Experimental estimates of peak skin dose and its relationship to the CT dose index using the CTDI head phantom

Abstract

A straightforward method is presented to estimate peak skin doses (PSDs) delivered by computed tomography (CT) scanners. The measured PSD values are related to the well-known volume CT dose index (CTDI_vol), displayed on the console of CT scanners. PSD measurement estimates were obtained, in four CT units, by placing radiochromic film on the surface of a CTDI head phantom. Six different X-ray tube currents including the maximum allowed value were used to irradiate the phantom. PSD and CTDI_vol were independently measured and later related to the CTDI_vol value displayed on the console. A scanner-specific relationship was found between the measured PSD and the associated CTDI_vol displayed on the console. The measured PSD values varied between 27 and 136 mGy among all scanners when the routine head scan parameters were used. The results of this work allow relating the widely used CTDI_vol to an actual radiation dose delivered to the skin of a patient.

INTRODUCTION

There has been an increasing concern among the scientific community^(1–3) and regulating bodies^(4–6) regarding the radiation dose delivered during computed tomography (CT) examinations. The US Congress⁽⁷⁾ has stressed the lack of clear and documented information about such doses. In order to estimate the risk to patients from CT scans, an estimate of the dose delivered to the skin and organs of a patient is essential. A need therefore exists to determine appropriate dosimetric quantities such as the organ dose and peak skin dose (PSD)^(8–11). These quantities provide real estimates of dose being delivered to patients and can serve to validate Monte Carlo simulations⁽¹²⁾. In particular, measuring PSD is ideal because it is a well understood dosimetric quantity that directly relates to radiation-induced skin injuries^(13–16). Furthermore, measurement estimates of PSD values using suitable phantoms can easily be made across all different types of CT units and scan protocols available in clinics. This is important for comparing doses for the same CT examination across different facilities, which can vary significantly⁽¹⁷⁾.

The CT dose index (CTDI) is widely used for quality control involving the radiation output of CT machines. In particular, the volume CTDI (CTDI_vol) is displayed on the console of all CT machines and is readily available to the operator. However, the CTDI_vol does not represent the dose delivered to a patient during an actual procedure^{(18, 19)}.

In this work, the peak (maximum) skin dose delivered to a patient was estimated experimentally by measuring the dose delivered to the surface of a head phantom. The use of a phantom to measure skin dose allows establishing a standard protocol to compare the results across different CT units and facilities. From a radiation protection point of view, determining the maximum dose delivered to the skin, i.e. the PSD, allows deriving quantities that can be compared with dose reference levels set by national and international standards. The measured PSD values for a range of CTDI_vol values were provided. As a result, a relationship can be established between both quantities from which the PSD for other CTDI_vol values can be derived. The authors have focused on the case of a head examination, but the method can be extended to other CT protocols such as a CT scan of the chest or abdomen. Since PSD values are very dependent on the protocol parameters used at each facility, similar measurements of the PSD can be performed at clinics for any given examination protocol in order to derive scanner- and exam-specific relationships between CTDI_vol and PSD. The aim of this study is to present this method and evaluate its feasibility in the clinical environment.

MATERIALS AND METHODS

The measurements were performed with four CT units (Table 1): GE Discovery CT750HD, 128 slices [GE Healthcare (see Disclaimer), Waukesha, WI, USA]; Philips iCT256, 256 slices (Philips Healthcare, Amsterdam, The Netherlands); Siemens Definition, 64 slices (Siemens Medical Systems, Erlangen, Germany) and Toshiba Aquilion ONE, 320 slices (Toshiba Medical Systems, Tochigi-ken, Japan).

Table 1.

Scanners, parameters and CTDI_vol values used in this study.

CT Unit	Location	Tube voltage and current × time	Scan length (mm) (r×NT)	Rot. time (s)	SIDa (cm)	CTDIvol (mGy) console/chamber
GE	Walter R.	120 kV, 250 mA s	20 (4 × 5)	1	54	69.7/67.0 (1.04)
Philips	NIH	120 kV, 400 mA s	20 (2 × 10)	0.875	57	73.8/69.1 (1.07)
Siemens	NIH	120 kV, 338 mA s	20 (4 × 5)	2	60	36.1/36.4 (0.99)
Toshiba	NIH	120 kV, 600 mA s	40 (1 × 40)	1.5	60	156.6/127.6 (1.23)

Excluding the scan length, all other values correspond to the routine axial scan head protocol used at each facility. The parameter r is the number of rotations and the parameter NT is the number of slices per rotation (N) multiplied by the slice thickness (T). The corresponding models for the CT units listed in the table were GE Discovery Model CT750HD, Philips model iCT256, Siemens Definition and Toshiba Aquilion ONE.

^aSource-to-isocentre distance.

A comparison was made between the CTDI_vol value displayed on the CT console and the measured CTDI_vol value. The CTDI_vol was measured following the AAPM protocol^{(20, 21)}. The standard CTDI head phantom and CT ionisation chamber (both from Radcal Corp., Monrovia, CA, USA) are shown in Figure 1a. For every examined scanner, the CTDI_vol was obtained from scans in an axial mode⁽²²⁾ using the scan parameters of the routine head scan as shown in Table 1. The corresponding CTDI_vol displayed on the console was recorded to allow a comparison with the measured CTDI_vol values (Table 1).

Figure 1.

(a) A set-up used for the measurements showing the head phantom and the ion chamber located in the central hole. The cylindrical phantom has a diameter of 16 cm and a length of 15 cm. The position of the top and bottom strips of radiochromic film is indicated with white circumferences (note that the bottom one cannot be seen because it is placed under the phantom. (b) The calibrated densitometer used to measure the optical density of the film strips.

The ionisation chamber was calibrated in the X-ray beam facilities at the National Institute of Standards and Technology (NIST) in terms of air kerma with the beam quality M120 beam (all data related to this beam can be found at http://www.nist.gov/calibrations/x-gamma-ray.cfm), which corresponds to a medium-filtered beam with a peak voltage of 120 kV_p and a half-value layer of 6.72 mm Al. Although one single reference X-ray beam cannot match all beams from all CT units due to the variability in the filtration used by different manufacturers, the M120 beam quality represents well the spectral characteristics of the X-ray emissions of the CT units studied in this work since it has an HVL within the range of HVLs expected for the CT units used in this work (5–9 mm of Al)⁽²³⁾.

The skin dose measurements were performed using a Gafchromic film model XR-QA (International Specialty Products, Inc., Wayne, NJ, USA). This type of film has been shown to have relatively small energy dependence (3 %) in the range of diagnostic energies^{(24, 25)} and a low angular dependence (less than 1 % decrease in response between 0° and 70°, 3 % decrease between 70 and 85°, 9 % decrease on average between 85 and 95° and a peak of 51 % decrease at 90°⁽²⁶⁾). Therefore, the M120 beam is an appropriate choice to use for the calibration of these films. The films were calibrated at NIST in the X-ray reference standard beams. Each batch (containing 20 pieces of film) requires a separate calibration but, once calibrated, one batch is enough to perform the shown method more than 50 times. As described elsewhere^{(27, 28)} 1-cm²-film pieces were mounted on a polymethylmethacrylate slab phantom to simulate the presence of the human body and were irradiated to seven different air kerma values K (10, 25, 50, 100, 200 and 500 mGy). A calibration of the films in terms of the absorbed dose to water (and therefore in terms of PSD) was later derived by multiplying the delivered air kerma K by the backscatter factor B, estimated to be equal to 1.49⁽²⁹⁾ and by the mass–energy absorption ratio between water and air , estimated to be equal to 1.04⁽²⁹⁾. Three different sets of films from a single batch were exposed for each exposure value. The change in the optical density of the films was measured using a densitometer (Model 331, X-rite Inc. Grand Rapids, MI, shown in Figure 1b with an exposed film) that was calibrated at the Food and Drug Administration (no additional filtration was used). The calibration curve in terms of PSD for the films is shown in Figure 2. The uncertainty bars shown in the figure represent the average standard deviation of the optical density obtained by the three films calibrated at the same dose (2.3 % relative uncertainty).

Figure 2.

Calibration curve for film in terms of PSD up to 200 mGy. The optical density is shown in arbitrary units (a.u.). Uncertainty bars (horizontal bars) represent the variation of the various film readings to a fixed dose (see text). Note that for smaller PSD values the uncertainty bars are smaller than the symbols.

Experimental set-up and procedure

The top surface of the 16-cm CTDI phantom is assumed to be equivalent to the forehead of a patient, where the PSD is expected. The head phantom was aligned at the isocentre of the scanner with the chamber in the centre hole of the phantom. The longitudinal axis of the chamber and cylindrical phantom were aligned parallel to the longitudinal axis of the CT gantry as shown in Figure 1a. A preliminary scan or ‘scout view’⁽³⁰⁾ was performed to check the alignment of the phantom and to select the scan length (or scan range). A scan length of 20 mm was selected except for the case of Toshiba which allowed a minimum scan length of 40 mm. The scan parameters used and corresponding routine protocol values used at each facility for an axial head scan are listed in Table 1. It must be noted that the scan parameters were chosen to optimise the reconstructed images without applying any dose reduction algorithms such as the adaptive statistical iterative reconstruction algorithm available for the GE device.

Two film strips of dimensions 5 × 1 and 10 × 1 cm were placed at the top and the bottom of the phantom, respectively, as shown in Figure 1. The film strip below the phantom is longer to facilitate its alignment with the centre of the phantom. The phantom was scanned over the scan length for a fixed value of the tube current. To assess reproducibility, this measurement was repeated three times using the exact same scan parameters. Beyond the routine protocol values, the entire procedure was repeated for five additional tube current values within the available range of mA s, over and below the routine protocol mA s setting. Since the lowest dose that resulted in a visible change in the colour of the film was ∼1.5 mGy, and this corresponded to 40 mA s in the Siemens scanner, the minimum value in all scanners was set as close as possible to 40 mAs. The maximum mA s value allowed in each scanner was also included in the study. This ensured that the highest dose a scanner can deliver was measured, which allows for risk estimation in case such dose is delivered by mistake. The change in the optical density of the films due to the exposures from CT scans was read using the densitometer (Figure 1). The films were handled following recommendations⁽³¹⁾, and non-irradiated pieces of film were preserved as controls together with the irradiated film pieces. The use of control films allowed monitoring any changes that might occur to the films, such as accidental exposure to radiation during transport or handling, physical degradation or fading.

For the purpose of obtaining beam profiles, films were read using a flatbed scanner (Hewlett Packard flat-bed scanner model 7650), as suggested previously by other authors^{(25, 26, 32–34)}. Flatbed scanners are useful to reveal relative intensity and to study beam profiles as discussed below. However, they can present difficulties for absolute dose measurements since scanning methods and parameters are not standardised⁽³⁴⁾.

RESULTS

Table 1 includes a comparison between the CTDI_vol values measured with the ion chamber and the CTDI_vol value displayed on the console. The values displayed on the console for the CTDI_vol agree in general within ±5 % of the measured CTDI_vol except for the case of the Toshiba scanner, where the predicted value is ∼23 % higher than the measured value. It is noteworthy that Toshiba appears to overestimate the CTDI_vol even for a scan length of 4 cm, which is the smallest allowed in the scanner and is much smaller than the chamber length of 10 cm.

Figure 3 shows a graph of the measured PSD against the displayed CTDI_vol for the four CT units investigated. The results from each CT unit are represented by a solid and a dashed curve. The solid curve refers to the PSD measured at the top of the phantom and the dashed curve refers to the PSD measured at the bottom of the phantom. As seen in Figure 3, the PSD measured at the bottom of the head phantom is lower than the PSD measured at the top in all cases with an overall average difference of 12 %. This is probably due to the extra attenuation of the table at the bottom of the phantom. But as observed in Figure 3, the variations are quite spread (around the average value of 12 %) depending on the CT unit and the value of the CTDI_vol. For example, in the case of the Siemens scanner for a CTDI_vol value of 65 mGy, the PSD at the bottom of the head phantom is 35 mGy while at the top it is 42 mGy (a 20 % difference). These variations are probably due to the differences in the intrinsic filtration of the different manufacturers; a harder spectrum is expected to provide lower differences between the measurements at the top and bottom of the phantom. The relationships between measured PSD and displayed CTDI_vol values, given by the slopes in Figure 3, are 0.84±0.04, 0.94±0.05, 0.75±0.04 and 0.87±0.04 for the GE, Philips, Siemens and Toshiba scanners, respectively. These values are consistent with the results of recent Monte Carlo calculations published by Zhang et al.⁽¹²⁾. Because of the linear relationship between PSD and CTDI_vol, the PSD can be obtained for any given CTDI_vol as shown using a head protocol. Similarly, a PSD versus CTDI_vol relationship can be obtained for other protocols used in the facility. Finally, a facility can thus construct a look-up table from each protocol for a quick reference to obtain the PSD for any given CTDI_vol.

Figure 3.

The relationship between the displayed CTDI_vol and PSD measured at the top and bottom surfaces of the head phantom. The uncertainty bars represent the variations observed from repeated measurements made with the film and ion chamber with a given set of scanning parameters (see text). The corresponding models for the CT units in the graph were: GE Discovery Model CT750HD, Philips model iCT256, Siemens Definition and Toshiba Aquilion ONE.

The uncertainty bars shown in Figure 3 represent the standard deviation of the PSD values obtained from repeated measurements. The overall uncertainty of the PSD values was estimated to be within ±5 %. It is important to note that for a given CTDI_vol there is a clear difference in the PSD value measured among the various CT scanners. As shown in Figure 3, for a fixed CTDI_vol value of 65 mGy, the lowest and highest PSD values measured at the top surface of the head phantom were 42 mGy (for the Siemens scanner) and around 65 mGy (for the Toshiba and Philips scanners), respectively. This represents a difference of ∼50 % between the different manufacturers.

An interesting aspect of performing measurements with films is that a cross section of the beam profile can be obtained in a quite straightforward manner. Figure 4 shows beam profiles of all four scanners used. Note that since only a cross section of the beam is shown, a relative intensity of the CT X-ray beam is plotted as a function of the position along the scan axis. The intensity readings, obtained from scanning the films with a flatbed scanner, are expressed in arbitrary units (a.u.).

Figure 4.

Beam profiles obtained from the films placed at the bottom of the phantom, produced with the highest mA s setting used for each scanner. The corresponding models for the CT units shown were GE Discovery Model CT750HD, Philips model iCT256, Siemens Definition and Toshiba Aquilion ONE.

As shown in Figure 4, there are clear differences in the shape of these beam profiles. In the case of the Philips system, the scan length was covered with two rotations and a table feed value of 10 mm. The irradiated film strips clearly show that there is a slight overlap between the two rotations that covered the scan length (seen as an intensity enhancement at 0 mm in Figure 4). In the case of the Siemens, the scan length was covered by four rotations and a table feed value of 5 mm. As shown in Figure 4, for this scanner, there is no overlap at the end of each 5-mm slice. There is instead a small drop in the intensity (as seen at −5, 0 and +5 mm in Figure 4), likely resulting from a slight separation between each rotation. In the case of the GE device, a fairly constant value for the intensity across the whole scan length is observed as shown in Figure 4. The lower PSD values observed in Figure 3 for the Siemens and GE scanners, relative to the Philips scanner, are consistent with the overall reduced intensity observed in the beam profile across the entire scan length as shown in Figure 4. This may suggest that the Philips scanner is using an exaggerated beam width⁽³⁵⁾, but these considerations are out of the scope of this work. Note that while the scan length for the Philips, Siemens and GE is 20 mm (beam profiles span between −10 and 10 mm), the minimum scan length value for the Toshiba scanner is 40 mm. In the case of this scanner the intensity was observed to have a constant value across the whole scan length.

For the PSD measurements described in this paper, the associated uncertainties were estimated. These include the uncertainty of the air kerma rates delivered at NIST, the uncertainty of the film calibration, the intra-batch variability and the reproducibility of the measured PSD values for a fixed set of CT scan parameters. The relative standard uncertainty of the air kerma rates delivered at NIST was 0.3 %, representing a small contribution to the total uncertainty of the PSD measurement. The uncertainty of the film calibration is driven by the relatively short dynamic range of the densitometer resulting in a standard uncertainty of 2.3 %. The maximum observed intra-batch variability of 3.5 % was obtained from calculating the standard deviation of the response of various films within a given batch. Finally, the reproducibility of the PSD values was obtained from the standard deviation of various PSD measurements made with exactly the same CT scan settings resulting in an average value of 3.4 %. No uncertainty was assigned to the factors used to convert the air kerma values to absorbed dose to water. The relative combined uncertainty of the measured PSD values was determined by combining in quadrature all the associated uncertainties described above by following published guidelines⁽³⁶⁾, and is estimated to be of the order of 5.4 %. A difference of up to 18 % was observed in the calibration of films belonging to different film batches and so only the films from one single batch were used for all the measurements reported in this paper.

DISCUSSION

Measured PSD values varied between 27 (for the Siemens scanner) and 136 mGy (for the Toshiba scanner) when the routine head scan parameters were used. The reader is reminded that in this work a phantom was used to simulate the typical size of a patient head. However, since the PSD depends on the distance from the source to the skin of the patient, the construction of CTDI_vol vs. PSD lookup tables for other protocols involving other parts of the body must account for differences in patient size (since one phantom cannot represent all patients).

Much higher doses can be delivered outside routine protocol settings. For example, in the case of the Toshiba scanner, a PSD as high as 160 mGy can be delivered in a single rotation as seen in Figure 3. Therefore, in 20 rotations the threshold for erythema or temporary epilation (2–3 Gy⁽¹⁴⁾) could be reached. As an example, interventional and brain perfusion scans may require a large number of rotations around the same spot of the body. However, it must be noted that perfusion scans are typically carried out at a lower kV_p and mA s values, since high image quality is not critical in these studies. The application of the suggested dosimetry method to these high-dose applications is trivial, since the results for one rotation simply needs to be multiplied by the corresponding number of rotations.

Another example that requires special attention is the use of body scan protocols including tube current modulation. In this case, a relatively high PSD is expected on the flanks of the body, not on the top or the bottom part as described in this work. The corresponding PSD for the same CTDI can be up to 70 % higher (because typical tube currents of 380 mA s can be used for a CTDI equivalent to 200 effective mA s⁽³⁷⁾). Therefore, pieces of film should be fixed to the flanks of a phantom with the appropriate shape to examine these protocols. During the time the measurements presented in this work took place, a number of safeguards and guidelines were published to help prevent CT scanners from delivering excessive radiation. These include safeguards published by the National Electrical Manufacturers Association⁽³⁸⁾ and guidelines by the American Association of Physicists in Medicine (AAPM)^{(39, 40)}.

Traditional densitometers are reliable for PSD measurements because they are calibrated to NIST traceable units and they are widely available because they are used in screen-film radiology and radiotherapy departments. Therefore, a dosimetry method based on these devices allows for comparable measurements in all clinics. The dosimetry described in this paper using radiochromic film and a densitometer may be used also in X-ray radiography and fluoroscopy for quick quality control purposes.

The AAPM Report No. 204⁽⁴¹⁾ on size-specific dose estimates provides a complementary approach to relate readings of CTDI_vol to patient dose. In that report, the patient dose values are estimated from calculations. In this work, instead, peak doses are obtained from doses delivered by real CT devices using dosemeters placed on the surface of a phantom which acts as a surrogate for a real human being. After taking the measurements, a CTDI_vol value obtained with the same protocol can be directly linked to an actual value of peak dose. These measured values can be used to validate the calculations from Report No. 204 for specific scanners and protocols.

Note that the method suggested in this paper does not address any aspects of the image quality. It is assumed that the routine protocols are set to optimise dose and image quality. If a facility wants to compare the performance of different devices using this method, a simultaneous analysis of image quality is imperative. Another limitation of this study is the required calibration of the film, which needs to be done at NIST or at a secondary standard dosimetry lab. As mentioned above, a single batch of films is enough to perform the procedure presented in this work more than 50 times (for example 10 times in five different scanners).

The use of a long strip of film placed on the surface of a phantom and along the z axis of a CT scanner can directly provide information about any irregularities in the intensity distribution of the X-ray beam (such as the overlap of slices). The easy measurement with a portable densitometer can provide an immediate alert of an excessive skin dose, whereas most other types of radiation detectors require many measurements to find the region of maximum skin exposure or PSD.

CONCLUSION

An actual measured value of the dose delivered to the skin during a CT examination, the PSD, is provided and related to the CTDI_vol, a widely used and recognised quantity that is readily available in all CT units. Thus, the method and analysis presented here provides valuable information to patients, radiological technologists, medical physicists and physicians to relate the displayed CTDI_vol to an actual measured dose delivered to the skin of a patient. Apart from the film calibration, the analysis described in this work can be easily followed in any CT facility for their specific scanners and protocols. Look-up tables for examinations of different body parts can be developed using appropriate phantom sizes. The results, based on densitometer measurements, can be compared worldwide.

DISCLAIMER

Certain commercial equipment, instruments and materials are identified in this work in order to specify adequately the experimental procedure. Such identification does not imply recommendation nor endorsement by the authors' institutions, nor does it imply that the material or equipment identified is the best available for the purposes described in this work.

FUNDING

Kish Chakrabarti acknowledges funding from Critical Path Project of FDA, which enabled him to fund H.H. and C.B. for this work.