Computed tomography (CT) has long been a powerful tool in the diagnosis of disease, identification of tumors and guidance of interventional procedures. With CT examinations comes the concern of radiation exposure and the associated risks. In order to properly understand those risks on a patient-specific level, organ dose must be quantified for each CT scan. Some of the most widely used organ dose estimates are derived from fixed tube current (FTC) scans of a standard sized idealized patient model. However, in current clinical practice, patient size varies from neonates weighing just a few kg to morbidly obese patients weighing over 200 kg, and nearly all CT exams are performed with tube current modulation (TCM), a scanning technique that adjusts scanner output according to changes in patient attenuation. Methods to account for TCM in CT organ dose estimates have been previously demonstrated, but these methods are limited in scope and/or restricted to idealized TCM profiles that are not based on physical observations and not scanner specific (e.g. don’t account for tube limits, scanner-specific effects, etc.).

The goal of this work was to develop methods to estimate organ doses to patients undergoing CT scans that take into account both the patient size as well as the effects of TCM. This work started with the development and validation of methods to estimate scanner-specific TCM schemes for any voxelized patient model. An approach was developed to generate estimated TCM schemes that match actual TCM schemes that would have been acquired on the scanner for any patient model. Using this approach, TCM schemes were then generated for a variety of body CT protocols for a set of reference voxelized phantoms for which TCM information does not currently exist. These are whole body patient models representing a variety of sizes, ages and genders that have all radiosensitive organs identified. TCM schemes for these models facilitated Monte Carlo-based estimates of fully-, partially- and indirectly-irradiated organ dose from TCM CT exams. By accounting for the effects of patient size in the organ dose estimates, a comprehensive set of patient-specific dose estimates from TCM CT exams was developed. These patient-specific organ dose estimates from TCM CT exams will provide a more complete understanding of the dose impact and risks associated with modern body CT scanning protocols.

Computed Tomography (CT) has been used for medical diagnosis for the past four decades and has made significant contributions to patient healthcare by providing fast and accurate diagnostic information. Besides the extraordinary medical benefits it has brought to society, it delivers radiation dose to the patients, which can be potentially hazardous. Therefore, it has been a significant interest in both scientific research and clinical practice to reduce radiation dose to the patients during CT scans, while still maintaining the diagnostic performance, so that the information provided through this procedure is not compromised and appropriate medical determinations can be made at the minimum cost.

In this research work, a Monte Carlo based simulation package was used to estimate radiation dose to individual radiosensitive organs of patients with a range of body habitus. This package is exam and protocol specific, and it takes into account technical details of CT scanners such as spectra, bowtie filtration, and beam geometry. Modifications were made to the Monte Carlo simulation package to perform detailed radiation dose assessments for both patients and phantoms. These include the estimate of radiation dose to individual organs, the peak radiation dose to a wide spread tissue (such as peak skin dose), and surface dose distribution in a complex CT irradiation environment. Meanwhile, the effect of a variety of traditional dose reduction methods, such as tilting the gantry in brain perfusion scans, was also investigated.

In addition to the traditional dose reduction techniques that are already being utilized in the clinic, an innovative method to reduce organ dose while maintaining image quality was investigated. The distribution of radiation dose within the scan volume was demonstrated to be dependent on the Tube Start Angle (TSA). A change of TSA can cause a shift of dose distribution along the longitudinal axis. This results in variations in the measurement of surface dose during helical scans. This dose variation along the longitudinal direction for patients in CT imaging inspired a novel innovation to reduce organ dose while maintaining image quality by adjusting the TSA and table height in CT exams. Monte Carlo simulations were performed to demonstrate the effectiveness of this method for different patients under various scenarios, including conventional fixed tube current CT scans, and tube current modulated (TCM) scans. Besides the dose benefit this new method brings, its effect on image quality was investigated and demonstrated that there was no significant compromise on the image quality.

Despite the efforts to reduce radiation dose while maintaining image quality, the ultimate tradeoff in the goal of maximizing the benefit to risk ratio in CT examinations is the tradeoff between radiation dose and diagnostic outcome. As radiation dose is decreased, the image quality may be degraded. However, the diagnostic outcome does not necessarily have to be compromised. In other words, the image quality used for specific CT clinical tasks today may have room to be degraded and still be able to maintain accuracy of diagnostic outcomes. In order to investigate this tradeoff between radiation dose and diagnostic outcome for a specific clinical task (appendicitis was selected in this dissertation), a preliminary observer study was conducted to determine the difference of diagnostic performance at various dose levels. Images at reduced radiation dose levels were simulated by adding noise to the projection data using a calibrated method. These methods were employed for a group of patients with right lower quadrant pain who were scanned because of a suspected appendicitis. The results of Receiver Operation Characteristic (ROC) analysis suggested that there was no significant difference between radiation dose levels of 100%, 70% and 50%. Detailed analysis of patient organ (liver) dose demonstrated that the diagnostic performance is nearly perfect when the liver dose is higher than 10mGy. The interrelationship between a simple image quality metric (noise), organ dose, and patient size was also investigated.

In summary, this work assessed dose reduction tools available today that do not affect image quality, proposed a new method to reduce organ dose while maintaining image quality, and evaluated the method to reduce radiation dose, which affects image quality but could maintain diagnostic outcome by investigating the tradeoff between radiation dose and diagnostic outcome, as well as their correlation with image quality metrics (noise).

Computed Tomography (CT) has been one of the leading imaging modalities in today's practice of Radiology. Since its introduction in 1970s, its unique tomographic capability has not only prevented countless number of unnecessary surgeries but also saved lives by early detection of disease. Radiation dose from CT has been estimated to contribute to almost 50% of all medical radiation exposures. Concerns about radiation-induced carcinogenesis have resulted in efforts that encourage monitoring and reporting radiation dose from CT examinations. It has been suggested that the most appropriate quantity for assessing risk of carcinogenesis from x-ray imaging procedures is the radiation dose to individual patients. Currently employed dose metrics used to report patient dose are CTDI_vol and DLP, neither of which is patient-specific dose, let alone dose to individual organs.

CTDI_vol is dose to a homogenous cylindrical phantom, which is defined for fixed tube current CT exams. With the implementation of Tube Current Modulation (TCM) feature in almost all clinical CT protocols as an intended means for dose reduction, while maintaining an appropriate diagnostic image quality, CTDI_vol definition was standardized across scanners to reflect dose to CTDI phantom based on the average tube current across the entire scan length. Depending on the type of CT exam, the average tube current used to report a CTDI_vol value may or may not represent the actual tube current at a specific table location. In addition to not taking into account variation of the tube current across a single exam, CTDI_vol is size-independent, i.e. patients with different sizes have the same CTDI_vol value if scanned using the same imaging parameters.

To adjust CTDI_vol for size, AAPM Task Group 204 was established and subsequently published a report containing conversions as a function of effective diameter which can be applied to scanner-reported CTDI_volto adjust for patient size. However, the generated conversion factors were based on fixe tube current measurements and Monte Carlo simulations and failed to take into account TCM. Additionally, the size metric used in TG 204 was entirely based on patients' physical dimensions and does not take into account variations in composition and density among patients, let alone within a single patient; i.e. differences between chest and abdomen in terms of attenuation properties could not be explained with a simple measure of dimension such as effective diameter. Instead attenuation-based metrics need to be implemented to explain these differences.

The overall purpose of this dissertation was to improve organ dose estimation from Computed Tomography exams by: (a) taking into account the commonly used feature in CT protocols, Tube Current Modulation (TCM), (b) employing a more appropriate way of reporting CTDI for TCM exams and (c) using a patient size descriptor capable of describing the attenuation properties of individual patients.

For this dissertation a validated Monte Carlo based MDCT model capable of simulating organ dose was utilized to estimate organ dose to voxelized patient models undergoing tube current modulated CT examinations. Both detailed TCM and z-axis-only modulation information were used in the simulations in case raw projection data was not accessible. In addition to simulated organ doses different CTDI_vol values based on the type of patient model, abdomen versus chest, were calculated. These CTDI_vol values included regional CTDI_vol,regional and organ-specific CTDI_vol,organ along with scanner-reported CTDI_vol, referred to as global CTDI_vol,global. Furthermore different size metrics, such as effective diameter and attenuation-based metrics, were calculated for every axial CT image within a series and averaged corresponding to the same regions and images used to calculated the above mentioned regional and organ-specific CTDI_vol values.

Using an approach similar to previous efforts and AAPM Task Group 204, the estimated organ doses were normalized by CT Dose Index (CTDI_vol) values. However, for TCM scans normalized organ doses by CTDI_vol,globalwere observed to not have a strong correlation with patient size. This result was quite different from that observed previously for fixed tube current exams. In contrast, when regional descriptors of scanner output (CTDI_vol,regional and CTDI_vol,organ) were used as a modified normalization factor, the results demonstrated significantly improved correlations with patient size.

Additionally, an attenuation-based patient size metric, the water equivalent diameter (WED), was investigated in terms of its ability to describe the effects of patient size on organ dose. WED was compared to the size metric introduced in TG204, effective diameter, which is based only on patient morphology (e.g. perimeter) and not on attenuation. Results of the comparisons demonstrated no statistically significant improvements of correlation between normalized organ doses and size metric once WED was utilized, except for normalized lung dose. Although there were no statistically significant improvements, the correlation of determination, R², increased for almost all organs once WED was employed. Similarly, there was no statistically significant difference between differently averaged size metrics, i.e. global average of size metrics versus regional average of size metrics, except for normalized lung dose, which showed a statistically significant improvement in R² once a regional WED was employed as a size metric compared to global WED.

Using improved normalization quantity and patient size metric for tube current modulated CT examinations, Generalized Linear Models were used to generate a predictive model capable of estimating dose from TCM exams using regional CTDI_vol) and WED. Different models based on scanners and organs were generated to establish the level of accuracy of each model and to determine the level of specification needed to achieve best organ dose estimates. Additionally, models with different response variables, normalized organ dose versus actual organ dose, were explored and compared.

When tested using a separate test set, investigated models with regional CTDI_vol) either as a predictor or normalization factor resulted in very similar results while models created with global CTDI_vol) as a predictor resulted in underestimation of organ dose across all organs. Additionally, it was shown that a model based on pooled data was not significantly different than scanner and organ-specific models since the pooled-data model resulted in employing significant categorical predictors such as scanners and organs. This observation confirms the fact that TCM algorithms are different across scanners and regional CTDI_vol) is not capable of eliminating these differences, but it can eliminate differences among TCM functions across a single CT scanner. Predictive organ dose estimates using generated models resulted in a mean percent difference of less than 10% when compared to actual Monte Carlo simulated organ doses.

The improvement of the newly generated model was also compared against currently used dose metrics, CTDI_vol) , SSDE, and ImPACT. While comparisons with actual Monte Carlo simulated organ doses resulted in statistically significant differences between conventional dose metrics and simulated organ doses, comparisons with organ estimates from the newly developed model resulted in no difference from Monte Carlo simulated organ doses.

This work demonstrated the feasibility of estimating organ dose from tube current modulated scans from three major CT manufacturers using an improved descriptor of tube current modulated scans as normalization quantity or predictor and a patient size metric based on patients attenuation properties.

Computed tomography (CT) has proven to be a critical component of clinical care, and at present there is a strong interest in quantitative imaging: augmenting or assisting human readers through the use of quantitative and computational techniques applied to the image data. While quantitative imaging is extremely promising and has been the focus of many research projects, widespread clinical adoption has not yet occurred, in part due to the susceptibility of many tests to CT acquisition and reconstruction parameters (such as radiation dose, reconstruction kernel, reconstruction algorithm, etc.). Previous efforts to illustrate and quantify the effects of parameter selection on various quantitative imaging tests have been limited in their ability to inform broader use of quantitative imaging; this is partly due to the number of parameters investigated (typically one) and/or the cohort size. This work builds on previous efforts by studying one well-established quantitative imaging test, namely emphysema scoring, using a newly-developed, high-throughput, quantitative imaging pipeline. Because of the number of conditions investigated and the cohort side, a key goal of this work is to provide recommendations for the clinical use of quantitative emphysema scoring.

The high-throughput pipeline was utilized to reconstruct a cohort of 142 subjects, scanned using the lung-screening protocol at our institution. Each scan was reconstructed under a variety of conditions: 100%, 50%, 25%, and 10% dose levels; 0.6mm, 1.0mm, and 2.0mm slice thickness; and smooth, medium, and sharp reconstruction kernels. Additionally, two reconstruction approaches were investigated: weighted filtered backprojection (wFBP), and an implementation of iterative reconstruction (Siemens SAFIRE). Thus, each scan was reconstructed using 72 unique parameter configurations.

First, the susceptibility of quantitative emphysema scoring was investigated and characterized by determining “safe” parameter configurations (i.e. resulting in small emphysema score change from a reference value computed on the 1.0mm, smooth kernel, 100% dose, wFBP reconstruction). Second, an adaptive denoising method (bilateral filtering, adjusted based on slice thickness and dose) was applied and safe parameter configurations were reassessed.

It was found that there exist small groupings of parameter combinations near the reference value that produce quantitative emphysema scores similar to the reference. This suggests that careful protocol adherence is not strictly necessary to obtain a reasonably accurate quantitative emphysema score, however there were still many parameter configurations that resulted in large deviations from the reference score. In terms of clinical translation, this suggests that in addition to a standardized, recommended protocol, one or two small changes would not typically compromise the results. However, if a scan or reconstruction was acquired using a parameter configuration deemed “unsafe,” this approach provides no means to obtain a valid emphysema score, other than to reacquire or re-reconstruct the data, which is not typically available.

With adaptive denoising applied, substantially more parameter configurations were found to result in acceptable levels of change. Only the parameter configurations using 10% dose resulted in problematic emphysema score changes. Thus, adaptive denoising provides a means to greatly improve the reliability of quantitative emphysema scoring, most importantly in cases where the scan or reconstruction fall outside of typically accepted standards.

While there is still more investigation needed, this dissertation illustrates that widespread quantitative emphysema scoring could be made more viable via the use of adaptive denoising, and in the absence of denoising only some parameter configurations yield acceptable quantitative results. Additionally, the high-throughput pipeline discussed can be applied to future, similar investigations regarding emphysema scoring, as well as investigations into other quantitative or computational imaging techniques.

Purpose

Methods

Results

Conclusions

INTRODUCTION

In 2011, the American Association of Physicist in Medicine (AAPM) devised the Size-Specific Dose Estimate (SSDE) quantity in Report 204. SSDE is a dose metric that adjusts the commonly-reported CTDIvol metric to account for patient size. SSDE was originally developed with fixed tube current (FTC) exams of the abdomen and was later extended to the chest. SSDE represents an average dose to the center of the scan volume. As such, it gives some information concerning the radiation dose received from Computed Tomography (CT) exam but does not provide a direct estimate of organ dose by definition. AAPM Report 204 notes that the difference between the actual patient dose and SSDE may differ by 10-20%. Currently, the International Electrotechnical Commission (IEC) has introduced a measure that would allow future CT scanners to report SSDE, meaning the SSDE will become widely available dose metric.

PURPOSE

Currently, adaptive dose reduction strategies such as attenuation-based tube current modulation (TCM) are in routine clinical use. Studies comparing SSDE to organ dose thus far have used patient models or TCM descriptions that are not seen clinically. Often times, these studies are also limited to routine chest and abdomen/pelvis exams. Moreover, recent developments in dose reduction technologies and, even federal recommendations, have produced other protocols that are commonly utilized, such as, low-dose CT lung cancer screening (LDCT-LCS) and organ-based tube current modulation (OBTCM) chest exams. The purpose of this dissertation was therefore to address the aforementioned shortcomings by evaluating an estimates of organ dose from routine and non-routine exams across a range of patient sizes to SSDE.

METHODOLOGY

This dissertation evaluated SSDE in light of organ doses from four routine protocols: (1) routine FTC head exams, (2) routine TCM chest exams, and (3) routine TCM abdomen/pelvis exams. Additionally, SSDE in relation to fetal dose from routine FTC and TCM abdomen/pelvis exams was also evaluated. Furthermore, this dissertation also evaluated SSDE in light of two non-routine protocols: (1) LDCT-LCS chest exams and (2) OBTCM chest exams. In contrast to previous studies, this investigation employed patient models generated from image data. Additionally, this study also employed “whole body” voxelized phantom models that are based on image data. Where appropriate (i.e., for scans employing TCM), tube current information was either extracted directly from raw projection data or estimated based on the methodology of one manufacturer. The voxelized patient models and tube current information were used in detailed Monte Carlo (MC) simulations, currently deemed the “gold standard” of CT dosimetry, for organ dose estimation. Organ doses from MC simulation were normalized by CTDIvol (CTDIvol,16 for head and CTDIvol,32 for body exams) and were compared with the SSDE f-factors from AAPM Report 293 for head exams and 204 for body exams. Specifically, this study investigated brain parenchyma doses from FTC routine head exams; lung and glandular breast tissue doses from TCM routine chest exams; and liver, spleen, and kidney doses from TCM routine abdomen/pelvis exams in relation to the SSDE f-factors. In addition, fetal doses from both TCM and FTC routine abdomen/pelvis exams were also investigated in relation to the SSDE f-factors. For non-routine exams, lung and breast doses both from LDCT-LCS and OBTCM chest exams were investigated in relation to the SSDE f-factors. Specifically, for each routine and non-routine protocol, a one-sided tolerance interval was utilized to estimate the upper tolerance limit needed to cover 95% of the population of cases (p = 0.95) with a confidence level (α) of 5% (α = 0.05). For each evaluation, the point of comparison in terms of the tolerance window is the 20% upper limit noted in AAPM Report 204.

RESULTS

For routine FTC head exams, this dissertation found that the upper tolerance limit for the difference between normalized brain parenchyma dose and the SSDE f-factors needed to cover 95% of the population with 95% confidence was observed to be 12.5%. This dissertation observed that, for normalized lung and breast dose from routine TCM chest exams, the upper tolerance limit for the difference between lung and breast dose the SSDE f-factors needed to cover 95% of the population with 95% confidence was observed to be 35.6% and 68.3%, respectively. For TCM abdomen/pelvis exams, this study found that that the upper tolerance limit for the difference between normalized liver, spleen, and kidney dose and the SSDE f-factor needed to cover 95% of the population with 95% confidence was observed to be 30.7%, 33.2%, and 33.0%, respectively. This investigation found that the upper tolerance limit for the difference between TCM and FTC fetal dose and the SSDE f-factor needed to cover 95% of the population with 95% confidence was observed to be 35.7% and 24.8%, respectively. For LDCT-LCS chest exams, this study observed that the upper tolerance limit for the difference between normalized lung and breast dose and the SSDE f-factor needed to cover 95% of the population with 95% confidence was observed to be 40.0% and 70.1%, respectively. For OBTCM chest exams, this study observed that the upper tolerance limit for the difference between normalized lung and breast dose and the SSDE f-factor needed to cover 95% of the population with 95% confidence was observed to be 50.5% and 64.0%, respectively.

CONCLUSION

The upper threshold limit of 20% between SSDE and organ dose was found to be insufficient to cover 95% of the population with 95% confidence for all of the organs and protocols investigated in this dissertation, with the exception of brain parenchyma dose from routine FTC head exams. Results of this dissertation suggest that a wider upper limit may be more appropriate if SSDE is to be used as an estimate for organ doses. For the routine body exams, a wider threshold difference of ~30-36% will be wide enough to cover 95% of the organs with 95% confidence investigated in this chapter, excluding the breasts. This tolerance difference may also be sufficient to cover 95% fetal dose with 95% confidence from abdomen/pelvis exams of pregnant patients pending adequate sample size. SSDE is likely to serve as a conservative estimate for breast dose from routine TCM chest and lung and breast dose from LDCT-LCS and OBTCM non-routine protocols. Another dose model that takes TCM into consideration may be needed.

Purpose

Methods

Results

Conclusions