Addressing Challenges in Diagnostic X-ray Dosimetry: Uncertainties and Corrections for Al2O3:C-based Optically Stimulated Luminescent Dosimeters

8 The use of Al 2 O 3 :C-based optically stimulated luminescent dosimeters (OSLDs) in diagnostic X-ray is a challenge 9 because of their energy dependence (ED) and variability of element sensitivity factors (ESFs). This study aims to 10 develop a method to determine ED and ESFs of Landauer nanoDot TM OSLDs for clinical X-ray and investigate 11 the uncertainties associated with ESFs. An area of 2 x 2 cm 2 at the central axis of the X-ray field was used to 12 establish the ESFs. A total of 80 OSLDs were categorized into “controlled” (n=40) and “less-controlled” groups 13 (n=40). The ESFs of the OSLDs were determined using an 80 kVp X-ray in free-air geometry. The OSLDs were 14 cross-calibrated with an ion chamber to establish the average calibration coefficient and ESFs. The OSLDs were 15 then irradiated at tube potentials ranging from 50 to 150 kVp to determine their ED. The uniformity of the X-ray 16 field was ± 1.5 % at 100 cm source-to-surface distance. The batch homogeneities of user-defined ESFs were 2.4 % 17 and 8.7 % for controlled and less-controlled OSLDs, respectively. The ED of OSLDs ranged from 1.125 to 0.812 18 as tube potential increased from 50 kVp to 150 kVp. The total uncertainty of OSLDs, without ED correction, 19 could be as high as 16 %. After applying ESF and ED correction, the total uncertainties were reduced to 6.3 % in 20 controlled OLSDs and 11.6 % in less-controlled ones. OSLDs corrected with user-defined ESF and ED can reduce 21 the uncertainty of dose measurements in diagnostic X-ray, particularly in managing less-controlled OSLDs.


Introduction
Optically stimulated luminescent dosimeters (OSLDs) are often used to monitor dose exposure in radiological procedures to ensure safety and the correct level of radiation delivery [1][2][3][4][5][6][7][8].These light-based detectors are now well-established and commercially used for personal dose monitoring, dose audits and in vivo dosimetry in radiotherapy [2-4, 6, 7, 9-17].The characteristics and radiation properties of various commercially available OSLDs have been well characterized and documented [10][11][12][13][18][19][20].The recent publication of the AAPM TG191 report provided a fairly comprehensive summary of works and guidelines on the use of thermoluminescent dosimeters (TLDs) and OSLDs in various clinical applications, as well as the various uncertainties associated with these dosimetry systems [9].The use of Al2O3:C-based OSLDs has gained little traction in kilovoltage (kV) energy photon beams, which are commonly applied in diagnostic imaging [8,19].One reason may probably be the huge uncertainty surrounding this dosimeter due to energy dependence (ED) issues and variable element sensitivity factors (ESF) associated with each detector [18,19,21,22].To our knowledge, no study has investigated the feasibility of establishing OSLD ESFs using diagnostic X-ray systems.This is because diagnostic X-ray fields are known for their inherent anode-heel effect, which produces a non-uniform field.In addition, although the OSLDs are reusable, their sensitivity has been reported to change with increased dose history [9,13,23].As it is not practical to monitor and track the dose history of each OSLD, it is thus crucial to understand the uncertainties in clinical dosimetry and diagnostic imaging so that general methods can be developed to resolve them.
Many uncertainties reported in the literature are based on methodologies that used megavoltage photon beams and high-dose radiotherapy applications [6, 7, 10-14, 16, 20, 23].There is still a paucity of methods to establish the sensitivity factors for diagnostic X-rays, and the uncertainties associated with the energy and lowdose ranges typically encountered in diagnostic imaging modalities.There are several advantages to characterizing and establishing the ED of OSLDs in diagnostic X-rays, particularly when there is a lack of access to linac and Co-60 irradiation facilities.Furthermore, the characterization of ED in the OSLDs may be more accurate within the region of beam energy application.Reft [21] reported that ED could be 3.5 times higher at diagnostic energy beams compared with megavoltage energy beams.
The procurement of OSLDs in different batches will entail different manufacturer-specified sensitivities.
Over time, the OSLDs may have also been used for various dose measurements in megavoltage (high-dose radiotherapy) or kilovoltage (low-dose diagnostic imaging) photon beams.Therefore, tracking the radiation dose history of each OSLD may be impractical.Thus, in this paper, we seek to answer the following questions: (1) Is establishing the ESFs of OSLDs using a diagnostic X-ray beam feasible?
(2) How reproducible is the ESFs established using this method?
(3) What is the ED of OSLDs in a typical diagnostic X-ray?(4) What are the uncertainties expected in using OSLDs with long or unknown radiation history compared with controlled OSLDs?
To address these challenges, we present a method to determine the ESF and ED of OSLDs using clinical diagnostic X-ray and investigate the uncertainties associated with variable radiation history.

Procurement of OSLDs
A total of 97 nanoDot™ (Landuer Inc., Glenwood, IL, USA) OSLDs in light plastic disks with a volume of 1.0 × 1.0 × 0.1 cm 3 each were used.The Al2O3:C crystals were contained in approximately 0.2 cm-thick plastic packaging with a diameter of 0.4 cm.The nanoDot™ dosimeters were connected to the MicroStar mobile dosimeter reader (Landuer Inc., Glenwood, IL, USA) to obtain and record the radiation data.
The total OSLDs were divided into two groups.The first group comprised 50 "screened" OSLDs that were newly acquired and used in a controlled manner, and all having the same manufacturer-specified ESF.These OSLDs had only been used for low-dose irradiation in diagnostic X-rays.The second group comprised 47 "lesscontrolled" OSLDs with manufacturer-specified ESFs varying from 0.80 to 1.09.The less-controlled OSLDs were acquired at different times and batches, comprising a mix of "screened" and "unscreened" detectors.They had been exposed to different radiation dose histories, including high-dose and megavoltage photon beams.These two groups of OSLDs represent a "controlled" and "less-controlled" set used to investigate the expected lowest and highest uncertainties.
The calibration of the OSLDs, estimation of ESF and ED were carried out using a Toshiba KXO-50S general X-ray unit (Toshiba Medical Systems Corporation, Tochigi, Japan).The Radcal 10X6-6 general-purpose ion chamber (Radcal Corporation, Monrovia, CA, USA) was used to calibrate the OSLD dose.The measurements were carried out at room temperature ranging from 20 °C to 22 °C and an atmospheric pressure of 999.9 to 1002.0 mBar.

Handling protocol of OSLDs
A total of 40 OSLDs from the controlled and less-controlled groups were selected and identified for irradiation.Seventeen were kept as control/constancy OSLDs.They were, however, always read out and bleached together with the rest of the OSLDs.All the OSLDs were bleached and pre-read to obtain the residual dose/counts before being irradiated.After irradiation, the OSLDs were rested for 24 hours before being read out using the MicroStar reader before being optically bleached for 24 to 48 hours.The optical bleaching was performed using a white LED light box.Precaution was taken to ensure that the bleached OSLD readouts were <0.1 mGy before reuse so that the minimal residual doses were retained.Three readings were taken from each OSLD during the readout process, and the mean counts were obtained.The mean count from pre-irradiated OSLDs was subtracted from irradiated OSLDs to obtain the net count.As the MicroStar reader was calibrated by the manufacturer and the process was traceable to a primary dosimetry laboratory, only the count values were used in this study.The manufacturer-specified calibration coefficient and sensitivity factors were not used in the air kerma computation.

Determination of X-ray field uniformity
Existing studies had reported methods to establish ESFs using megavoltage X-ray with a linac.However, no studies had reported using clinical diagnostic X-ray units, likely because of large variations in field uniformity due to the anode heel effect.However, we postulated that this was possible if one could identify the region where the field uniformity was acceptable.In a uniform field, it would be possible to ensure the consistency and reproducibility of X-rays delivered to a group of OSLDs, which was essential in determining the ESF.
The uniformity of a 30  30 cm 2 X-ray field was determined using an Unfors Xi R/F & MAM (Unfors Raysafe AB, Billdal, Sweden) solid-state detector at selected equidistance intervals from the central axis.The air kerma was measured at equidistance from the central axis to ±12.5 cm outward, in the direction parallel and perpendicular to the anode-cathode direction.The measurements were normalized to one at the center axis.The X-ray field uniformity was then determined for selected field sizes.

Determining the element sensitivity factors
The field size of 30 x 30 cm 2 was preferred in this experiment because both ion chamber and OSLDs could be positioned in the irradiation field without intrusively affecting OSLD measurements.Figure 1 shows a customized jig made of polystyrene board with specific cut-outs to fit the Radcal ion chamber that was positioned 5.5 cm away from the OSLDs.The replacement method was used when calibrating the OSLDs with the standard reference chamber by shifting the central axis of the X-ray laterally to the center of the specific dosimeters used.There were two methods to establish the OSLD calibration coefficient and ESFs; the individual detector method and the batch method [9].

Individual detector method:
The calibration coefficient (Cali), of individual OSLDs was determined by taking the ratio of the raw counts for the respective OSLDs and dividing them by the air kerma (AKcal ) reading measured by the ion chamber in Equation 1.
Where, the i is the ith OSLD and the unit is count/mGy.

Radcal nanoDots
The individual Cali can also be converted to ESF within a batch of OSLDs by taking the ratio of the individual calibration coefficient to the average calibration coefficient,  ̅̅̅̅̅ of the batch of OSLDs using Equation 3. 3)

Batch method:
For the batch calibration method, the average calibration coefficient () ̅̅̅̅̅̅ for the group was taken as the ratio of the average count of all OSLDs to AKcal, as shown in Equation 4. 4) where the unit for Cal is count/mGy.
The ESF of each OSLD were determined using Equation 5, 5) To correct for ESF, the raw counts corrected for the  ̅̅̅̅̅ and the individual ESF, as shown in Equation 6. 6) Note that the value derived from Equation 3 will be almost the same as those derived from Equation 5.
The ESFs were calibrated based on air kerma because of two reasons.First, the free-air geometry setup eliminated unwanted backscattered radiation to the OSLDs, ensuring a more accurate ESF estimation.Second, this setup could also be applied if the ion chamber was replaced with solid-state detectors with lead backing.These detectors were commonly used in many radiological facilities.

Reproducibility of ESF
We evaluated the ESF's reproducibility by repeating the irradiation at 80kVp in a 30 x 30 cm 2 X-ray field size at 100 cm SSD.A total of 20 OSLDs from each group was irradiated with a standard exposure in free-air geometry and another set of 20 OSLDs were irradiated on a 15 cm water phantom.For each exposure, only four OSLDs were placed within the central 2 x 2 cm 2 area.The setup variation in free-air geometry and on water phantom was also used to verify the impact of increased scattered radiation on ESF determination.The ESFs from the second setup (ESF2) were compared with those obtained from the first measurement (ESF1).

Determining ED in diagnostic imaging range
The same experimental setup used for ESF determination was also used to determine the ED of OSLDs.
However, the tube potentials were varied between 50 and 150 kVp.The half-value layer (HVL) of each tube potential was measured using the Unfors Xi R/F & MAM detector and equivalent beam energies were calculated using the measured HVLs and the National Institute of Standards and Technology (NIST) physics database [24].
The net counts obtained from the OSLDs were corrected for their individual ESF.The dose-response of OSLDs was taken as the ratio of the corrected air kerma measured by OSLDs to the standard reference ion chamber (IC), normalized to one using an 80 kVp beam (Equation 7).7) where, EDkVp is the ED of the specific tube potential.
This could be specified in HVLs as well.AKm,I,kVp is the air kerma measured by the OSLDs, corrected for ESF, for a specific tube potential.ICkVp is the air kerma measured using the standard reference chamber for the specific tube potential.AKm,I,80Vp is the air kerma measured by the OSLDs, corrected for ESF, using an 80 kVp beam.The IC80kVp is the air kerma measured by the standard reference chamber for 80 kVp beam.In some literature, the ED was also reported as the ED correction factor, which is the inverse of the ED described here [18,19].
The ESF and ED corrections were applied using Equation 8.
where, the AKESF,kVp is the air kerma measured by the OSLDs, corrected for ESF and ED.
The ED measurement was repeated once by shuffling the OSLDs used for each tube potential.The reproducibility of the OSLD ED response was evaluated for both controlled and less controlled OSLDs.

Measuring dose linearity of OSLDs
The dose linearity of the OSLDs for 60, 80, 100 and 140 kVp X-rays with doses ranging from 0.5 to 22 mGy were determined using the Toshiba KXO-50S general X-ray unit.This dose range covered the typical diagnostic X-ray examination range.The OSLDs were irradiated in free-air geometry at 100 cm SSD, at the central axis of a 30 x 30 cm 2 field size.The Radcal ion chamber was used to determine the air kerma.The OSLDs' net counts were corrected for ESF and ED, and plotted against the Radcal ion chamber measured air kerma values.

Statistical analysis
The data were analyzed using IBM SPSS version 22 (IBM Corp, Armonk, NY, USA).A paired t-test was used to compare the user-defined ESFs and the manufacturer-specified ESFs.Pearson correlation was used to determine the correlation between the two ESFs.

Uncertainty budget determination
The uncertainties identified in this study included the repeatability of the MicroStar readout, calibration of standard dosimeters, X-ray beam uniformity and consistency, dose linearity, ESF variations and ED of OSLDs.
The uncertainty of the standard reference dosimeter was the calibration uncertainty obtained from the calibration certificate.The consistency of the X-ray beam was taken as 1SD of the mean of five exposures using the same exposure parameters.The repeatability of the MicroStar reader was obtained by taking each OSLDs readout three times to obtain the mean and 1 SD value.The repeatability of the MicroStar reader for three sequential readouts were determined for the dose level of 3 -4 mGy and at the level of the residual dose level, post-bleaching.The total uncertainty was taken as all the individual uncertainties added up in quadrature.

X-ray field uniformity
Figure 2: X-ray field uniformity of the Toshiba KXO-50S general X-ray unit Figure 2 shows the X-ray field uniformity of the Toshiba KXO-50S general X-ray unit for the 25 x 25 cm 2 field size within a 30 x 30 cm 2 X-ray field.Measurement at the edge of the field was avoided due to penumbra.
There was a pronounced anode heel effect in the direction parallel to the anode-cathode direction, while in the direction perpendicular to the anode-cathode, the dose was reasonably uniform.In the anode-cathode direction, the field uniformity changed from 0.74 to 1.07 due to the anode-heel effect.While in the direction perpendicular to the anode-cathode, the field uniformity was reduced from 0.98 to 0.95 at ±12.5 cm away from the central axis.
However, within the central 2 x 2 cm 2 field, the uniformity was better than ±1.5 %.Thus, it was decided to use the central 2 x 2 cm 2 area for subsequent calibration as well as the ESF and ED determination of the OSLDs.
The large non-uniformed X-ray field in the anode-cathode direction makes it impossible to use a large square area for ESF calibration, a method typically used to determine the ESF under linac beams.However, it was still possible to use a narrow rectangular area at the center of the X-ray field perpendicular to the anode-cathode direction for this purpose.Based on the measurements, the 2 x 10 cm 2 area still provided uniformity of ≤1 %.

ESF determination
The average calibration coefficient established for 40 OSLDs using the Radcal ion chamber ( ̅̅̅̅̅ ) was 3023 ± 83 counts/mGy and 3006 ± 262 counts/mGy for controlled and less-controlled OSLDs, respectively.The batch homogeneity was 2.8 % for the controlled group and 8.8 % for the less-controlled group.The histograms in Figure 3 show a broader distribution of sensitivity in less-controlled OSLDs compared with controlled ones.

Reproducibility of ESF
There were slight deviations between ESF1 and ESF2.The mean absolute differences for the free-in-air setup were 1.7 ± 1.3 % and 2.8 ± 2.0 % for the controlled and less-controlled OSLDs, respectively, whereas, for the on-water phantom setup, the mean absolute differences were 1.2 ± 1.1 % and 2.3 ± 1.6 % (Table 1).Compared with the manufacturer-specified sensitivity, the user-defined ESFs were more reproducible.

Energy dependence
Figure 5 shows the OSLD dose response as a function of HVL.Four sets of data show the repeated ED measurements for the controlled and less-controlled OSLDs.The average of the dose-response was normalized to 1 at 80 kVp beam (HVL = 3.2 mmAl, 35.5 keV equivalent energy) and tabulated in Table 2.A strong agreement between the controlled and less-controlled datasets indicated that the ED was reproducible.The Al2O3:C-based OSLDs over-responded by 12.5 % at 50 kVp beam and under-responded by 18.8 % at 150 kVp.The ED of the OSLD, as a function of HVLs, was fitted with a third-degree polynomial as in Equation 9.
ED = 0.0003x 3 + 0.0123x 2 -0.1663x + 1.4246 (Equation 9) where, ED is energy dependence and (x) is the HVL in mm Al.Using the ED correction factors in Table 2, the OSLDs were corrected for their ED and compared with the Radcal ion chamber measurements (Figure 6).The OSLD-measured air kerma was corrected for ESF and ED, which resulted in measurements closer to the ion chamber readings.The absolute mean differences between ESF and ED-corrected OSLD readings with the Radcal ion chamber were 1.2 ± 0.6 % and 0.5 ± 0.4 % for the controlled and less-controlled groups, respectively.
Compared with the MicroStar reader direct-dose readings, the ion chamber's readings were significantly different, and the distribution was much broader.The absolute mean differences between the MicroStar directdose readings with the Radcal ion chamber were 8.3 ± 6.0 % and 12.0 ± 8.7 % for the controlled and less controlled groups, respectively.The ESF and ED correction reduced the variation in OSLD measurements, particularly for the less-controlled group.The average standard deviation of the four OSLD readings was reduced from 6.4 ± 3.1 % to 2.2 ± 1.0 %.However, in the controlled group, the variation was almost similar between the MicroStar directdose readings (2.0 ± 0.7 %) and ESF and ED corrected readings (2.8 ± 0.7 %). Figure 8 shows the counts corrected for ESF and ED, with different positively linear lines converging towards zero.The linearity curves for 80 kVp and 140 kVp were measured using controlled OSLDs, while the linearity curves for the 60 kVp and 100 kVp were measured using the less-controlled OSLDs.This might explain why there appeared to be larger error bars in the 60 kVp curve.The excellent agreement between linearity curves for different tube potentials proved that the ESF and ED corrections were accurately applied.

Uncertainty budget
Table 3 shows the uncertainty budget to establish the OSLD's ESFs using a diagnostic X-ray beam.The maximum uncertainty of three sequential readouts using the MicroStar reader was 2.8 %.The repeatability deteriorates with lower doses.The OSLDs' reading at post-optical bleach was 100 ± 11 (11 %) counts and 151 ± 40 (27 %) counts for controlled and less-controlled OSLDs, respectively.At such low doses, slight fluctuations in the readouts appear to be unrealistically magnified.Therefore, in the uncertainty budget, we considered the uncertainty at the typical dose encountered during radiological examinations to be more relevant.The X-ray constancy was taken at 1 SD of the mean of five air kerma measurements using the Radcal ion chamber.We obtained the uncertainty of the calibration of the standard dosimeters from their calibration certificates.The uncertainty associated with batch homogeneity included inherent uncertainty due to repeatability of the MicroStar reader.Thus, we extract the uncertainty associated solely with the batch homogeneity using Equation 10.
where  ℎ ℎ is the uncertainty associated solely with ESF variability,  ℎ ℎ, is the total uncertainty comprising both the MicroStar repeatability and ESF variation, and   is the MicroStar repeatability uncertainty.
Without applying the appropriate ED correction, the uncertainties were 12.6 % and 16.0 % for controlled and less-controlled OSLDs.With ED correction, the total uncertainties were reduced to 6.6 % and 11.9 % for controlled and less-controlled OSLDs.

X-ray field uniformity
The X-ray field uniformity of the clinical diagnostic X-ray unit was suitable to determine the ESF of OSLDs, provided the measurements were made within 2 x 2 cm 2 at the center of the X-ray field.It is possible to extend the usable flat field area of the X-ray field by placing OSLDs along the perpendicular direction of the X-ray at the center axis.

ESF correction methods
Two methods were proposed to correct the sensitivity variation of OSLDs.We will discuss the advantage and disadvantages of each methodology here.The individual detector calibration method provided the individual calibration coefficient (  ), which could be treated as a singular correction for the OSLD response to the ion chamber used to establish this calibration.The   was specific to the standard reference chamber used.This meant that if the reference chamber was interchanged with another detector, such as another ion chamber or Unfors detector, the   value might also change.The OSLD-measured dose could be traceable to the specific standard reference chamber and not others.On the other hand, the use of   allowed each OSLD to be used independently from the batch that it was calibrated on, as long as the standard reference chamber and the MicroStar reader used in the OSLD readout remained the same.
We postulated that the calibration coefficients of the OSLDs might depend on the reference chamber used.With the batch calibration method, the individual calibration coefficients were transformed into a fractional representation in the form of individual ESFs (ESFi).However, the site-specific average calibration coefficient ( ̅̅̅̅̅  ) needed to be established if the standard reference chamber or the MicroStar reader was different.When the MicroStar reader readout the OSLD counts, the readings had an inherent sensitivity variation, i.e. the ESF.
Thus, one needs to remove the effect of the inherent sensitivity variation on the  ̅̅̅̅̅ using Equation 11.
̅̅̅̅̅   = (Equation 11) where,  ,  is the net count measured at the site. ,  is the air kerma measured by the different detector at the site.n is the number of OSLDs simultaneously irradiated at the same time.The average of the individual calibration coefficient was taken as the  ̅̅̅̅̅   .
The ESFs were transferrable with different standard reference chambers or MicroStar readers.The  ̅̅̅̅̅  could be established by performing a simple 80 kVp beam exposure at a standard setup detailed in the methodology, and the net counts of the OSLDs corrected for ESFi.However, this method required the OSLDs to be used in the batch they were calibrated together.Inter-mixing different batches of OSLDs might introduce additional uncertainty [9].
The ESFs showed a wider spread in less-controlled OSLDs compared with controlled ones.Although the controlled group comprised of screened OSLDs with exactly the same manufacturer-specified sensitivity of 0.87, there was still a variation in the OSLD sensitivity that resembled a Gaussian distribution.However, this might be due to the uncertainty brought on by the repeatability of the MicroStar reader, which in the case of our study, had a maximum value of 2.8 %.This might be due to the design flaw of the MicroStar reader, which had an optical port of 4 mm in diameter, identical to the diameter of the optical port of the plastic case of the nanoDot OSLD [25].Several studies had reported higher maximum values, such as Jursinic (2020) who reported a 3.5 % uncertainty due to this design flaw.Al-Senan & Hatab (2011) reported values ranging from 2.9 % to 3.6 % [18].
Although the user-established ESFs were correlated to the manufacturer-specific sensitivity, they were nonetheless significantly different.The effect was more obvious in the less-controlled OSLD group.The different radiation history of OSLDs would also change the sensitivity over time and, thus, would require periodic reestablishment of the ESFs [23].

Reproducibility of the ESFs
The maximum absolute differences in ESFs were 4.8 % in controlled OSLDs and 7.1 % in less-controlled ones.It is practically impossible to ensure exact replication of the ESFs due to the uncertainty of the MicroStar repeatability.Within the measurement uncertainty, the ESFs established using the free-in-air and on-water phantom setup were in close agreement.The AAPM TG191 recommended that for newly acquired OSLDs, selected OSLDs should be tested to verify that the manufacturer-specified sensitivity was within 5 % [9].Thus, our results appeared to concur with the magnitude of uncertainty expected for OSLD ESFs, either obtained empirically or specified by the manufacturer.

ED correction
The ED of the OSLD for the controlled and less controlled groups were similar and not affected by the sensitivity difference and the radiation dose history of the OSLDs.This work found an ED variation of 31.3 % for the equivalent energy ranging from 30.2 keV to 44.1 keV.This is quite comparable with the results reported by Al-Senan and Hatab (2011) at 36 % ED [18].
The ED model based on HVL was introduced to enable easy determination of ED value, without having to convert HVL to equivalent energy.This was also because the HVL of diagnostic X-ray beams might differ slightly for different tube potentials and X-ray units.Relying solely on the tube potential to determine the ED correction may be inaccurate.There were also suggestions to use the combination of tube potential and HVL, known as the quality index, to indicate beam quality [26].

Dose linearity
The linearity of OSLDs was excellent in the dose range typically encountered in diagnostic radiology.
The less-controlled OSLDs showed larger uncertainties within the four OSLDs measuring the same dose.We note that the calibration coefficient derived from the dose linearity was in good agreement with those established using a fixed exposure for the batch OSLD calibration.We propose it as a suitable average calibration coefficient for batch calibration.The calibration coefficient derived from dose linearity was also postulated to be sensitive to the standard reference dosimeter used to derive it.Thus, the calibration coefficient might produce inaccurate conversions if the corrected OSLD measurements were compared with a different reference dosimeter.

Uncertainty
Compared with the AAPM TG-191 report on uncertainties of OSLD performance under high-efficiency applications, the uncertainties observed in this study were larger.They reported uncertainties (k=1) of 3.5 % and 4.9 % for controlled and less-controlled OSLDs, respectively [9].This could be because the reported uncertainties were primarily based on studies done in megavoltage X-ray and at higher dose levels typically used in radiotherapy applications.The slight fluctuations in OSLD readout did not significantly change the counts.Furthermore, ED was not a concern in the megavoltage energy range, where the Compton scattering was the main interaction.
In contrast, when using OSLDs in diagnostic imaging applications, the dose levels encountered were typically in the mili-Gray range.Thus, slight variations in the readout process, which could be introduced by the repeatability of the MicroStar reader, could have a more significant impact on the dose measured.Furthermore, in the kV energy range, the ED of the OSLDs was also substantial and should not be neglected.

Conclusion
We found that it was possible to establish the ESFs of OSLDs using a diagnostic X-ray beam.The ESFs were reproducible to 4.8 % and 7.1 % in controlled and less-controlled OSLDs, respectively.The Al2O3:C-based OSLDs over-responded by 12.5 % at 50 kVp tube potential and under-responded by 18.8 % at 150 kVp.Selected uncertainties associated with this methodology were determined.If the ESFs and ED were not corrected, the total uncertainty would be as high as 16 % in less-controlled OSLDs.Our approach allowed for more accurate characterization of OSLD ED in the region of beam energies encountered in diagnostic imaging.It provided a means to correct ESF variation using clinical diagnostic X-ray units.This methodology could be useful at centers where access to a radiotherapy linac or Co-60 irradiation facilities is limited or unavailable.

Figure 1 :
Figure 1: Customized detector holder made of polystyrene.

Figure 3 :
Figure 3: Histograms of ESF for controlled and less-controlled OSLDs

Figure 4
Figure 4 shows the scatter plot of ESF established at 80 kVp beam compared with the manufacturer's ESF.Although the manufacturer-specified sensitivity was significantly different from user-established ESFs, a significant correlation was found between the ESF established at 80 kVp beam with the manufacturer-specified ESF for the OSLDs (r = 0.728, p<0.001).

Figure 4 :
Figure 4: Scatter plot showing the distribution of the ESF established at 80 kVp beam compared with the manufacturer-specified sensitivity.The diagonal line is the line of identity.

Figure 5 :
Figure 5: The ED of OSLDs normalized to 1 at 80 kVp beam, equivalent energy of 35.5 keV and HVL of 3.2 mmAl.The error bars represent 1SD of the mean of four sets of measurements taken on controlled and lesscontrolled OSLDs

Figure 6 :Figure 7
Figure 6: Boxplots showing the distribution of four OSLDs, where the readouts are directly obtained using the MicroStar reader and, after applying the ESF and ED corrections, compared with the ion chamber readings

Figure 7 :
Figure 7: Dose linearity of the OSLD counts as a function of air kerma.The OSLD counts were corrected for ESFs of individual OSLDs.The linear models describing the relationship are to intercept at 0. The error bars are 1SD of the mean of four OSLDs.

Figure 8 :
Figure 8: Dose linearity of the OSLD counts as a function of air kerma.The OSLD counts were corrected for the ESF and ED.The linear models describing the relationship are set to intercept at 0. The error bars are 1 SD of the mean of four OSLDs.

Table 1 :
Reproducibility of the ESF, established using different setups *ESF1 was established using free-in air; ESF2 was established using either free-in air or on a water phantom; N is the number of OSLDs

Table 2 :
The ED of OSLDs at various tube potentials, half-value layers and equivalent energies

Table 3 :
Uncertainty budget of OSLDs