Optimisation of tube voltage range (kVp) for AP abdomen, pelvis and spine imaging of average patients with a digital radiography (DR) imaging system using a computer simulator

Abstract

Objectives:

To investigate via computer simulation, an optimised tube voltage (kVp) range for caesium iodide (CsI)-based digital radiography (DR) of the abdomen, pelvis and lumbar spine.

Methods:

Software capable of simulating abdomen, pelvis and spine radiographs was used. Five evaluators graded clinical image criteria in images of 20 patients at tube voltages ranging from 60 to 120 kVp in 10 kVp increments. These criteria were scored blindly against the same patient reconstructed at a specific reference kVp. Linear mixed effects analysis was used to evaluate image scores for each criterion and test for statistical significance.

Results:

Score was dependent on tube voltage and image criteria; both were statistically significant. All criteria for all anatomies scored very poorly at 60 kVp. Scores for abdomen, pelvis and spine imaging peaked at 70, 70 and 100 kVp, respectively, but other kVp values were not significantly poorer.

Conclusions:

Results indicate optimum tube voltages of 70 kVp for abdomen and pelvis (with an optimum range 70–120 kVp), and 100 kVp (optimum range 80–120 kVp) for lumbar spine.

Advances in knowledge:

There are no recommendations for optimised tube voltage parameters for DR abdomen, pelvis or lumbar spine imaging. This study has investigated and recommended an optimal tube voltage range.

Introduction

A computer algorithm capable of simulating digital radiographs (DR) of the chest, abdomen, pelvis and lumbar spine has been recently developed and validated by our group.^1,2 This algorithm has been used to examine optimised tube voltage (kVp) parameters for computed radiography (CR)³ and more recently DR⁴ imaging of the chest. Based on these results, our institution adopted the exposure parameters recommended.

Currently, there is no guidance from national or international professional bodies on the optimum tube voltage parameters for imaging of the abdomen, pelvis and lumbar spine anatomies with DR systems, and the guidance from the literature is scarce. Jacobs et al⁵ recommended pelvis DR imaging to be carried out at 135 kVp, but their study used a single phantom and measured physical image quality metrics only (i.e., signal and noise). They did not use images that contain realistic projected anatomy, which is essential to ensure the optimisation task is as clinically relevant as possible, nor did they ask expert image evaluators to grade the images. Other pelvis optimisation studies have focused on CR imaging only^6,7 or parameters other than tube voltage.^8,9 Of the official guidance that does exist, the Council of European Communities (CEC) Quality Criteria¹⁰ recommends 75–90 kVp for pelvis and 80–100 kVp for lumbar spine but this is for outdated film-screen technology. No such guidance exists for abdomen imaging.

In the UK, under the Ionising Radiation (Medical Exposure) Regulations 2017 (IR(ME)R17), all medical radiation doses to patients must be optimised, consistent with the intended purpose. It is therefore very important that optimised exposure parameters for X-ray examinations are investigated, especially where these parameters are out of date or do not exist. Exposure guidance, if appropriate, should subsequently be issued to the radiology community. This becomes even more important given the recent European wide study by McFadden et al,¹¹ who concluded that there was wide variation and lack of standardisation in exposure technique for abdomen and pelvis imaging.

This paper follows the successful approach we have used in recent years for CR and DR chest imaging using computer simulated images to investigate optimised tube voltage range for anteroposterior (AP) abdomen, pelvis and lumbar spine caesium-iodide (CsI)-based DR imaging. The evaluation of simulated images was carried out by experienced image evaluators (radiologists and reporting radiographers), so this work presents the results of a virtual clinical trial. Model observers (computational tools that aim to mimic the performance of real observers) were not used because there is currently no single accepted model that is universally applicable to the detection task in clinical images containing anatomical noise, and they cannot be adapted to the grading analysis required by this study.

Methods and materials

Synopsis of the computer simulator

The simulator is based on a digitally reconstructed radiograph (DRR) algorithm, that is a computer simulation of a conventional 2D X-ray image created from CT data. Although fully described in previous studies,^1–4 a brief synopsis of the algorithm is described below (all software coding in this paper was carried out using Matlab version 2012b, Mathworks, Natick, MA):

1. The virtual phantom is created from real patient CT datasets. The voxel resolution of the phantom is 0.34×0.39×0.34 mm (width x height x depth).

2. CT number of the virtual phantom is converted into linear attenuation coefficient (LAC) using formulae derived from the Gammex RMI (Gammex-RMI, Broadway Business Centre, Nottingham, UK) tissue equivalent phantom (model no. 467).

3. X-ray spectra are generated using the Spectrum Processor software which is based on Report 78 published by the Institute of Physics and Engineering in Medicine.¹²

4. Simulated X-ray pencil beams are projected and attenuated exponentially through the CT dataset using a ray casting method of DRR production. The intensity of photons in 0.5 keV energy bins emerging from the virtual phantom is calculated.

5. A DRR of pixel values in terms of absorbed energy per area (keV/mm²) in the DR system’s converter layer (caesium iodide phosphor) is then created. This is the raw DRR.

6. Frequency-dependent noise correct for detector incident dose and beam quality is added to the raw DRR.

7. Scatter measured experimentally across the diagnostic energy range using an anthropomorphic phantom with an anti-scatter grid on a real DR system are added to the raw simulated DRR.

8. DRR pixel values are converted to DR system pixel values using the measured system transfer properties (STP) of the real imaging system.

9. DRR images were validated quantitatively with a phantom² and real patient¹ images. Signal-to-noise ratio (SNR) values of the DRR images in the lung, spine and diaphragm areas where within 10% of the real images, irrespective of tube voltage. Histograms of pixel values were similar in shape. Qualitative validation was carried out by expert image evaluators and they all agreed that the DRRs adequately simulated real images, and that they were acceptable to use for optimisation studies.

Simulation of a real DR imaging system

The algorithm is capable of simulating AP abdomen, pelvis and lumbar spine DR images of average adults at various tube voltages, detector air kerma (DAK) values and with or without scatter rejection (using an anti scatter grid). For this study, the software was configured to simulate the Agfa DX-D400 (Agfa, Peissenberg, Germany) DR system with imaging panels 35×43 cm in size (pixel pitch 0.14 mm) utilising a caesium iodide (CsI) converter layer. Furthermore, the associated ceiling suspended Toshiba X-ray tube with total filtration equivalent to 3.2 mm of aluminium and an anode target angle of 12° was incorporated into the algorithm.

AP abdomen images produced by the simulator with tube voltages of 60 and 120 kVp are shown in Figure 1(a) and (b). Note that only examples of abdomen images are shown here because they also incorporate the other two anatomies investigated (pelvis and lumbar spine).

Figure 1.

(a) Simulated low tube voltage (60 kVp) abdomen image and (b) high tube voltage (120 kVp) abdomen.

As illustrated in Figure 1 (a) and (b), contrast is slightly poorer in the 120 kVp image, as expected, due to the reduction in the photoelectric interaction in the body. However, the 60 kVp image (Figure 1(a)) is noisier, especially in the spine region.

Simulated image reconstruction and preparation

For each of the three examination types, anonymised CT data sets from 20 average sized patients (70 ± 10 kg – identified with input from the expert scanning radiographer) were used to create simulated DR images. Permission from our local R&D committee was obtained prior to using retrospective CT data. Health Research Authority (HRA) or ethics committee approval was not necessary. To conform to local exposure protocol, the images were simulated with scatter rejection using an anti-scatter grid (strips per mm = 4, grid ratio = 12:1), at seven different tube voltages – 60 to 120 kVp in steps of 10 kVp. All abdomen images were produced with a matched effective dose of 0.32 mSv (±1%) , all pelvis images 0.27 mSv and all lumbar spine images 0.33 mSv by using an appropriate tube-current time product (mAs) required at each tube voltage to provide this effective dose. These values of effective dose were used because our radiology department’s local diagnostic reference levels (LDRL) are 1.8 Gycm² for abdomen, 1.9 Gycm² for pelvis and 1.5 Gycm² for lumbar spine imaging, respectively. Using conversion coefficients recommended in the Health Protection Agency’s (now Public Health England) report HPA-CRCE-028 returns the above values of effective dose. It should be noted that these values of effective dose were checked with commercial software capable of calculating effective dose corrected for beam quality (PCXMC Dose Calculations, v.2.0.1.4, STUK, Finland). All HPA calculated values were within 10% of PCXMC values. Images of a given patient were transferred to a folder and given the name Patient_N, were N was the sequential patient number. For example, patient 1 had 21 images (images 60 to 120 kVp in steps of 10 kVp for each of the three anatomies) transferred to its folder called Patient_1. This was repeated for all 20 patients. All patient folders were placed in a location accessible to image evaluators.

Evaluation of clinical image quality using a blind observer study

The methodology described in our chest imaging optimisation study⁴ for scoring simulated images was followed exactly in same manner in this study. A brief synopsis is described here:

1. Five experienced image evaluators (three Radiologists and two reporting Radiographers) graded all images on calibrated reporting PACS (Picture Archiving and Communications Systems) workstations with a dual monitor configuration (Barco Ltd, Brussels, Belgium). On a third screen, bespoke software (‘IQ_scoring’) was used for image grading; the software automatically opens three windows 1) the reference image for grading, 2) a test image and 3) a scoring panel. The reference image was kept on the right hand PACS monitor, the test image on the left hand PACS monitor and the scoring panel on a third monitor.

2. The reference image for each examination type was reconstructed with a tube voltage of 90 kV, as a reference value based on local protocol could not be ascertained. All evaluators were blinded to what tube voltage any of the images represented.

3. IQ_scoring displayed test images in a random manner and evaluators were asked to grade this image against the reference, using a grading panel. Changing window and level settings was permitted.

4. All image evaluators graded each test image on a flexible numerical continuous scale, using slider bars along which any point may be selected by the evaluator, rather than using an ordinal scale. This approach has recently been recommended by Keeble et al¹³ to circumvent some of the problems associated with the analysis of ordinal data for medical image optimisation. Placing the bar to the left/right of centre indicated inferior/superior image quality of the test image with respect to the reference for the specific criteria being scored. The positioning of each bar was subjective but to account for this a linear mixed effects algorithm was used to model the scores with evaluator as a ‘random effect’ (see next section).

5. The possible range for scoring spanned −3 to +3, for example, if an evaluator placed the bar all the way to the left this would record a score of −3 for that criteria. Each evaluator was blinded to this.

6. When this process was finished for a specific test image, the next random test image was automatically displayed and the scoring process was repeated until all test images were graded for all 20 patients. Note – each evaluator was forced to compare the reference image with itself to check for any observer bias.

Scoring criteria were based on the Council of European Communities Quality Criteria,¹⁰ revised to reflect modern diagnostic requirements and expertise of our image evaluators (Table 1).

Table 1.

The image criteria used for grading simulated images. All criteria were compared with the reference image

Image Criteria
Abdomen	Pelvis	Lumbar spine
Overall quality of image compared to reference	Overall quality of image compared to reference	Overall quality of image compared to reference
Overall quality of contrast	Quality of iliac wings	Quality of vertebra
Quality of trabeculae in iliac wings	Quality of sacral foramina	Quality of posterior vertebral edges
Quality of anatomical detail in bowel (such as haustra)	Quality of pubic and ischial rami	Quality of posterior elements (such as facets, pedicles, spinous process)
	Quality of sacroiliac joints	Quality of the L5 to S1 lumbosacral joint
	Quality of femoral heads	Quality of cortex

Analysis of evaluator scores using a linear mixed effects algorithm

Based on the recommendation of Keeble et al,¹⁴ the relationship between average evaluator score for each criteria and tube voltage was modelled with a linear mixed effects (LME) algorithm. The R statistical software package¹³ (R Foundation for Statistical Computing, Vienna, Austria) and lme4 (Bates et al¹⁵) were used for this exercise. The LME algorithm adjusts the average evaluator score for each criteria by modelling random effects such as intra- and inter-variability of image evaluators. The following full model was formulated in R for each anatomical region investigated:

Score ~TubeVoltage*ImageCriteria + (1|Patient) + (1|Evaluator) (1)

where ‘Score’ is the image quality LME adjusted score given by the observer (in the range −3 to +3), ‘TubeVoltage’ is the tube voltage being evaluated, ‘ImageCriteria’ is the structure that the score is being given for from Table 1, ‘Evaluator’ is the expert scoring the image and ‘Patient’ is the simulated patient being evaluated. Fixed effects in this model are ‘TubeVoltage’ and ‘ImageCriteria’, as we expect these will determine the score given by the observer in a determinate manner. Furthermore, we expect an interaction between the two fixed effects (as denoted by the *), as the evaluator may give different scores for a given tube voltage setting depending on which region of the image they are scoring, for example, 60 kVp may be good for soft tissue detail such as bowel, but poor in the trabeculae in iliac wings. ‘Patient’ and ‘Evaluator’ have been classed as random effects as they will influence the score but in a way that is impossible to fully control as the study only sampled 20 patients (per anatomical region) and five evaluators out of a much larger population, all of which may have slightly different ‘baseline’ scores. The aim of the study is to find the best option overall for all image evaluators.

The following reduced models were created to assess whether there is a statistically significant relationship between score and both tube voltage and image criteria:

Score ~ImageCriteria + (1|Patient) + (1|Evaluator) (2)

Score ~TubeVoltage + (1|Patient) + (1|Evaluator) (3)

Score ~1 + (1|Patient) + (1|Evaluator) (4)

Using R, a likelihood ratio analysis using the ANOVA statistical test was carried out to compare the likelihood of the full model with each reduced model in turn. This was done completely independent to one another. A p-value was returned for each test indicating whether removing tube voltage, image criteria or both in combination from the model had a statistically significant effect on the evaluator score. A p-value of less than 0.05 was considered statistically significant.

Results

For the abdomen region, it was found that:

• Tube voltage affected LME-modelled evaluator score in a statistically significant manner (χ²(24)=654, p < 10⁻¹⁵).

• Image criteria affected LME-modelled evaluator score in a statistically significant manner (χ²(21)=85, p < 10⁻⁹).

• Tube voltage and image criteria in combination affected LME-modelled evaluator score in a statistically significant manner (χ²(27)=665, p < 10⁻¹⁵).

For the pelvis region, it was found that:

• Tube voltage affected LME-modelled evaluator score in a statistically significant manner (χ²(36)=1407, p < 10⁻¹⁵).

• Image criteria affected LME-modelled evaluator score in a statistically significant manner (χ²(35)=176, p < 10⁻⁹).

• Tube voltage and image criteria in combination affected LME-modelled evaluator score in a statistically significant manner (χ²(41)=1421, p < 10⁻¹⁵).

For the lumbar spine region, it was found that:

• Tube voltage affected LME-modelled evaluator score in a statistically significant manner (χ²(36)=1536, p < 10⁻¹⁵).

• Image criteria affected LME-modelled evaluator score in a statistically significant manner (χ²(35)=64, p < 10⁻²).

• Tube voltage and image criteria in combination affected LME-modelled evaluator score in a statistically significant manner (χ²(41)=1546, p < 10⁻¹⁵).

The results of the optimisation study for each anatomical region are shown graphically in (Figure 2).

Figure 2.

2(a) Abdomen, 2(b) Pelvis and 2(c) Lumbar spine. Average image quality scores modelled using a linear mixed effects algorithm across all image criteria. Error bars demonstrate the standard error of the mean.

AP abdomen region

All criteria demonstrate very poor performance at 60 kVp scoring between −1.23 (overall) and −0.49 (contrast). However, scores for all criteria improve above this tube voltage, largely fluctuating around zero (taking error bars into account). Nevertheless, the highest scoring tube voltage in each region is 70 kVp, scoring between 0.05 (trabecular) and 0.02 (overall). All other tube voltages remain below zero.

AP pelvis region

In common with the abdomen, all criteria demonstrate very poor performance at 60 kVp scoring between −1.33 (overall) and −0.69 (femoral). Scores for all criteria above this tube voltage fluctuate around zero (taking error bars into account) suggesting any tube voltage above 60 kVp would be acceptable (as long as the tube current (mAs) is adjusted appropriately to compensate for effective dose). Nevertheless, in the overall, femoral, iliac and sacroiliac regions, the highest scoring tube voltage is 70 kVp (scoring between 0.04 and 0.00). In the pubic and ischial rami region, 100 kVp performed the best with a score of 0.01. The highest score in the sacral foramina region was 0.00 (100 kVp).

AP lumbar spine region

Once again, all criteria demonstrate very poor performance at 60 kVp scoring between −1.48 (L5 to S1) and −1.24 (cortex). Scores for all criteria follow a similar trend in that they peak at 100 kVp.

It should be noted that the scores for the 90 kVp images across all anatomies and image criteria are very close to zero (average less than 0.01) demonstrating each evaluator gave the 90 kVp test image a score at or very close to zero.

Discussion

This study has investigated optimised tube voltage range for abdomen, pelvis and lumbar spine imaging with a DR imaging system, the results of which may inform the medical physicist advising imaging departments for optimisation purposes. This has been demonstrated through expert grading of computer-simulated images.

For all anatomies, 60 kVp performed very poorly. In the abdomen and pelvis regions, all other tube voltages returned improved scores but largely fluctuated around zero, that is, the scores plateau from 70 kVp onwards. Taking error bars into account, the results suggest anything in the range 70–120kVp works equally well for matched effective dose. In the lumbar spine region, 70 kVp also performed relatively poorly, with other tube voltages returning improved scores, albeit fluctuating around zero, suggesting anything in the region 80–120 kVp would provide sufficient image quality.

Nevertheless, abdomen and pelvis imaging returned 70 kVp as the highest score in most criteria, while lumbar spine provided highest score at 100 kVp. CEC guidance recommends 75–90 kVp for film-screen pelvis imaging. This result of this study do not disagree, except suggesting a wider tube voltage range is possible, the lowest being 70 kVp. Similarly, recommendations for film-screen lumbar spine imaging is 80–100 kVp. As above, the results here for DR imaging do not contradict this guidance. There is no such guidance for abdomen imaging.

As we argued in our recent work on chest optimisation,⁴ the results of the virtual clinical trial presented in this work follow a roughly similar trend to the efficiency of CsI DR phosphors, which peaks between 70 and 90 kVp and drops off at tube voltages outside of this range. This would partly explain the poor performance of 60 kVp in all regions of all anatomies. Furthermore, given this work has investigated inherently dense regions of the body, it is probably that 60 kVp with its inferior beam quality and therefore penetration, coupled with the tube current-time product (mAs) used to provide matched effective dose, has resulted in images much poorer than the higher tube voltage counterparts. Similarly, 70 kVp did not perform well in the lumbar spine region for any criteria; again, this may be due to poorer penetration outweighing CsI absorption efficiency.

It is interesting that scores for high tube voltages do not drop off as suggested by the physics of CsI absorption efficiency. Furthermore, there is contrast loss (less photoelectric absorption within the patient) as the mean energy of the X-ray beam increases. However, the results presented in this paper suggest the use of higher beam qualities to image dense anatomical regions outweighs the poorer CsI efficiency and contrast loss inherent with using these higher energies.

The results also show that for matched effective dose, image quality was not as dependent on image criteria (p-values 10⁻³ to 10⁻¹⁰) as tube voltage (p-value ~ 10⁻¹⁶). This is probably because all regions of each anatomy are relatively dense and similar in terms of X-ray attenuation. This is contrary to the findings of our chest optimisation study but this is to be expected given the regions of the chest are very different (e.g., diaphragm and lung).

Although there was no evidence base to suggest using 70 kVp prior to this study, our radiology department adopted a standard operating protocol for average abdomen, pelvis and lumbar spine exposures of 70 kVp when the department transitioned from CR to DR in 2017. Fortuitously, this tube voltage matches the most superior in terms of scores returned for abdomen and pelvis derived in this study. In the intervening years, there have been no complaints about image quality and large-scale dose audits have consistently shown median dose-area-product (DAP) for standard sized patients are lower than the relevant national diagnostic reference levels (DRLs) for abdomen and pelvis imaging. Based on the results of this study, we will retain the current exposure protocols. However, this work has demonstrated scores peaked at 100 kVp for lumbar spine imaging and that 70 kVp was not optimal in any region. Moving away from 70 kVp for lumbar spine imaging will be justified.

Limitations

Limitations associated with this work are the same as that listed in our previous study.⁴ Briefly these are:

1. Limited simulated patient information set (n = 20).

2. Only ‘average’ patients have been investigated so the optimal tube voltages recommended here can only be used for patients of this size. Nevertheless, at the very least, they can be used as a starting point.

3. The results were derived using modelled computer-simulated images. Although we have thoroughly validated the algorithm, the results reported must be interpreted in the context of this limitation.

4. The simulated images do not take into account exposure time or patient movement artefacts, so one could argue that the tube voltages with shorter exposure times may be the most suitable. However, we investigated minimum possible exposure times for pelvis and spine imaging using a clinically appropriate phantom. The phantom was exposed on the imaging system where our algorithm is configured to simulate - it was possible to keep the exposure time below 400 ms (range 160 ms at 70 kVp to 12.5 ms at 120 kVp), for matched effective dose (60 kVp not considered due to relative poor image-quality scores). This meets the requirement of the CEC guidance which states pelvis and spine imaging should be carried out with exposure times of <1000 ms and <400 ms, respectively. Therefore, although our algorithm does not account for patient movement, in reality it is possible to minimise this anyway, irrespective of tube voltage chosen, and so should not be seen as a limiting factor.

5. This work is specific to Agfa CsI-based DR phosphor. However, given that most DR image devices are based on CsI phosphors, it is likely the conclusions of this work are transferable to other vendors.

Conclusion

We have demonstrated with simulated images containing realistic clinical content that abdomen and pelvis DR imaging is optimal at 70 kVp and lumbar spine imaging at 100 kVp, but taking error bars into account, the results suggest anything in the range 70–120kVp for abdomen and pelvis, and 80–120 kVp for lumbar spine, works equally well for matched effective dose. This will allow the medical physicist to advise imaging departments where to start in respect to optimised tube voltage for these examinations.