Article Text

Download PDFPDF

Original research
Radiation dose reduction during neuroendovascular procedures
  1. Pearse P Morris1,
  2. Carol P Geer2,
  3. Jasmeet Singh2,
  4. Waleed Brinjikji1,
  5. Rickey E Carter3
  1. 1 Department of Radiology, Mayo Clinic, Rochester, Minnesota, USA
  2. 2 Department of Radiology, Wake Forest University School of Medicine, Winston-Salem, North Carolina, USA
  3. 3 Department of Biomedical Statistics, Mayo Clinic, Rochester, Minnesota, USA
  1. Correspondence to Dr Pearse P Morris, Department of Radiology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, USA; morris.ppearse{at}


Aim To describe the impact of steps towards reduction of procedural doses of radiation during neuroendovascular procedures.

Methods Phantom exposures under controlled circumstances were performed using a Rando-Alderson adult-sized head phantom. Customized imaging protocols were devised for pediatric and adult imaging and implemented in clinical use. Outcome data for estimated skin doses (ESD) and dose–area product (DAP) following pediatric and adult diagnostic and interventional procedures over 4.5 years were analyzed retrospectively.

Results Dose estimates were reduced by 50% or more after introduction of customized imaging protocols in association with modification of personnel behavior compared with doses recorded with previously used vendor-recommended protocols.

Discussion Substantial reductions in radiation use during neuroendovascular procedures can be achieved through a combination of equipment modification and operator behavior.

  • angiography

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Increased use of ionizing radiation in the course of medical imaging has doubled the per capita annual effective dose in the United States to a figure >6 mSv per year compared with data from 1987.1 This has raised concern about the increased burden of stochastic effects of radiation with future malignancies.2–4 Population-based outcome studies from the United Kingdom, Australia, and Taiwan of patients exposed to CT radiation in childhood indicate a detectable level of tumor genesis compared with controls for dose levels <100 mSv.5–8 The extremely low level of tumor genesis in these studies provides reassurance that the stochastic risks of medical radiation are extremely small. However, in pediatric CT scanning it is now generally accepted that reducing the very small number of delayed stochastic complications to the lowest possible range should form the basis of medical practice. This has resulted in significant modification of technique and indications for CT use in children.9 10 The same stringent standards are warranted in the use of neuroendovascular equipment.

This paper reports the results of a single-center effort to reduce radiation doses associated with neuroendovascular procedures over a 4.5-year period.


Phantom data

Phantom exposures using a Rando-Alderson adult-sized head phantom were performed in anteroposterior and lateral projections on a Siemens Artis Zee biplane system to compare vendor-recommended imaging protocols with customized protocols for digital subtraction angiography (DSA), fluoroscopy, and roadmap. Geometric conditions such as table height, source–image distance (SID), magnification levels, collimation, duration of exposures, frame count, etc, were all controlled. Dose variables manipulated in the customized process consisted principally of dose per frame and frame rates for DSA, dose per pulse and pulse rates for fluoroscopy and roadmapping, and copper filtration. Radiation readings were taken from the integrated dosimeters installed in the biplane equipment. During the period covered by the analysis, the internal dosimetry of the Artis Zee biplane room was monitored quarterly by a departmental physicist using an independent dosimeter. Agreement of the independent dosimeter and that of the biplane equipment was within a margin of <10% throughout the study.

Variables that affected final image quality but were not directly related to radiation dose included reconstruction algorithms and edge enhancement. Variables affecting dose deriving from operator behavior such as SID (height of the image receptor) and grid insertion/removal were evaluated over a range of magnification options.

Clinical data

A retrospective study of radiation readings from adult and pediatric (<18 years) diagnostic angiograms and neurovascular interventional cases over a 4.5-year period was performed with approval of the Wake Forest Medical Center institutional review board committee. Retrospective informed consent was not required. In the first 2 years of the study (baseline), vendor recommendations were used in all patients. Thereafter, modifications to the imaging protocols were introduced and further modified over time with a consistent trend towards dose reduction with each change. Data included total fluoroscopy time, total dose–area product (DAP), estimated skin dose (ESD), number of digital subtraction angiography (DSA) runs, and number of DSA images in each plane of imaging. All cases were conducted under the direct supervision of the senior author, but trainee and other faculty participants performed variable and substantial proportions of the cases. Only conventional cerebrovascular and craniofacial cases were included in the analysis. Spinal angiograms and minor procedures such as sclerotherapies were excluded.

Data analysis

Phantom and clinical radiation readings were stored on Excel files (Microsoft) and transferred to JmpPro (SAS, Cary, North Carolina, USA) for parametric statistical analysis.

Results I: phantom data

Virtually identical results were obtained in both planes for all non-clinical portions of this study, and hereafter only data from the analyses of phantom readings in the anteroposterior plane will be presented.

Effect of reduction of dose per frame and dose per pulse

The ranges of detector doses per frame of DSA and per fluoroscopy pulse with corresponding rates per second are shown in table 1.

Table 1

Customized imaging protocols

Comparison of the lowest-dose infant imaging protocol with the standard vendor adult protocol, including modification of frame rates, for a 10 s DSA run showed a 40–42-fold reduction in DSA readings (table 2). The readings for DAP and ESD for a 30 s exposure of fluoroscopy showed up to an eight- to ninefold decrease and for roadmap up to a sevenfold decrease. These numbers were consistent over the entire range of magnifications and in both planes of imaging.

Table 2

Phantom exposures of customized protocols

Effect of grid removal

In cardiology, removing the grid in front of the flat-panel detector has become a standard stratagem for dose reduction, where it is described as most effective in children weighing <20 kg.11 12 DSA data for grid removal are provided in table 3, showing a reduction in doses of 14–57% in DAP and skin dose, the effect being more modest at higher magnification levels. Diminished quality of imaging after grid removal can be compensated for by a variety of stratagems, including close attention to SID and enhanced use of image postprocessing default settings for edge enhancement (figure 1).

Figure 1

(A–D) Four phantom DSA images taken with a low-dose pediatric protocol. Images (A) and (B) are taken with the grid in place and edge-enhancement (EE) settings of 10 (low) and 30 (manufacturer default), respectively. Images C and D are taken without the grid and with EE settings of 30 (default) and 90 (highest), respectively. A perceived loss of resolution or sharpness on the grid-removed images can be compensated for through augmentation of EE settings.

Table 3a

Effect of grid removal on phantom fluoroscopy doses 30 s, source–image distance 100 cm

Table 3b

Effect of grid removal on phantom DSA doses, pediatric protocol, anteroposterior plane, source–image distance 100 cm, 15 frames

Effect of source–image distance

Beam intensity varies in proportion to the square of the source-image distance. Therefore, inattentive imaging with an unnecessarily elevated detector location results in unnecessary elevation of dose and deleterious elevation of the kVp (figure 2). For fluoroscopy the effect on skin dose was an increase of 70% over a range of 100–120 cm, while DSA showed an increase of 26% over the same range.

Figure 2

The impact of increased source–image distance (SID) on skin dose (A) is illustrated for fluoroscopy (left) and DSA (right) using a pediatric imaging protocol under controlled conditions with a phantom in the anteroposterior plane. Source–skin distance was constant through all conditions, field of view 22 cm, Cu filtration 3 mm, grid in. The impact of SID on output kV is illustrated in figure 2B. Therefore, it becomes easy to understand that the cumulative penalty for inattention to the SID over an entire case is twofold: higher skin doses and proportionately higher dose–area product (not illustrated), and deterioration of image quality due to unnecessary elevation of the kV.

Result of increased copper filtration

The interplay of machine output in kVp and mA, magnification levels, and load (patient size) is a complex relationship. A minimal degree of beam filtration is desirable to eliminate low-energy photons. Excessive metal filtration, however, drives up the kV (figure 3, table 4). High levels of copper filtration are associated with lower dose readings but with a penalty of elevated kV. This kVp penalty is less prominent when the grid has been removed, implying that higher copper filtration levels might be considered advantageous when the grid has been removed, amplifying the radiation-reduction benefits of grid removal even further.

Figure 3

Interaction of grid removal and variation of copper filtration using DSA output kV readings with and without the grid are illustrated using a pediatric DSA protocol, field of view 22 cm, source–image distance 100 cm. Fixed copper filtration levels were used with and without the grid, all other variables being controlled. The data show that the deleterious impact of rising kV is attenuated when the grid is removed.

Table 4

Effect of copper filtration on phantom DSA readings. Low-dose pediatric protocol, DSA 0.54 nGy/frame, field of view 22 cm, source–image distance 100 cm

Results II: Clinical data

Outcome data for adult diagnostic and interventional cases and for pediatric diagnostic and interventional cases are listed in table 5. Linear regression analysis over time was performed showing a statistically significant reduction of approximately 50% for dose measurement in all groups, with the exception of the pediatric interventional cases, where the number of cases was limited (figure 4). Multiple linear regression analysis was also performed to test if the effect persisted after adjustment for other factors that might influence total dose–namely, general curtailment or inhibition of the use of radiation by the operators over time through reduction of the number of DSA runs per case, shortening of DSA runs, or curtailment of the use of fluoroscopy overall.

Table 5

Patient demographics and data

Figure 4

Analysis of linear regression analysis of radiation readings versus time (50 months) for adult and pediatric cases. In the interests of space, only the results for the dose–area product (DAP) versus time in adult (A) and pediatric (B) diagnostic procedures are illustrated. However, equivalent results were obtained in adult diagnostic estimated skin dose (ESD) versus time (R2=0.03, p<0.009), adult interventional DAP versus time (R2=0.07, p<0.0001) and ESD versus time (R2=0.028, p<0.0094), and pediatric diagnostic ESD versus time (R2=0.2, p<0.0014). For pediatric interventional ESD versus time (R2=0.019, p<0.59) and DAP vs time (R2=0.033, p<0.48) the trend was towards reduction in radiation parameters but did not reach statistical significance.


Modern biplane neuroangiography machines can achieve astounding imagery of the cerebrovascular system. Tiny, vital structures which Thomas Willis and Christopher Wren struggled to discern and sketch by dim candlelight in the 17th century are routinely manifest to us every day, a privilege that can sometimes be easy to accept as commonplace. Engineers and vendors have achieved this imaging quality through technical complexity while still complying with standards of durability, safety, consistency of performance, and simplicity of use. Physicians can focus on the clinical problem despite the technological complexity through the use of vendor-provided ‘black box’ protocols that yield excellent images. This is desirable in many ways, and undoubtedly the most pragmatic solution. However, it removes some flexibility in establishing a hierarchy of priorities by the operator and results in higher than necessary doses of radiation.

Striving for consistently excellent digital imaging may no longer be the foremost imaging priority, as a balance between imaging quality and ‘lowest possible radiation doses’ is important. In cardiology, for instance, where radiation safety has become a pressing concern for patient and staff safety, striving for ‘adequate’ image quality rather than ‘excellent’ has become a standard of safety,13–16 a concept yet to make an impact on neuroendovascular imaging.

Stochastic risks of radiation dose from neuroendovascular procedures

The scientific and public scrutiny of the use of CT scanning in medical practice and its attendant radiation risks has not yet been focused on neuroendovascular procedures. Such a course of events would probably serve the field well, considering the impact of such scrutiny on the design of CT scanners and on their use over the past 15 years. A landmark paper in 2007 by Brenner and Hall4 was one of many publications which drew attention to the rapidly rising rate of CT scanning per head of population and the possible delayed stochastic complications which might be expected, particularly in the pediatric patient group. Their argument was that dwelling on the effective dose alone and thinking about radiation dosimetry averaged over populations was obscuring the implications of the disproportionately higher organ doses involved in pediatric CT scanning. A head CT scan in a small child under technical conditions of that time might be expected to deliver a brain radiation dose of between 50 and 100 mGy, considerably more than the approximate dose of 20 mGy in an adult relevant at that time. Consequently, a different scale of stochastic risk is necessary in thinking about organ doses to the brain and marrow in children than might be considered reasonable in adults.

Recent papers support a view that even though the stochastic risks of radiation exposure from medical imaging might be extremely low, nevertheless low doses do carry a detectable risk and that this risk might be measurably higher in children. Pearce et al 6 performed a retrospective review of data from the British NHS of patients under the age of 22 years scanned between 1985 and 2002. Outcome rates of leukemia indicated an excess relative risk of 3.18 after a marrow dose of 30 mGy and for brain tumors an excess relative risk of 2.82 for brain doses of 50–74 mGy. Mathews et al 5 performed a similar study using Australian Medicare records and compared cancer incidences in 680 211 patients aged <19 years exposed to CT scanning with incidences in those without CT exposure between 1985 and 2005. An increased relative risk of cancer of 24% overall was observed, but this number was 35% in patients aged 1–4 years at the time of scanning. The estimated effective radiation dose per scan was approximately 4.5 mSv, and the data suggested a dose–response curve—that is, more CT scans with higher doses implied higher risk, with an approximately 16% increase in the relative risk per scan. The mean duration of follow-up for the study was 9.5 years, so these outcomes may represent an underestimation of the cancer risks attributable to CT scanning. Moreover, a study of an international cohort of radiation workers provides evidence of a radiation dose–cancer association even with protracted low-level radiation exposure of the order of 1.1 mGy a year.17

From these studies, it is reasonable to conclude that a linear no-threshold model of a dose–response curve of the risk of carcinogenesis following low doses of radiation is most likely and that these risks can be reduced through improved operator training and technique. A sustained drive to train interventional cardiologists in dose-reduction techniques has demonstrated that substantial reductions in skin dose and DAP can be achieved even among trainees for diagnostic and interventional coronary procedures.18–21 The increased interest in radiation reduction among interventional cardiologists has been probably honed, in part, by anecdotal studies of health effects, particularly left-sided brain tumors.13 22 23 This concern may be exaggerated, but as a stimulus to caution it can only redound to increased patient care.

Within modern radiology, growth of concern about angiographic procedures has been slower. A recent study of radiology technologists detected a higher tumor incidence in those exposed to fluoroscopic procedures than in unexposed controls, prompting a call for renewed attention to this topic.24 25

The authors acknowledge the limitations of the data presented here. The paper reflects the clinical work of a single center over 5 years, during which time several implicit and tangential variables might have had an impact on the radiation readings recorded during diagnostic and interventional cases. The effect, for instance, of variable degrees of collimation could not be measured during this time and might have affected the DAP outcome. This was a retrospective study without a contemporaneous control group. However, the nature of known risks from radiation during the time-frame of this study would have presented an insurmountable ethical obstacle towards creating a control group of patients. Vendor innovations with specific radiation reduction features have been marketed since completion of this retrospective study, but in the opinion of the authors the operator-driven problems discussed in this study are still relevant. Air kerma readings have since become the radiation dose recorded in many hospitals. We did not record air kerma specifically in this study, but the ESD in mGy reported here is functionally equivalent as a measure of likely peak skin dose. The retrospective methodology used in this study was not specifically designed to detect an effect of trainee participation. However, the protracted 4.5 years of data collection, during which time approximately 18 neuroradiology fellows rotated on and off the neuroendovascular service, probably eliminated any effect related to progression of trainee skills. Finally, the study did not show a robust reduction in pediatric interventional doses over time, although this might have been a reflection of the small numbers involved.


This study compares the radiation used by an operator who performs a neuroendovascular case with default settings with that of an operator who considers how to achieve greater control of the flexibility of the room. Lesion detectability, as a semiquantifiable surrogate for ‘image quality’, is proportional to, among other factors, the square root of the detector dose. This implies that a small improvement in image quality is inextricably tied to a non-linear increase in radiation dose, and that a prioritization of image quality in all circumstances will inexorably lead to a substantially higher dose per image. We have shown that several small adjustments to exposure parameters have a profound cumulative and multiplicative effect on total radiation use during a diagnostic or interventional case while still maintaining sufficient image quality. Furthermore, we have demonstrated that inattentive use of angiographic equipment results in unnecessarily elevated doses and also deterioration of image quality. Greater attention to operator training and radiation awareness, in particular, would serve the long-term interests of our patients by reducing the risks of high skin doses in long interventional cases, and by reducing the potential stochastic risks of tumor genesis related to radiation.


The authors acknowledge the assistance of Sonia Watson, PhD, in editing the manuscript.



  • Contributors PPM is senior author and participated in data collection and manuscript preparation; CPG and JS participated in data collection and manuscript review/rewriting: WB participated in manuscript review and statistical analysis; REC participated in manuscript preparation and directed the statistical analysis and data review.

  • Competing interests None declared.

  • Ethics approval Wake Forest University Medical Center institutional review board.

  • Provenance and peer review Not commissioned; externally peer reviewed.