Opportunities to Reduce CT Radiation Exposure, Experience Over 5 Years at the NIH Clinical Center

Abstract

Our current study was undertaken in order to compare CT exposures during various dose-reduction initiatives at the National Institutes of Health Clinical center, to show trends in exposure reduction over a 5-y period, and to provide benchmarks that other facilities may use. Using an in-house extraction tool (Radiation Exposure Extraction Engine), we derived CT exposure data from Digital Imaging and Communications in Medicine (DICOM) headers over 5 y. We present parameters used and compare most common exams between 2010 and 2015. During a period of exposure-reduction initiatives, data of 79 396 exams from nine CT scanners on 87 scan protocols were analyzed. Adult chest exposures were reduced 53% and chest, abdomen and pelvis exams were reduced 43%  (p < 0.001). Only extremity exams did not show significantly reduced exposure. Collecting data over several years allowed us to confirm and compare several initiatives. We demonstrated significant exposure reductions during continued reduction efforts on common exams. Our results may provide benchmarks for similar centers.

INTRODUCTION

Monitoring CT dose parameters is an important early step in reducing radiation exposure. For example, in any organization, one must study before acting, known in the corporate literature as PDSA⁽¹⁾ (Plan, Do, Study, Act) or the need to know where you are (e.g. benchmark exposures), to see where you are going (exposure reductions). Greenwood et al. describes application of this and other efficiency concepts applied to exposure monitoring and reduction⁽²⁾. There are recent publications available that share radiation exposure experiences providing benchmark expectations^{(3, 4)}.

The purpose of this study was to review patient specific CT exposures during various dose-reduction initiatives over 5 y at our institution, during a period of active initiatives to reduce radiation dose. Dose optimization included body mass index (BMI) based kVp reduction in 2010, model-based iterative reconstruction (MBIR) starting in 2011, automated tube voltage selection and application of virtual non-contrast (VNC) in 2013. This large, retrospective review of our exposures can act as benchmark exposure reductions.

METHODS

Our institution's Office of Human Subject Research determined this study as exempt from institutional review board review.

MONITORING DOSE DATA

We established an upper exposure threshold of 50 mSv to alert ordering physicians that exposures were among our highest⁽⁵⁾. To this end, we have used the open source program Radiance⁽⁶⁾ since 2011 for radiation dose tracking and outlier detection.

For more detailed analysis, we developed an in-house program that we termed RE3 (Radiation Exposure Extraction Engine)⁽⁷⁾ used to collect the data, available as open source code. We applied this tool, to extract CT exposure data over a 5-y period from Digital Imaging and Communications in Medicine (DICOM) headers and radiation dose structured reports (if available) during active dose-reduction initiatives.

This allowed us to compare CT radiation exposures across research protocols and ordering physicians, and demonstrate our successful replication of targeted radiation exposure reductions relative to prior and new benchmark exposures. We performed this extraction from PACS dating from 2009 to our most recent full year of available exposures, 2015.

RE3 interfaces with an image-processing tool that calculates scanned body volume (SBV) and water-equivalent diameter (D_w)⁽⁸⁾ for each slice. With this information, we can calculate SSDE (Size Specific Dose Estimation)⁽⁹⁾ when applicable. Our extraction method is built on Perl and DCMTK and connected to our PACS and configured to run 24/7 extracting dose and other values from the DICOM headers⁽⁹⁾.

We were able to mine data on specific parameters including age, gender, anatomic region(s), scan range, scan protocol settings, research protocol number and ordering provider. We can determine the number of acquisitions per study, which is useful in analyzing studies like multiphase abdomen and/ or pelvis CT. These data were successfully retrieved for studies from all but one, now unused, scanner, which did not provide dose info in the images’ DICOM headers.

Scanner independent dose-reduction initiatives

A summary of the following dose-reduction initiatives is included in Table 1.

Table 1.

Summary of dose initiatives undertaken by this institution from 2010 to 2015.

Dose initiative	Impact
Scanner independent
Multidisciplinary workflow team	Ensure dose initiatives are being followed, and discusses outliers
Standardized technical protocols	Improves consistency of exposures between patients of similar body sizes
BMI-based kVp reduction	Reduces dose to smaller patients
Organ dose measurements	Allows for better understanding of dose impacts
Neck positioning	Reduces doses to more radiosensitive organs in the head
Informational services	Informs other groups of the availability of exposure reduced exams
ACR NRDR participation	Allows for comparison to other hospitals to identify any protocols that can be improved
Scanner dependent
New scanners	Overall improvements reduces exposure on most exams
New reconstruction methods	More efficient methods require less exposure to provide sufficient images
VNC	Dual energy enables reconstruction of non-contrast exam when needed
Spectral shaping	For certain chest exams, a tin filter is placed that allows for reduced dose with adequate images

During these initiatives, we had a continual feedback system where technologists, radiologists and ordering investigators reviewed image quality in multidisciplinary sessions to agree on new technical parameters.

Initial priorities included developing a multidisciplinary workflow team with stakeholders (lead CT radiologist and CT technologist, medical physicist, CT nurse, reception, etc.) in early 2010. This team adheres closely to the style of the one in Siegelman et al., and seems to be a useful catalyst in the establishment of exposure reducing procedures⁽¹⁰⁾.

We then standardized technical protocols in a consistent format on our shared drive for universal access. These protocols include parameters such as kVp, mA s, filters and iterative reconstruction strength. See Figure 1 for an example chest, abdomen and pelvis procedure. This standardization reduces human error when selecting these parameters, resulting in consistency among doses for a given protocol. This was especially helpful across four vendors and many models with variable terminology.

Figure 1.

This figure shows an example of our technical protocol format for an adult abdomen and pelvis CT. Note that the ACR required (for accreditation) scanner parameters are included (e.g. indication, positioning, kVp, mA s, Kernel, positioning, contrast, etc.). We include expected CTDI and DLP as a reference for technologists and radiologists.

We initiated a BMI-based kVp reduction on all chest and some pediatric body CTs⁽¹¹⁾. Patients with lower BMI can be exposed to lower amounts of radiation while retaining similar image quality. While the overall dose reduction may be difficult to notice on a large scale as it only affects patients with lower BMIs, those smaller patients benefit greatly.

As the current exposure metrics, such as CT Dose Index (CTDI) vol, retrieved from the scanner only gives us a general sense of dose exposure, we started organ dose estimations on select exams in 2012. We have begun estimating organ doses that leverages both automated segmentation and Monte Carlo calculations⁽¹²⁾. With this information, we can better evaluate certain exposure reducing techniques, exemplified with lens irradiation.

Since irradiation of the lens increases the probability of developing cataracts in a dose-dependent manner⁽¹³⁾, special care needs to be taken when performing scans around this region. For brain CTs, technologists switched to orbitomeatal line, while in neck CTs, the head was extended. These measures excluded the globes from the axial scan range in each procedure.

These dose initiatives were followed by participation in the American College of Radiology's National Radiology Data Registry (ACR NRDR), more specifically their Dose Index Registry (DIR)⁽¹⁴⁾. The DIR allows for cross-institutional comparison of radiation dose indices, thereby providing benchmarks that we can compare against.

Supporting our dose-reduction initiatives included promotional and informational sources (further encouraging dose-reduction efforts), for example, we provided information on our intranet, and newsletters for our ordering providers. After all, the introduction of new protocols can only be useful if other people use them.

We also collaborated with NCI (National Cancer Institute) on a mobile-friendly website for patients⁽¹⁵⁾. This site answers commonly asked questions including ‘What is CT?’ and ‘What should I expect?’. It also tries to allay patients’ concerns by describing radiation risks in context of these scans.

CT scanners and exposure parameters

Figure 2 shows our CT scanner vendor and model timeline over a 5-y period. We started MBIR, for body, chest and sub-millisievert cardiac exams⁽¹⁶⁾. Then we installed CT scanners that include third generation iterative reconstruction, fast scan speeds and dual energy.

Figure 2.

CT scanner timeline shows the use, decommissioning and trade-in for new scanner models and vendors at our institution over the last 5 y. Scanner F's data were not available for its entire duration, while the exposure data for scanner A were stored in the DICOM header beginning late 2009. The two H's refer to the same model of scanner, but different physical units. ‘n’ refers to the number of exams performed by the scanner.

With newer scanner's ability to reconstruct images at a similar quantity compared to older scanners, we were able to alter the scanner settings to reduce exposure, while maintaining adequate image quality. See Table 2 for example scanner exposure parameters before and after dose reductions, used for our most common exams.

Table 2.

Example scanner exposure parameters before and after dose reductions in our most common exams.

Anatomic region	Exposure parameters	2010	2015
All exams	Reconstruction algorithm	Filtered back projection	Model-based iterative reconstruction
Chest	mA sa/kVp	240–300/120	118/120
Lowest exposure chest	mA sa/kVp	240/120	118/Automated kV Selection
Chest, abdomen and pelvis	mA sa/kVp	240/120	118/120
Dual energy	mA sa/kVp	240/120	150/100 and 150
	Phases:	Triple phase	Double phase (VNC)b

^aThe mA s on all scanners sampled was dose modulated, so the corresponding value is the reference mA s.

^bThe VNC replaced the ‘real’ non-contrast; thereby allowing one less pass.

Scanner-dependent dose-reduction initiatives

With the advent of new technologies in scanners, new opportunities to further minimize irradiation arise. For instance, one of our most successful exposure reductions in abdominopelvic imaging was achieved by applying dual energy, then processing the VNC (substituting the real non-contrast passes) on our G and H scanners⁽¹⁷⁾. This allows us to reduce triple phase urograms to a single phase (split-dose urography⁽¹⁸⁾) on patients where we need to evaluate anatomic and functionality of post-operative or tumor invading renal collecting systems.

Another especially useful technique is to use spectral shaping via tin filter in conjunction with third generation MBIR. While the image quality is not high enough to be used for assessment for soft tissue details, it is sufficient for analysis of other disease types, and have proven especially useful for patients with Chronic Granulomatous Disease. These immunodeficient children and young adults require frequent repeat imaging, demanding the need for greatly reduced radiation exposures, now possible with this spectral shaping.

Comparative study and statistical analysis

Because some dose initiatives occurred at the same time, it can be difficult to parse out which caused more impact. For instance, the introduction of automated kVp selection on a new scanner could have a combined dose reduction due to new scanner and automated kVp selection. Performing cross-scanner comparisons can be challenging due to the plethora of confounding variables. Instead, we first show an overall comparison of exams between 2010 and 2015 to demonstrate the large differences that occurred because of the implementation of these dose initiatives. We follow this with evaluations of specific dose initiatives. For certain techniques, such as testing automated vs manual, BMI-based kVp reduction, it was difficult to reliably produce two samples that would not have confounding variables, so they were not tested. In the aforementioned example, the automated selection was only available on newer scanners, while the manual was performed on legacy scanners. Others, including the creation of a workflow team, are unable to be quantified, so we relegate them to the discussion.

Radiologists and ordering physicians reviewed quality with continual incremental exposure reductions as per ALARA.

For both the general and specific analysis, we performed Student's t-test for unpaired samples using Microsoft Excel. For comparisons that involved comparisons of the same subjects across time, we used Student's t-test for paired samples. Boxplots were generated using the matplotlib library in Python.

RESULTS

Data sets

We successfully extracted exposure data from 79 396 of 94 882 total CT exams from nine scanners made by three vendors. Scanner F did not have the capability to provide exposure data while it was being used here, while the exposure data from scanner A was not being stored at the very beginning of this study. We compared age and size specific exams throughout the last 5 y with exposure-reduction initiatives on 554 research protocols and 87 CT scan protocols.

The scan protocols refer to the general, technical specifications of the exam that will be performed; this includes the number of acquisitions. This means that when a scan protocol requiring three scans (e.g. for a multiphase study) is performed, it is considered a single exam. A single patient can undergo a specified scanning protocol on multiple visits, and these are considered as separate exams for this analysis. Meanwhile, a research protocol describes the purpose of the exam, as it fits in with research (e.g. for chronic granulomatous disease), which are associated with protocol numbers that help identify which groups are ordering what kind of exams.

Figure 3 represents anatomic region distribution of CT exams at our center where CT has become a major diagnostic imaging modality for chest, abdomen and pelvis; the areas we have focused our exposure-reduction efforts. There are intensive therapies conducted at NIH with increased demands on imaging with some patients getting up to 10–20 CT's a year (in 2015, 88 patients received 10 or more CTs).

Figure 3.

CT exam distribution at our institution. This represents anatomic region distribution at our research center. The y-axis represents the number of exams that include the specified body region as denoted in the x-axis. This may cause a single exam to be split among multiple columns (e.g. a single CAP (Chest Abdomen and Pelvis) exam will be counted in each of the chest, abdomen and pelvis exams).

The yearly demographics is displayed in Table 3. It is important to note that there is a substantial number of young patients scanned here, with many of them requiring frequent follow-up imaging.

Table 3.

Demographics of patients who received CT scans.

Year	# Patients	Female	Male	Avg age	≤18	19–40	41–60	61–80	≥81
2009	5429	2492	2937	47.2 ± 17.2	420	1245	2523	1197	44
2010	5563	2489	3074	47.6 ± 17.5	411	1310	2432	1347	63
2011	5600	2533	3067	47.7 ± 17.5	431	1275	2497	1335	62
2012	5614	2564	3050	47.8 ± 17.9	444	1276	2433	1402	59
2013	5554	2616	2938	48.6 ± 17.8	398	1248	2344	1497	67
2014	5715	2672	3043	49.2 ± 17.8	397	1249	2387	1603	79
2015	5905	2759	3146	49.7 ± 17.8	380	1320	2379	1737	89

Note: The counts for each year can include patients from other years, but each patient is only counted once in a given year.

Overall Exposure Reductions

Example CT exams presented showed significant exposure reduction when comparing 2015 (n = 10 373) to 2010 (n = 12 395). For example, our neuro exams, including cerebrum, sinus and necks CTs, were reduced by 29% (1117 ± 413 to 796 ± 313 mGy-cm), 55% (377 ± 188 to 170 ± 80 mGy-cm), and 37% (562 ± 166 to 354 ± 177 mGy-cm), respectively, during this time period, as seen in Figure 4a, b and c. For the sinus exams, there was a large difference between the scanners that were used to perform this exam which causes the skewness evident in the figure. During the same time period, there was a 310 mGy-cm average difference between scanners. However, even when comparing the lower exposure scanner to the newer one, there was a 42% difference (283 ± 136 to 165 ± 76 mGy-cm).

Figure 4.

Boxplot comparisons of the DLP for the most common protocols between 2010 (left) and 2015 (right) exams. (a) shows the reduction in cerebum CT (29% reduced), (b) sinus reductions (55%) and (c) neck CT reductions (37%). (d) Shows exposure reduction of 53% on our routine chest CT. (e) shows decreased exposures on combined CAP for a reduction of 34%, p < 0.001.

For body exams, a majority of our exams, the average Dose Length Product (DLP) of our adult chest CTs was reduced from 463 ± 155 to 215 ± 162 mGy-cm, a 53% reduction, and combined chest, abdomen and pelvis CT exams reduced from 996 ± 342 to 657 ± 316 mGy-cm, a 34% (p < 0.001). For pediatric cases, chest CTs were reduced by 62% from 211 ± 110 to 81 ± 70 mGy-cm, while CAP exams were reduced by 38% from 433 ± 205 to 266 ± 208 mGy-cm.

See Figure 4d and e for boxplot comparisons of chest, CT, chest abdomen and pelvis DLP from 2010 compared to 2015 reduction. The number of sampled scans is less in the earlier sample due to fewer scanners with minable data in DICOM headers. Figure 5 shows exposure reductions in pediatric exams.

Figure 5.

Dose reductions in pediatric studies. (a) 62% reduction in chest CT and (b) 38% reduction in chest, abdomen and pelvis combined exams.

These reductions were less noticeable in our extremity exams between 2010 and 2015. Using our femur exams as an example, it is more difficult to see any changes between those years, if all of the scans are included in the comparison: 652 ± 235 mGy-cm in 2010 vs 677 ± 414 mGy-cm in 2015, which was not significant, p > 0.63. However, if this comparison is performed between the most common scanners for the protocol in each year (Scanner D in 2010 and Scanner H in 2015), the DLP in 2010 was 641 ± 226 mGy-cm compared to 571 ± 341 mGy-cm, though again, this difference was not significant, p = 0.16. This is partly due to focusing exposure reductions on our most common exams (chest, abdomen and pelvis) and the challenges of reducing exposure in cases where the quality trade-off is not worth the reduction; at least for our purposes. Furthermore, the number of exams may affect this p-value, as there is a substantially less number of extremity exams performed than other types, see Figure 3.

SPECIFIC DOSE-REDUCTION INITIATIVES

One of our first major changes to protocols, the BMI-based kVp reduction, occurred during the second half of 2010. To check its efficacy, we compared chest scans in the pediatric population (the target of this initiative) in the first half of 2010 to those in the first half of 2011, before and after this change took place. We also performed this comparison within the most commonly used scanner that we had data for, while also keeping other parameters, such as pitch, collimation width and reference mA s, the same. As was expected of decreasing the tube potential, we noticed an overall decrease in patient DLP for those who underwent this new protocol (Figure 6). When comparing the DLP of patients with a SBV of <5000 cm³, this was a significant decrease (p < 0.0002): an average of 111 ± 16 mGy-cm compared to 79 ± 20 mGy-cm. We excluded patients above 5000 cm³ due to the lack of patients in 2010 who had undergone a similar protocol. Following this  BMI-based method, 55% of our pediatric patients had a 100 kVp scan performed instead of 120 kVp.

Figure 6.

A comparison of two protocols, one from 2010 and one from 2011, whose only difference was the use of 100 kVp, when manually selected to do so due to their small size. As expected, a significant difference (p < 0.0002) appears between these two protocols.

Some newer scanners have the ability to automatically select an appropriate kVp based on the size of a patient, and we have utilized this feature when it became available on scanners G and H. As such, it is difficult to compare manual kVp selection with automated kVp selection, as they were used on different scanners. The most relevant statistics would be the frequency of the exams selected to be performed at the lower kVp, as we already know that the corresponding doses should be lower. We have had two scanners with this capability. In the first 6 months after installation, the first scanner had 95% of pediatric chest scans performed below 120 kVp (80 or 100 kVp). Interestingly, for the second scanner, for its first year, only 57% had scans performed below 120 kVp (70, 80, 90, 100 or 110 kVp) for pediatric chests, while 97% of the scans were done at one of the lower settings in the following 6 months.

As mentioned previously, organ dose estimations allowed us to better assess dose initiatives that would not be obviously effective otherwise. These estimations were performed using NCICT software⁽¹⁹⁾. A good example of the utility of these measurements is in the evaluation of neck tilting in neck CT exams as a dose initiative. If we were to look at standard measurements for exposure, such as DLP and CTDIvol, there would be little difference, as these are based on scanner properties, not the patient. We have previously described how altering the neck CT protocol can greatly reduce doses to organs in the head. The modifications included reducing the scan range, as well as modifying the neck position such that the mandibular line would be perpendicular to the table. In this study, we compared 10 patients who had undergone a neck CT before and after these changes and found that doses to the lens were reduced 15-fold. Other organs also showed reduced doses: the brain, 33% reduction, and the pituitary gland, 66% reduction⁽²⁰⁾.

By utilizing the VNC procedure, we were able to remove an entire series from multiphase exams which required a non-contrast phase. This removal is mirrored in the exposure metrics for these patients. Because the VNC protocol was only able to be performed on a scanner that did not have many exams without the VNC portion, protocols from a different scanner was compared. Ten months of data was compared between a protocol without VNC, its corresponding protocol with VNC, and another scanner with the VNC protocol (Figure 7). The protocol without VNC had a DLP of 1846 ± 732 mGy-cm, while the ones with had 793 ± 248 mGy-cm for scanner G and 1011 ± 471 for scanner H. The difference between the two scanners using VNC arises due to a difference in the parameters of the scan: scanner H's protocol uses a higher reference exposure for the AEC than scanner G's.

Figure 7.

Comparison of a standard triple phase exam (left boxplot) with the significantly (include statistics) reduced dose of the VNC exams on two different scanners: scanner H (middle) and scanner G (right).

The final major scanner-specific dose initiative was to introduce a new chest protocol that utilized spectral shaping to achieve very low exposures⁽²¹⁾. By comparing the SSDEs of scans from 61 patients who had undergone this protocol with scans performed using a routine chest protocol on the same patients, we noticed a drastic decrease: from 6 ± 3 mGy with the routine chest to 0.6 ± 0.2 mGy with the spectral shaping (Figure 8). Although it is unsuitable for many protocols due to less soft tissue detail, when it can be used, it provides immense reductions.

Figure 8.

Comparison of routine chest CT (left boxplot) and the striking reduction with Sn filter (spectral shaping/ hardening) shown on the right (tiny right rectangle) boxplot. This technique resulted in a 90% SSDE reduction (p < 0.0001).

DISCUSSION

We reviewed 5 y of CT radiation exposures demonstrating exposure reductions and lessons learned at the National Institutes of Health Clinical Center with example comparison cases. Our comparisons between current day data and exposures from 5 y ago demonstrate consistent exposure reductions on most exams. We were also able to automatically obtain SBV segmentation of scanned regions with our in-house extraction tool for size estimation in most cases. From there we can calculate SSDE, a commonly recognized and accepted measure⁽²²⁾.

One of the motivating factors for sharing our experience was the challenge of monitoring and reducing exposures across multiple scanners and vendors; each having their own issues with dose reporting. This quality assurance and improvement exercise emphasizes the importance of technical protocols that allowed us to have similar settings across vendors with a variety of terminologies for scanner settings. This may allow our results to be generalizable for facilities with similar challenges of multiple platforms to analyze and reduce exposure.

Our weekly outlier detection notifies principal investigators about unusually high exposures by secure emails. We also have monthly exposure reviews with team meetings and discussions on dose reduction. Our outreach program informs investigators of exposure-reduction efforts and helps them establish new research protocols to select optimal scanner settings and exposure reductions for their studies.

We have also participated in the ACR DIR (Dose Index Registry) for several years comparing our exam exposures with participating radiology departments across the country. This and periodic comparisons of imported outside exams on referral patients, allowed us to confirm exposure reductions by comparison, and reduce exposures.

By analyzing our data, we were able to notice some important trends. The most obvious of which is that with better technology comes a significant decrease in dose, whether it be from more efficient scanners and reconstruction algorithms or from dual energy's capability of producing VNC images. However, during our time using this technology, we have noticed that it has its own issues. For instance, during the initial implementation of MBIR in one of our scanners, reconstructed images were a bit too low quality to be confidently used. Implementing this technology requires communication between radiologists, who are able to ensure the quality of the images, and those who are implementing the initiatives, to ensure that patient care does not diminish.

While new technology is important, there are still a variety of options that allow for a significant reduction in dose that does not require the use of new technology. These initiatives are more focused on personalizing the exam, such as tailoring the kVp to match the BMI of a patient and ensuring the patient's position is optimal, monitoring the data, and fostering a dose conscientious environment.

The utility of close monitoring of protocols is exemplified when our chest CTs on a new scanner had multiple protocols with a variety of settings associated with it, some higher exposure than they could have been. By discovering this inconsistency, we were able to standardize the protocol so that exposures would remain consistently low. Even with newly installed scanners with the latest reconstruction algorithms, attention to scan protocols is paramount. While it is easy to see the benefits from the personalization and monitoring, the community can be easily overlooked, even though it is just as important.

The community encourages the usage of all of our available resources, while providing a dose-reduction mindset. Furthermore, with this mindset, there is more incentive to put in the extra effort that is required to implement these initiatives. By informing ordering physicians of newly available protocols, they will be able to order the lower dose exams, if appropriate, rather than consistently using routine protocols. Furthermore, by working together with these physicians, we can develop more specific protocols that provide lower exposures while maintaining adequate image quality for their purposes, as we have done with our immunocompromised patients.

The authors realize several limitations, for example, although our extraction engine program is open source⁽²³⁾, some informatics expertise is required to extract DICOM headers, set up a server and obtain meaningful outliers. We found some discrepancies between DICOM header data and dose pages that varied with each vendor, however, direct comparisons of select exams allowed us to narrow down the reasons.

Our unique opportunity to greatly reduce chest CT may be a niche protocol for when viewing fine details of the lung is unnecessary, like with some of our immunocompromised patients.

We replaced our BMI-based kVp reduction on chest CT once we had the possibility of automated kV selection on the G and H scanners. A limitation results such as HU quantification, however, we did not apply this when quantifying tumor density (for example).

Since we continually reduced doses in ways that we have learned during the last 5 y, it is difficult to attribute dose reductions to isolated events; however, we believe the major component was teamwork and monitoring of exposures, while promoting a dose conscientious culture.

We did not objectively assess quality of exams while incrementally lowering doses, rather, like most centers, we continually reduced exposure parameters per ALARA with feedback from radiologists and ordering providers to assure acceptable quality.

Lastly, not all exams could have their exposures reduced over the years, as the quality trade-off was not acceptable to meet our research and clinical needs.

CONCLUSIONS

Collecting size specific CT exposure and other data over several years has allowed us to confirm and compare many types of exposure-reduction initiatives, especially in our most common exams across several hundred research and scan protocols. Our dose reductions success is only partially due to updated CT parameter settings and new technologies. For example, we believe our education and outreach has inspired a dose conscientious culture. We share our successful dose-reduction initiatives as well as the lessons learned, to provide benchmark dose reductions from a large research center with multiple CT scanners.

FUNDING

This work was supported by the National Institutes of Health via a postbaccalaureate intramural research training award. Dr. Bluemke reports non-financial and financial support from Carestream for PACS-related research programs. Dr. Folio reports non-financial and financial support in the form of a research agreement from Carestream for PACS-related research programs.