Figure 5 shows the results for the AP projection of varying the value of the median OD between 0.2 and 3.0 using the Lanex Regular (Eastman Kodak Campany, Rochester, NY) screen with TML film on (a) effective dose, (b) the contrast of the L1, L3 and L5 transverse processes and (c) the SNR of the trabecular structures on the L1, L3 and L5 vertebrae. The effective dose increases linearly with OD between 0.4 and 1.6. There is a rapid increase in effective dose above an OD of 1.6 due to the shape of the TML H and D curve. The effective dose is 22% greater at a median OD of 1.6 than at the median OD of 1.36 used in the reference system. The transverse processes have a maximum contrast at different median ODs due to the differing OD beside each anatomical detail and, therefore, their position on the H and D curve. The L1, L3 and L5 transverse processes have maximum contrasts at ODs of 1.6, 1.4 and 1.2, respectively. The contrast of the L3 process at an OD of 1.6 is very similar to that at 1.36. The trabecular structures also have a maximum SNR at different median ODs. The details on the L1, L3 and L5 vertebrae have maximum SNRs at ODs of 2.6, 2.4 and 2.2, respectively. The maximum SNR values are 47%, 33% and 20% greater than the SNR values for the L1, L3 and L5 trabecular structures using the reference system.

View larger version (28K): [in this window] [in a new window]   Figure 5. The effect of median optical density (OD) on (a) effective dose, (b) OD difference and (c) signal-to-noise ratio (SNR) for the anteroposterior projection. The effect of median OD on (d) effective dose, (e) OD difference and (f) SNR for the lateral projection.

Results of optimization
Scatter rejection technique
Table 2 shows the tube voltages for the six important details which give 0.9 of the contrast and SNR values for the lumbar spine AP reference system. These results are for an imaging system using a grid with N=40 cm^-1 and r=8 and a Lanex Regular/TML screen–film system (400 speed class). The table shows that there are differences in the voltage required for each detail. The lowest tube voltage is found for the L1 trabecular structure. The imaging requirement for this detail then limits the tube voltage to be less than or equal to this value so that the image quality criterion is met for all six details. The L1 trabecular structure was found to be the limiting detail for this grid, and for some of the other grids investigated (Table 3), as in these cases the SNR for this detail had the largest response to tube voltage. Thus for the grid under investigation, the largest dose saving that can be achieved is 18%.

View this table: [in this window] [in a new window]   Table 2. The tube voltages which produce 10% lower contrast (OD, difference in optical density) or signal-to-noise ratio (SNR) for the six anatomical details (see Important Contrast Details section) than obtained with the lumbar spine anterior-posterior reference system using an imaging system with a N=40 cm-1, r=8, d=40 µm grid and a Lanex Regular/TML screen–film system (400 speed class). The corresponding effective dose relative to the value for the reference imaging system (Eastman Kodak Campany, Rochester, NY) is also given. The detail which limits the optimization, i.e. the one which requires the lowest tube voltage, is written in bold italics

View this table: [in this window] [in a new window]   Table 3. The best tube voltages and the corresponding values of the relative effective dose for each grid studied with a Lanex Regular/TML screen–film system (Eastman Kodak Campany, Rochester, NY) (400 speed class) for the anterior-posterior projection. The limiting detail (see Important Contrast Details section) and the image quality measure (difference in optical density (OD) or signal-to-noise ratio (SNR)) are also given. The systems which give the lowest patient dose are written in bold italics

Table 4 shows the optimization of different grid designs using the Lanex Regular/TML screen–film system for the LAT projection. It was found that there are no dose savings for these grids compared with the reference system. This is due to the reference system using the more sensitive Lanex Fast screen–film system (600 speed class). Therefore, a compromise for a 400 speed class imaging system would be to use a N=40 cm^-1 and r=8 grid in both the AP and LAT projections. This provides a small overall dose saving of 5% compared with the respective AP and LAT reference imaging systems.

View this table: [in this window] [in a new window]   Table 4. The best tube voltages and the corresponding values of the relative effective dose for each grid studied with a Lanex Regular/TML screen–film system (Eastman Kodak Campany, Rochester, NY) (400 speed class) for the lateral projection. The limiting detail (see Important Contrast Details section) and the image quality measure (difference in optical density (OD) or signal-to-noise ratio (SNR)) are also given. The system which gives the lowest patient dose is written in bold italics

View this table: [in this window] [in a new window]   Table 5. The best tube voltages and the corresponding values of the relative effective dose for each grid studied with a Lanex Fast/TML screen–film system (Eastman Kodak Campany, Rochester, NY) (600 speed class) for the anterior-posterior projection. The limiting detail (see Important Contrast Details section) and the image quality measure (difference in optical density (OD) or signal-to-noise ratio (SNR)) are also given. The system which gives the lowest patient dose is written in bold italics

Optical density
Figure 6 shows the optimization of the median OD, OD_med in the AP projection. The median OD was varied between 80% and 150% of the reference system value of 1.36. The system studied used a grid with N=40 cm^-1 and r=12 and a Lanex Regular/TML screen–film system. Figure 6a shows the variation of the highest tube voltage that fulfills the image quality requirement as a function of OD_med. The corresponding limiting detail and image quality parameter type are shown for each data point. The highest tube voltage increases with increasing OD_med until a maximum value of 81 kV is reached at an OD_med of 1.36 and then decreases. Below an OD_med of 1.36, the contrast and SNR of each detail all increase with increasing OD_med. The limiting detail is the trabecular structure on the L1 vertebra as its SNR has the largest response with OD_med. Above an OD_med of 1.36, the contrasts of the L5 and L3 transverse processes decrease with increasing OD_med. Therefore, the tube voltage has to be decreased in order to recover the reduced contrast to meet the required image quality criterion. The L5 transverse process is the limiting detail as its contrast has the largest decrease with increasing OD_med. Figure 6b shows that there are dose savings below an OD_med of 1.50 with the effective dose being 25% lower than the value for the reference system at an OD_med of 1.09.

View larger version (27K): [in this window] [in a new window]   Figure 6. The optimization of median optical density with (a) the highest tube voltage consistent with the requirement to obtain at least 90% of the image quality of the reference system for all details considered and (b) the corresponding values of the effective dose relative to the reference system values for the anterior-posterior view. The optimization of median optical density with (c) the highest tube voltage and (d) the corresponding effective dose for the lateral view.

	Discussion

Top Abstract Introduction Methods and materials Results Discussion Conclusions References

 
Our work on the optimization of the scatter rejection technique has shown that the tube voltage could be decreased or increased in order to produce a dose reduction depending on the grid design. Vañó et al [5] increased the tube voltage from 60 kV to 90 kV to produce a dose reduction of 35% whilst maintaining image quality for the lumbar spine AP examination. However, Almén et al [4] have shown that increasing the tube voltage from 70 kV to 90 kV significantly alters the image quality of AP films, whereas increasing from 77 kV to 95 kV does not significantly alter the image quality of LAT films. In our optimization studies, the tube voltages that fulfilled the image quality criterion were less than 85 kV for the AP films and 90 kV for the LAT films.

Further evidence that our work closely corresponds to that of Almén et al is given by Sandborg et al [30]. Sandborg showed that the physical parameters such as contrast and SNR could be used to predict the order that the imaging systems were ranked by the European radiologists [4]. For example, in the AP projection Almén et al found significant differences in image quality for changing tube voltage, but not for changing speed class. This can also be demonstrated from our study of changing the image parameters and observing their effect on calculated image quality. By increasing the tube voltage from 70 kV to 90 kV, a large decrease of 28% was observed in the calculated contrast and SNR whereas only a small decrease of 10% was seen in the SNR for increasing the speed class from 400 to 600. It is therefore reassuring that the work in this paper is consistent with changes in image quality observed clinically.

	Conclusions

Top Abstract Introduction Methods and materials Results Discussion Conclusions References

 
The results of varying the different imaging parameters shows how straightforward it is to have high image quality and high patient dose, e.g. low tube voltage and to have low image quality and low patient dose, e.g. high tube voltage. Conversely, it is difficult to balance high image quality and low patient dose. The optimization of radiographic imaging involves several different parameters. Therefore, it is very useful that a Monte Carlo model can be used to point out imaging systems that give low patient dose whilst still maintaining the same image quality as reference systems. These systems are worth investigation in future, more time-consuming, clinical trials.

For 400 speed class systems using grids in the AP projection, a dose reduction of between 8% and 22% can be achieved. A further dose reduction of 13% is possible with a 600 speed class system using a grid. Table 6 shows the imaging system configuration that produced the largest dose reduction in our work. Dose reductions of a similar size can be obtained for a grid with a high grid ratio (r=16), a high strip density (N=70 cm^-1) and a small lead strip width (d=20 µm).

Our work clearly shows that the largest dose reductions are for 600 speed class systems. However, in a recent review [28] of patient doses from screen–film imaging in the UK for the year 2000, the National Radiological Protection Board (NRPB) shows that there are significantly fewer 600 speed class systems in use compared with 400 speed class systems. The review [28] also shows the continuing trend for lower dose per lumbar spine radiograph of 37% in the period from 1984 to 1995 and 18% in the period from 1995 to 2000. The NRPB state that this is due to the increasing use of faster screen–film combinations. Therefore, our work highlights that there are still potential optimizations to be made in lumbar spine radiography. It is also reassuring to know that the systems found by the optimizations are similar to those recommended by the CEC guidelines [3] as given in Table 6.

	Acknowledgments

 
Dr F R Verdun (Lausanne, Switzerland) is thanked for supplying measured H and D curves, modulation transfer function and noise power spectra of the screen–film combinations used in this work. Alexandr Malusek is acknowledged for the image of the voxel phantom in Figure 1.

	Footnotes

 
This work was supported by grants from the Commission of European Communities (Nos. FI4P CT950005 and FIGM-CT2000-00036). The Swedish authors were supported by grants from the Swedish Radiation Protection Institute, SSI (Nos. SSI P1018.97 and SSI P1083.98), and the Swedish Foundation for Strategic Research (No. R98:006).

Received for publication April 8, 2002. Revision received August 30, 2002. Accepted for publication October 21, 2002.

	References

Top Abstract Introduction Methods and materials Results Discussion Conclusions References

 

1. Espeland A, Korsbrekke K, Albreksten G, Larsen JL. Observer variation in plain radiography of the lumbarsacral spine. Br J Radiol 1998;71:366–75.[Abstract]
2. Hart D, Wall BF. Radiation exposure of the UK population from medical and dental X-ray examinations. NRPB-W4, Didcot, Oxon: NRPB, 2002.
3. Carmichael JH, Maccia C, Moores BM, Oestmann JW, Schibilla H, Van Tiggelen R, Wall B, editors. European guidelines on quality criteria for diagnostic radiographic images. Report EUR 16260 EN. Luxembourg: European Commission, 1996.
4. Almén A, Tingberg A, Mattsson S, Besjakov J, Kheddache S, Lanhede B, et al. The influence of different technique factors on image quality of lumbar spine radiographs as evaluated by established CEC image criteria. Br J Radiol 2000;73:1192–9.[Abstract]
5. Vañó E, Oliete S, González L, Guibelaude E, Velasco A, Fernández JM. Image quality and dose in lumbar spine examinations: results of a 5 year quality control programme following the European quality criteria trial. Br J Radiol 1995;68:1332–5.[Abstract]
6. Sandborg M, McVey G, Dance DR, Persliden J, Alm Carlsson G. A voxel phantom based Monte Carlo computer program for optimization of chest and lumbar spine X-ray imaging. Rad Prot Dosim 2000;90:105–8.
7. Sandborg M, Tingberg A, Dance DR, Lanhede B, Almén A, McVey G, et al. Demonstration of correlations between clinical and physical image quality measures in chest and lumbar spine screen–film radiography. Br J Radiol 2001;74:520–8.[Abstract/Free Full Text]
8. Sandborg M, McVey G, Dance DR, Alm Carlsson G. Schemes for the optimization of chest radiography using a computer model of the patient and X-ray imaging system. Med Phys 2001;28:2007–19.[CrossRef][Medline]
9. Dance DR, McVey GH, Sandborg M, Alm Carlsson G, Verdun FR. The optimization of lumbar spine AP radiography using a realistic computer model. Rad Prot Dosim 2000;90:207–10.
10. Sandborg M, Dance DR, Persliden J, Alm Carlsson G. A Monte Carlo program for the calculation of contrast, noise and absorbed dose in diagnostic radiology. Comput Methods Programs Biomed 1994;42:167–80.[Medline]
11. Dance DR, Lester S, Alm Carlsson G, Sandborg M, Persliden J. The use of carbon fibre material in radiographic cassettes: estimation of the dose and contrast advantages. Br J Radiol 1997;70:383–90.[Abstract]
12. Zubal G, Harrell CR. Voxel based Monte Carlo calculations of Nuclear Medicine images and applied variance reduction techniques. Image and Vision Computing 1992;10:342–8.[CrossRef]
13. Zubal G, Harrell CR, Smith EO, Rattner Z, Gindi G, Hoffer PB. Computerised three dimensional segmented human anatomy. Med Phys 1994;21:299–302.[CrossRef][Medline]
14. International Commission on Radiological Protection. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60. Ann ICRP 1991:21.
15. International Commission on Radiation Units and Measurements. Phantoms and computational models in therapy, diagnosis and protection. ICRU Report No. 48. Bethesda, MD: ICRU, 1992.
16. Dance DR, McVey G, Sandborg M, Persliden J, Alm Carlsson G. Calibration and validation of a voxel phantom for use in the Monte Carlo modeling and optimization of X-ray imaging systems. Proc SPIE 1999;3659:548–59.
17. International Commission on Radiation Units and Measurements. Photon, electron, proton and neutron interaction data for body tissues. ICRU Report No. 46. Bethesda, MD: ICRU, 1992.
18. Kramer R. Determination of conversion factors between body dose and relevant radiation quantities for external x- and -radiation. Neuherberg, Germany: GSF, 1979.
19. Birch R, Marshall B, Ardan GM. Catalogue of spectral data for diagnostic X-rays. Scientific Report Series 30. London: The Hospital Physicists' Association, 1979.
20. International Standards Organization. Photography–sensitometry of screen–film systems for medical radiography—ISO method for determination of sensitometic curve shape, speed and average gradient. ISO 9236. Geneva, Switzerland: ISO, 1993.
21. International Commission on Radiation Units and Measurements. Medical imaging—the assessment of image quality. ICRU Report 54. Bethesda, MD: ICRU, 1996.
22. Tapiovaara MJ, Sandborg M. Evaluation of image quality in fluoroscopy by measurements and Monte Carlo calculations. Phys Med Biol 1995;40:589–607.[CrossRef][Medline]
23. Tapiovaara MJ, Sandborg M, Dance DR. A search for improved technique factors in paediatric fluoroscopy. Phys Med Biol 1999;44:537–59.[CrossRef][Medline]
24. Lubberts G. Random Noise produced by X-ray fluorescent screens. J Opt Soc Am 1968;58:1475–83.
25. Swank RK. Absorption and noise in X-ray phosphors. J Appl Phys 1973;44:4199–203.
26. Nishikawa RM, Yaffe MJ. Model of the spatial-frequency dependent detective quantum efficiency of phosphor screens. Med Phys 1990;17:894–904.[Medline]
27. Tingberg A. PhD thesis: quantifying the quality of medical X-ray images: an evaluation based on normal anatomy for lumbar spine and chest radiography. Lund, Sweden: Lund University, 2000.
28. Hart D, Hillier MC, Wall BF. Doses to patients from medical X-ray examinations in the UK—2000 Review. NRPB-W14. Didcot, UK: NRPB, 2002.
29. Hughes JS. Ionising radiation exposure of the UK population: 1999 review. NRPB-R311. Didcot, UK: NRPB, 1999.
30. Sandborg M, McVey G, Dance DR, Alm Carlsson G. Comparison of model predictions of image quality with results of clinical trials in chest and lumbar spine screen–film imaging. Rad Prot Dosim 2000;90:173–6.

This article has been cited by other articles:

				G McVey The effect of phantom type, beam quality, field size and field position on X-ray scattering simulated using Monte Carlo techniques Br. J. Radiol., February 1, 2006; 79(938): 130 - 141. [Abstract] [Full Text] [PDF]

				H. Geijer and J. Persliden Varied tube potential with constant effective dose at lumbar spine radiography using a flat-panel digital detector Radiat Prot Dosimetry, May 17, 2005; 114(1-3): 240 - 245. [Abstract] [Full Text] [PDF]

				G. Ullman, M. Sandborg, D. R. Dance, R. Hunt, and G. Alm Carlsson Distributions of scatter-to-primary and signal-to-noise ratios per pixel in digital chest imaging Radiat Prot Dosimetry, May 17, 2005; 114(1-3): 355 - 358. [Abstract] [Full Text] [PDF]

				L Court and T Yamazaki Technical note: a comparison of antiscatter grids for digital radiography Br. J. Radiol., November 1, 2004; 77(923): 950 - 952. [Abstract] [Full Text] [PDF]

				A Tingberg, C Herrmann, B Lanhede, A Almen, M Sandborg, G McVey, S Mattsson, W Panzer, J Besjakov, L G Mansson, et al. Influence of the characteristic curve on the clinical image quality of lumbar spine and chest radiographs Br. J. Radiol., March 1, 2004; 77(915): 204 - 215. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by McVey, G Articles by Alm Carlsson, G Search for Related Content PubMed PubMed Citation Articles by McVey, G Articles by Alm Carlsson, G