Organ doses evaluation for chest computed tomography procedures with TL dosimeters: Comparison with Monte Carlo simulations

Abstract Purpose To evaluate organ doses in routine and low‐dose chest computed tomography (CT) protocols using an experimental methodology. To compare experimental results with results obtained by the National Cancer Institute dosimetry system for CT (NCICT) organ dose calculator. To address the differences on organ dose measurements using tube current modulation (TCM) and fixed tube current protocols. Methods An experimental approach to evaluate organ doses in pediatric and adult anthropomorphic phantoms using thermoluminescent dosimeters (TLDs) was employed in this study. Several analyses were performed in order to establish the best way to achieve the main results in this investigation. The protocols used in this study were selected after an analysis of patient data collected from the Institute of Radiology of the School of Medicine of the University of São Paulo (InRad). The image quality was evaluated by a radiologist from this institution. Six chest adult protocols and four chest pediatric protocols were evaluated. Lung doses were evaluated for the adult phantom and lung and thyroid doses were evaluated for the pediatric phantom. The irradiations were performed using both a GE and a Philips CT scanner. Finally, organ doses measured with dosimeters were compared with Monte Carlo simulations performed with NCICT. Results After analyzing the data collected from all CT examinations performed during a period of 3 yr, the authors identified that adult and pediatric chest CT are among the most applied protocol in patients in that clinical institution, demonstrating the relevance on evaluating organ doses due to these examinations. With regards to the scan parameters adopted, the authors identified that using 80 kV instead of 120 kV for a pediatric chest routine CT, with TCM in both situations, can lead up to a 28.7% decrease on the absorbed dose. Moreover, in comparison to the standard adult protocol, which is performed with fixed mAs, TCM, and ultra low‐dose protocols resulted in dose reductions of up to 35.0% and 90.0%, respectively. Finally, the percent differences found between experimental and Monte Carlo simulated organ doses were within a 20% interval. Conclusions The results obtained in this study measured the impact on the absorbed dose in routine chest CT by changing several scan parameters while the image quality could be potentially preserved.

dose protocols resulted in dose reductions of up to 35.0% and 90.0%, respectively.
Finally, the percent differences found between experimental and Monte Carlo simulated organ doses were within a 20% interval.

Conclusions:
The results obtained in this study measured the impact on the absorbed dose in routine chest CT by changing several scan parameters while the image quality could be potentially preserved. substantially decreasing the need of exploratory surgery. 1 Since the development of the first CT equipment, this diagnostic imaging modality has been rapidly expanding, mainly due to the speed of image acquisition, and high-quality images. 2 Surveys such as the conducted in the United States in 1987 estimated that in 1980, only few years after its implementation, 2.2 million CT procedures were performed in general hospitals. 3 In 2007, it was estimated that more than 62 million CT procedures had been performed, from which at least 4 million were pediatric examinations. 4 Chest CT is one of the most common imaging examinations performed, accounting for approximately 16% of all CT procedures. 5 Notwithstanding, its utilization is increasing due to relatively recent efforts to implement low-dose chest CT for lung cancer screening in highrisk populations. As a consequence of the increasing number of CT examinations, the radiation dose absorbed by patients has become a concern among radiologists, researchers, and manufacturers. 4,6 Currently, CT utilization faces challenges related to justification of the procedure (i.e., benefits should outweigh potential risks) and dose optimization. 7,8 With the development of the CT technology, scanners have become more complex and efficient, challenging the accuracy of traditional dosimetry methods. 1 Although the computed tomography dose index (CTDI) and the dose length product (DLP) are well stablished metrics nowadays, these quantities only provide the information about how the machine was operated. 9 However, much more important and complex to assess is the information on the patient dose from any arbitrary examination. This information depends on a number of parameters, such as patient size and the anatomical region scanned. 10 Efforts have been made to develop robust methodologies to allow direct estimation of organ doses from patients undergoing CT exams. New ancillary metrics for CT dose quantification are being developed, such as the effective diameter and water-equivalent diameter, which are adopted to assess the size specific dose estimates (SSDE). 11,12 The correlation between the aforementioned quantity and organ doses is still under investigation. 13 Estimation of organ dose values is not a trivial task. In general, three approaches have been adopted over the past decades: (a) direct measurements with different kinds of dosimeters, anthropomorphic phantoms, and postmortem subjects, (b) calculations using Monte Carlo methods combined with computational human phantoms, and (c) biological dosimetry based on blood samples. 10 Several advantages and disadvantages can be discussed regarding each approach. Anthropomorphic phantoms for dosimetry, for instance, have been in use for more than 30 yr, and researches indicate the ongoing development of phantoms according to new CT technologies. 14 The use of postmortem subjects provides a wide range of different sizes and anatomies. However, they do not replace the use of phantoms. This technique is difficult to perform and dose measurement is limited to some points, thus it is difficult to measure the average dose to a given organ. 10  | 309 dependence in the energy range of radiology and radiotherapy. 16 Moreover, their small sizes provide accurate spatial localization of the doses inside the studied organs. These measurements were compared with dose estimates obtained with Monte Carlo simulations using National Cancer Institute dosimetry system for CT (NCICT a ), an organ dose calculator based on Monte Carlo radiation transport technique combined with a series of computational human phantoms. 17 In addition, this investigation also addresses the effects of tube current modulation (TCM) on organ dose in comparison with fixed tube current protocols, particularly in pediatric examinations in which TCM protocols have been recently applied for chest CT irradiations. However, its efficiency has been questioned for pediatric patient irradiations. 18,19 As the standard protocol for pediatric chest CT in InRad involves TCM, the effects on TCM on organ dose in comparison to protocols with fixed tube current were evaluated.

2.A | Thermoluminescent dosimeters
Lithium Fluoride doped with Magnesium and Titanium (LiF:Mg,Ti) thermoluminescent dosimeters (TLD), in the format of 3 × 3 × 1 mm 3 chips (TLD-100, Harshaw Chemical Company, OH, USA) were used in the present work. These TLD chips were processed by a Risϕ TL/OSL reader, model DA-20, (DTU Nutech. Inc., Roskilde, Denmark). During the reading process, the dosimeters were heated from room temperature to 350°C at a constant rate of 10°C/s, generating the LiF:Mg,Ti characteristic TL curve (photon counts against temperature). The socalled "TL value" was then obtained by numerically integrating the TL curve and the resulting quantity is directly proportional to the dose deposited by the radiation in the dosimeter. 20 In order to correlate the TL value to the Air Kerma (K Air ), calibration curves were constructed using both an RQT 9 X ray beam quality 21 generated by a Philips MCN 421 equipment (Philips, Germany) and a Philips Brilliance 64 CT scanner. 22 Two SSDL calibrated ion chambers (30 cc from PTW, Freiburg, Germany, and 0.6 cc from Radcal Corporation, Monrovia, CA, USA) were used to measure the air kerma. These calibration curves were adopted for the organ doses estimations with the anthropomorphic phantoms. Dosimeter holders were specially designed using polyoximethylene to accommodate up to 5 TLDs inside the drilled holes of the anthropomorphic phantoms. 25 Figure 1 shows two dosimeter holders together with TLDs and a centimeter scale for perspective.

2.C | CT scanners
The irradiations were performed using two different 64-slice CT scan-  26 The ULD protocol was designed as part of an ongoing investigation approved by the institutional review board to address the diagnostic information of CT scans with doses comparable to chest radiographs. 27 Other investigators have previously reported this practice for dose optimization. [28][29][30] Both LD and ULD protocols seek to reduce the dose by adjusting the scanner's tube current. LD tube current is set at 120 mA, with 48 mAs, whereas ULD is set at an even lower value of 40 mA, with 16 mAs, which represents a significant decrease compared to the value of 300 mA used for STD chest CT protocol.
F I G . 1. Thermoluminescent dosimeter holder, specially designed to be introduced into RANDO phantom internal holes, and the TLD chips placed beside a scale for perspective view.
Two phantom irradiations were performed to investigate the impact of TCM on lung dose reduction in the GE scanner. One irradiation consisted of longitudinal TCM ("Auto mA"), whereas the second irradiation consisted of both longitudinal and angular modulation combined ("Auto + Smart mA"). The angular modulation in GE scanners can only be selected in combination with longitudinal modulation. 31 Acquisition parameters of the studied protocols are presented in Table 1. Since scan projection radiographs (SPR) are often performed before TCM protocols, the imparted doses due to double SPRs were also evaluated.

2.D.2 | Pediatric protocols
Diagnostic pediatric chest CT were also surveyed using information obtained from the institutional PACS, using DICOM header metadata. The target protocol for this study was named "Chest for Children," which is the standard chest protocol for pediatric population.
In order to compare doses under different operating conditions, four variations of this protocol were assessed: two values of tube voltage were used (120 and 80 kV), and for each tube voltage, first a fixed mAs value was chosen and then longitudinal TCM was used. However, this approach differs from clinical practice, as TCM is always selected regardless of tube voltage for dose reduction. The acquisition parameters of the studied protocols are presented in Table 2.

2.E | TLD positioning
TLD groups were positioned inside the phantoms according to the thyroid and lung distributions 25,[32][33][34] (Tables 3 and 4). In every irradiation, one group of TLD was left outside the examination room in order to estimate the background radiation dose, which was subtracted from all TL values corresponding to the irradiations during data analysis. The placement of the groups inside each phantom is described below.

2.E.1 | Adult phantom
All adult chest irradiations were performed using 40 groups of three TLDs each distributed into the lungs of the adult phantom. The distribution of the groups within each slice of the phantom along with the lung tissue fraction is presented in Table 3. In Table 3, f i values correspond to the lung mass fraction contained inside each physical slice i of the phantom.
T A B L E 1 Acquisition parameters for the adult phantom irradiation using the GE CT scanner. The values for CTDI vol and DLP displayed by the scanner, relative to a 32 cm CTDI phantom, are also shown. were placed in the thyroid.

SPR
The distribution of the groups within each slice of the phantom along with the lung and thyroid tissue fraction is presented in Table 4. The determination of the fractions of the total lung mass (f i ) is described elsewhere. 32 In a typical chest CT procedure, the lungs are entirely irradiated and the thyroid is at least partially irradiated, according to the position of the patient on the couch. Since those are radiosensitive organs, 2 it is important to evaluate the radiation dose absorbed by these organs during such procedures. Thyroid doses evaluation is particularly relevant for pediatric patients due to their long life expectancy. Therefore, the pediatric phantom was irradiated from the middle of the neck through the lung bases and the resulting doses to the lungs and to the thyroid were evaluated.

2.F | Organ doses estimate
In order to convert the TL values into organ-absorbed doses, the following 4-step procedure was adopted: • The TL values were converted into K Air , using the calibration curve previously described (Section 2.A).
• For each phantom slice i, a mean value of K Air (K i Air ) is calculated, as shown in eq. (1). 25,35 where G is the total number of TLD groups accommodated into ith slice and σ 2 n is the variance of the TL values from TLDs in the nth group. Equation (1) assumes purely statistical uncertainties from each TLD Group, since each group is not affected by partial volume irradiations, and it represents the weighted mean of individual airkerma means calculated from each TLD group inserted in the ith slice. 36 • K i Air values were converted to organ average absorbed dose in the organ fraction present at ith slice, D i , according to 25,37,38 : where (μ/ρ) Organ and (μ/ρ) Air are the mass-energy absorption coefficients for the target organ and air 39 respectively, which vary according to the effective energy of the X ray beam ( Table 5). The determination of those values is described elsewhere. 25,32 • Last, the mean absorbed dose for the entire organ was estimated by summing up the contributions regarding each slice, where f i is the organ fraction contained in ith slice. 40,41 D The uncertainties on organ dose values were considered within a 68.3% interval (k = 1) and are described in Appendix A.

2.G | Comparison with NCICT
The results obtained with the experimental method proposed in this study were compared with the organ doses calculated by NCICT software. NCICT is based on a series of pediatric and adult computational human phantoms representing the reference individuals defined in the ICRP Publication 89 with several CT scanner models. 17,42,43 The program features a graphical user interface so that the user can introduce the scan parameters specific to each examination. 17 Moreover, the software comprises a batch module that enables the calculation of organ doses for a large number of patients and for a TCM protocol. 17 The organ dose calculated from the software has been extensively tested by measurements. 44,45 Comparison results are presented along with the percent differences between experimental (D exp ) and simulated (D sim ) values per organ, as follows:

2.H | Statistical evaluation
The agreement between experimental and simulated methods was quantified according to the Bland-Altman analysis. 46 This analysis is used to evaluate the mean differences between two different methods by estimating an agreement interval, in which 95% of these differences fall. 46,47 In this study, the percent differences between

3.A | CT acquisition protocols
The evaluation of the CT examinations conducted at InRad showed that more than 50 modalities of CT are performed annually. In 2016, a total of 95,000 patients were identified. About 5% of these patients were pediatric (0-15 yr old). The most frequently applied protocols for both adult and pediatric patients were identified (Fig. 2).

3.B.1 | Adult lung doses
The lung mean absorbed doses due to the Chest CT protocols previously described are summarized in Table 6, along with further dosimetric quantities (dose/mAs, dose/mAs eff , CTDI vol , and DLP values).

3.B.2 | Pediatric lung and thyroid doses
For the pediatric phantom, doses to the lungs and thyroid were evaluated. These organs were directly irradiated by the primary beam of the chest CT scan. Results are presented in Table 7, along with further dosimetric quantities (dose/mAs, dose/mAs eff , CTDI vol , and DLP values).

3.C | Comparison with NCICT
The experimental acquisition parameters for each phantom and CT  for any future sample, the differences between both methods should fall within this limit in about 95% of the trials. The upper limit of agreement is higher than the limit adopted in this study (20%): the highest difference found was (19.3 ± 0.8)% for the thyroid using 80 kV and TCM, which is in agreement with the 20% limit that has been adopted. Therefore, the results presented in both Table 8 and Bland-Altman plot of the percent differences against the mean of the organ doses obtained with the NCICT software and TLD measurements. The mean of the percent differences is presented in blue (8.9%) and the 95% limits of agreement are presented in the dashed lines.
T A B L E 9 Lung-absorbed doses due to the Standard, Low Dose, and Ultra Low-Dose chest CT protocols estimated by the present work (with TLD measurements) and by the methodology proposed by Huda and Sandison. 49 The relative difference was calculated as the percentage difference between the values estimated by both methodologies. In the study conducted by Mathews et al., 53 the authors evaluated the cancer risk in pediatric patients after their exposure to ionizing radiation from CT examinations. The cohort had examinations performed from 1985 to 2005 and, overall, cancer incidence was 24% higher for exposed people than for unexposed people. In particular, an increased incidence rate ratio (IRR) was reported for several types of cancer (e.g., digestive organs, melanoma, brain), including thyroid. The authors argue that even though modern CT scanners are likely to yield to lower radiation doses, it is essential to limit CT examinations to cases that present a clear clinical indication, particularly for pediatric patients. Table 11 shows the comparison of the absorbed organ doses when TCM modulation was turned on for both tube voltages (80 and 120 kV).

4.B.2 | TCM protocols
According to the results in Table 11, TCM can reduce the organ doses by 49.1 ± 3.3% in the pediatric phantom when setting a tube voltage of 80 kV and by 40.8 ± 2.9% when using 120 kV. In the clinical practice extracted from the data collected, the majority (>95%) of examinations were performed with 120 kV and TCM, while a few examinations were performed with 80 kV and TCM. From Table 11, switching the kV from 120 to 80 keeping the TCM in both cases would save up to 25.4 ± 1.9% of thyroid doses and up to 28.7 ± 1.8% of lung doses, maintaining the necessary image quality for diagnostic purpose. Therefore, a possibility of optimization was identified, which is in progress of implementation and validation.
In particular, it is essential to evaluate the image quality when aiming at protocol optimization. There are several studies reporting different tools to assess clinical image quality, 54-57 although on the other hand there are several studies showing that a radiologist tend to select images in which a given objective parameter (e.g., contrast resolution) is higher. 54 In the study proposed by Rehani 54  However, some studies report small increases in absorbed organ dose in pediatric subjects due to TCM. 18 In the study conducted by

4.C | Comparative evaluation with NCICT
Experimental and simulated results were in agreement within 20%.
Small differences are mainly related to anatomical difference between the computational human phantoms built in NCICT and the physical phantoms used for dose measurements. Despite such differences, the experimental methodology presented in this study showed to be adequate for dose evaluation.
T A B L E 1 0 Comparison among absorbed doses for thyroid and lungs when using fixed mA with 120 and 80 kV.

4.D | Clinical benefits and limitations
The clinical motivation for this study was the general evaluation of the practices related to CT procedures performed in a clinical institution. The experimental measurements were performed in order to have a more reliable estimate of the organ doses in such procedures.
A limitation of this study refers to the use of two sizes of anthropomorphic phantoms and two organs only. However, this is an accurate method that can be applied in a wide range of phantoms and

| CONCLUSIONS
An experimental approach was applied in this study to evaluate organ doses in anthropomorphic phantoms in different chest CT protocols.
This methodology has proven to be efficient for measurements of doses to organs within the scan regions but its applicability to different situations must be evaluated, especially when the organ is not directly irradiated by the primary CT beam. Nonetheless, because of the limita-

CONFLI CT OF INTEREST
The authors declare that they have no conflict of interest.

E N D N O T E
a NCICT program can be obtained from National Cancer Institute by contacting Dr. Choonsik Lee at http://ncidose.cancer.gov.

UNCERTAINTIES ESTIMATION
The uncertainties considered to calculate the overall uncertainties of lung-and thyroid-absorbed dose estimates are summarized in Fig. A1.
σ MQ is the uncertainty in the ionization chamber reading (in Coulombs) for a X ray beam quality Q, σ N k;Q 0 is the uncertainty of the calibration coefficient given by the IC calibration report, σ kQ;Q 0 refers to the correction factor for a radiation beam quality Q regarding the ionization chamber's calibration beam quality Q 0 , σ kTP is due to the correction factor for temperature and pressure, σ kAir is the composed uncertainty for air kerma values, σ n is the uncertainty of the TL values from the F I G . A1. Scheme illustrating all the uncertainties used to the overall uncertainty for lung-absorbed dose estimation.
different TLDs inside a measuring group, σ g is the systematic uncertainty regarding the TLD group selection (i.e., 6.5%), 25 σ a refers to the calibration curve of TL values and air kerma, σ f corresponds to uncertainties on the organ mass fraction inside each physical slice of the phantoms, 32 and finally σ DOrgan is the overall uncertainty to the organ dose estimates obtained with the propagation of all components. Confidence level considered is 68.3%. (k = 1).