Management of pediatric radiation dose using GE fluoroscopic equipment

Dose and image quality are bound together in the cost–benefit proposition of medical imaging using ionizing radiation to diagnose and treat patients; therefore, the two should be discussed together. Adherence to the ALARA principle requires judgments on the part of the clinician and the manufacturer. The clinician must judge the level of image quality required for diagnosis and/or therapy in a given procedure and the manufacturer must make technical judgments about what defines ‘optimal–image quality and the levels of radiation dose required to reach the clinical objectives. GE’s philosophy in this regard can be summarized as follows:

Although dose is well-defined and relatively easy to measure or estimate, image quality is a multidimensional entity judged by the clinician in performing critical visualization tasks. Lacking a complete understanding of the human visual system and higher-level visual brain functions, we must rely on other quantitative methods to characterize image quality, and try to do so in a manner that best correlates with clinicians–perception. The classical dimensions of image quality include contrast, gray scale, noise, spatial resolution, temporal resolution, low contrast detectability, modulation transfer function, and various distortions. For the purposes of this article, however, we propose a simpler, higher-level list of imaging performance characteristics that are appropriate for modern digital imaging systems, including those with flat-panel detectors:

AutoEx is described in more detail with the aid of Fig. 3. AutoEx provides the link between patient information (deduced from the detector signal) and technique values to use (relayed to the generator and collimator). A two-step approach is employed. In step one, an equivalent patient thickness is determined based on the average dose to the detector and the technique values used for that exposure [6]. In step two, the technique parameters are determined for the next exposure using the patient thickness value and predetermined look-up tables. Design of the look-up tables (or ‘trajectories– is based on the cost–benefit relationship of image quality and dose. Defining image quality as contrast-to-noise ratio (CNR), trajectory tables are designed to give a targeted CNR for the lowest possible patient entrance dose. An example table is shown in Fig. 4. This is an example of a fluoroscopic trajectory for adult angiography. The corresponding pediatric angiographic trajectory employs a shorter pulse width of 4 ms maximum for equivalent patient thicknesses less than 30 cm PMMA. Other constraints are also factored into trajectory design, such as maximum and minimum generator settings, X-ray tube capabilities, regulatory dose limits, etc. A trajectory table is developed for each set of available system parameters (e.g., image magnification, frame rate, dose level, etc.) for each operational mode (e.g., fluoroscopic, spot, DSA, etc.). This provides a great deal of flexibility in design and also enables enhancement of system performance via periodic software upgrades.

In order to fit the needs of different clinical applications, regional regulatory norms, and varying preferences of clinicians, several trajectory families are offered with the Innova and Precision 500 D products. Such families typically feature different dose levels and related optimization criteria. Figure 5 lists the offerings for the Innova 4100 and their typical dose levels. The first dose customization is performed at installation; one of five AutoExposure preferences is selected that best fits the needs of the customer. In addition, the user can select at installation between two distinct strategies for dose reduction in lower frame rate fluoroscopic modes. The first option, called ‘Maximum Dose Reduction– uses the same X-ray techniques for 15 fps and 7.5 fps as it would in the case of 30 fps, thus reducing the dose by factors of 2 and 4, respectively. Because of the characteristics of human visual perception, the application of this strategy results in a reduction of the perceived image quality in the lower frame rate modes compared to the 30 fps mode. The second option, ‘Balanced IQ/Dose– provides similar perceived image quality for all frame rates, which can be obtained by reducing the dose corresponding to 30 fps fluoro to approximately 75% for 15 fps and 50% for 7.5 fps [7]. For the families and frame rates offered, a 20:1 range of doses can be achieved. Figure 6 lists the trajectory families offered with Precision 500 D and typical dose levels. Families are offered for adult and pediatric modes. The pediatric mode is further divided into ‘grid out–and ‘grid in–modes. The grid out mode is primarily for young children, where the grid is kept out because of the lower scatter-to-primary ratio for dose reduction. The grid in mode is provided for larger children and adolescents, where scatter can cause significant image quality degradation and necessitates use of the grid. The maximum dose rate for both pediatric trajectories is 5 R/min, while for the adult trajectory it is 10 R/min. In all trajectory families, the variation in dose rate with frame rate follows the recommendations of Aufrichtig et al. [7]. Finally, trajectories are also developed to meet specific country regulations. The German trajectory is shown as an example. A dose range of about 6:1 is spanned proceeding from the adult trajectory family (30 fps) to the pediatric grid out family (3.75 fps). Including the German trajectory family would present an even larger range.

The control of dose with field-of-view has evolved as imaging chains have become more sophisticated. In early image intensifier-based systems, the average entrance dose to the intensifier varied inversely with the field-of-view as a result of the decrease in gain of the intensifier with smaller fields-of-view (and the fact that no further compensations were made in the automatic brightness control loops). As a result, patient skin dose would increase significantly when using smaller fields-of-view. Of course, image quality improved, as well, as a result of the higher signal-to-noise ratio associated with a higher X-ray photon flux. However, as image systems became more sophisticated, it became possible to separate any need for higher image quality with smaller fields-of-view from the behavior of an imaging component. On the subject of the need for higher image quality with smaller fields-of-view, there is a spectrum of opinion with two extremes. At one end is the belief that dose does not need to be changed with field-of-view, which is consistent with the way that traditional film/screen cassettes inherently behave. (But bear in mind that films are not magnified for display.) The only consideration in deciding on dose should be the signal-to-noise ratio required in the image. At the other end of the spectrum is the belief that the perceived signal-to-noise ratio in the displayed image should be kept constant. Simply magnifying the image, without increasing the dose, will tend to improve the visual performance of the clinician, which will also make the noise in the image more apparent and often objectionable. Perhaps implicit in this belief is that the selection of smaller fields-of-view implies a desire to see smaller detail and smaller contrast differences, which is improved by increasing the dose. GE has chosen to consider both of these opinion camps in the design of its current fluoroscopic systems, by adopting a general principle of varying the dose inversely with the linear dimension of the field-of-view, e.g., diameter or edge. The logic is that some increase of dose is warranted in selecting smaller fields-of-view, but not as much as 1/area. Imagine, for example, threading a guidewire and catheter from the groin to the aortic arch using the largest field-of-view afforded by the system for speed, followed by use of a smaller field-of-view to perform the maneuver of entering the vertebral artery from the arch. One could argue that a significantly higher level of image quality is warranted in the latter case compared to the former. And, in any case the clinician is afforded several dose selections in every field-of-view, making it possible to follow the ALARA principle in any imaging situation.

Image processing techniques can have profound effects on the qualities of a displayed image produced from a given X-ray dose. In this section we discuss two important types of image processing, dynamic range management (DRM) and fluoroscopic noise reduction (FNR), through temporal filtering.

For a quantum-limited system, the noise reduction achieved (relative to no averaging) is to a good approximation equal to \( {1 \mathord{\left/ {\vphantom {1 {\sqrt {N_{eff} } }}} \right. \kern-\nulldelimiterspace} {\sqrt {N_{eff} } }} \). Using these relations, it is apparent that significant noise reduction occurs for small values of X_live. For example, for X_live = 20%, N_eff = 9 (using Eq. 1), and the noise reduction is \( {1 \mathord{\left/ {\vphantom {1 {\sqrt 9 }}} \right. \kern-\nulldelimiterspace} {\sqrt 9 }} \), or 0.33. This means that the apparent noise in the FNR-filtered image sequence would be the same as an unfiltered sequence at a nine-times higher dose.

The noise reduction achieved by FNR comes at the expense of increased motion blurring artifacts in the image. If anatomy is moving on a frame-to-frame basis, averaging a number of those frames will cause undesirable blurring and/or multiple images of moving features. The advanced FNR algorithm employed by GE mitigates these undesirable effects by employing motion compensation. Motion compensation varies N_eff, depending on the degree of motion deduced for a given pixel. Motion is deduced by comparing the pixel value for the live frame with the previously displayed frame, and the value used for X_live is dependent upon the difference, denoted as δ. Typically for large δ, X_live is set to a larger value, thus decreasing N_eff and the amount of temporal filtration applied. For small δ, X_live is set to a smaller value, increasing N_eff. In such a manner, noise reduction is achieved across the entire image but in a manner that does not present the distracting effects of blur and multiple images. By reducing quantum noise in this way, FNR is consistent with the ALARA principle in that it enables lower dose operation.

In order to satisfy different clinician preferences and to be effective for the various clinical modes of operation, a variety of FNR settings are available for the Innova and Precision 500 D products. Each setting can be thought of as corresponding to a different mapping of N_eff versus δ. In the Innova products, a unique setting can be made in each user-defined procedure protocol. FNR settings are also automatically linked to clinical operation. For example, with lower frame rate, the likelihood of significant anatomical motion between consecutive frames increases. Therefore, the preferred mapping of N_eff versus δ can be set differently as a function of frame rate.

In this section we attempt to provide a comprehensive list of all the features that exist for dose management and minimization. We provide more complete explanations of those features that have not been discussed in previous sections.

New imaging techniques can also have significant impact on patient dose. Two are discussed here: Innova Spin and Innova Chase. Innova Spin is a rotational acquisition technique utilizing DRM image processing discussed above to prevent image burnout in the lung and at skin boundaries during the rotation. It has been found that rotational acquisition sequences can replace several fixed-angle acquisition sequences, thereby saving radiation dose and contrast medium. This has been demonstrated both in adult coronary angiography and pediatric patients [9, 10]. Innova Chase, originally designed as an unsubtracted bolus chase technique, also utilizes DRM processing to prevent image burnout and blackout while panning over the abdomen and extremities. The Innova Chase technique has been found to be useful in reducing adult dose in uterine fibroid embolization (UFE) procedures compared to DSA [11, 12], without compromising clinical image quality. The latter result occurs because the DRM processing reduces the black to white variation in the image without reducing vessel contrast, something that DSA does extremely well, but without the possibility of misregistration artifacts. It is, therefore, reasonable to propose that this technique could also be useful in pediatric procedures where patient motion often limits the performance of DSA.