The use of bismuth breast shields for CT should be discouraged

Arguing against the Proposition is Robert G. Gould, Sc.D. Dr. Gould obtained his doctoral degree from Harvard University in 1977 and then took an appointment at the University of California San Francisco, where he has stayed for his entire career and is currently Professor of Radiology. He is a Past-President and Fellow of the AAPM and has served on or chaired numerous AAPM committees. His major research interests are developments in molecular imaging, small animal and high-resolution SPECT, and artifact and dose-reduction strategies in CT. Breast shielding wastes radiation already delivered to the patient. The dose reduction from breast shields is due to the attenuation of x-rays coming from the anterior direction. When the tube rotates to the posterior and lateral positions, the shield absorbs photons exiting the patient—x-rays that would have otherwise reached the detector. This increases image noise without any commensurate reduction in radiation dose, which is clearly inconsistent with the ALARA principle. Simply by decreasing the tube current, one can reduce not only anterior (breast) dose by the same amount but also reduce dose to the entire scan volume.2 Even though the tube current is reduced, image noise is equivalent to when bismuth is used.3 Breast shielding degrades image quality and CT number accuracy. Bismuth shielding increases image noise across the entire image, not just directly under the shields.3–5 It also causes streak and beam hardening artifacts, which can artifactually increase CT numbers, again not just below the shield, but throughout the entire image.3,6 If the increase in image noise caused by bismuth shields is diagnostically acceptable (as claimed by proponents of the method), it is better to simply decrease the tube current to reduce breast dose. This avoids the artifacts and CT number errors caused by the shields. Furthermore, technologists do not have to spend time positioning the shields carefully and cleaning them between patients. Using bismuth shielding in conjunction with automatic exposure control (AEC) systems leads to unpredictable and potentially undesirable dose and image quality performance. AEC is widely used to adapt tube output according to the specific patient's attenuation profile and diagnostic task. Placing a shield on the patient before the CT radiograph is acquired will lead to overestimation of patient attenuation and result in an increase in tube current, thereby defeating the purpose of using the shield.7,8 If the shield is placed after the CT radiograph is acquired, the image quality prescribed by the user will not be obtained due to the unanticipated attenuation of the shield. Either way, the sophisticated “phototiming” algorithm is thwarted; the prescribed dose or the prescribed image quality is not delivered. Thus, we strongly advocate that medical physicists discourage the use of bismuth breast shields. Many improvements have been made in CT scanners that lower radiation dose and many methods to reduce patient dose during scanning have been published.9,10 Bismuth shields placed over the patient during CT imaging are shown to reduce the dose to the region immediately below the shields by 30% or more.6,11–14 Both the thyroid and breasts are considered stochastically sensitive organs and have individual weighting factors for calculating effective dose. Indeed, the ICRP weighting factor assigned to the breast equals the highest of any organ (0.12).15 Bismuth shields used to protect the breast are commercially available at modest cost and can be placed in sterile plastic bags and reused. They are positioned on the patient after the scout image is acquired because many CT scanners use the scout image as the basis for the adjustment of tube current during scanning. If the bismuth is placed prior to scouting, the algorithm for current adjustment may compensate for the attenuation of the shields, reducing their effectiveness. On CT scanners that adjust the tube current based on a direct measure of the radiation intensity during scan acquisition, notably those made by Siemens, the use of shields is not recommended. Care must be taken in placement of the shields since they produce streaks, which should not be allowed to project into the patient's anatomical image. This is accomplished by offsetting the shields from the patient's surface by several centimeters and positioning the shields so that their surface is tangential to the curvature of the torso. Placement is not difficult and a few minutes of technologist training is sufficient. Bismuth shields affect the CT number of the tissue, most noticeably immediately below the shields. The CT-number increase in soft tissue can be quite significant close to the shield, more than 100 HU, but the effect decreases rapidly with distance from the shield.6 If the shields are offset from the skin by 2 cm, the CT number increase is 40–50 HU near the skin, less centrally, and a few HU near the skin opposite the shield.6 The offset does not significantly affect the dose reduction that is achieved.6 When breast shields are used on adult women, the effect is primarily seen outside the ribs. While inaccuracy of the CT numbers may be offensive to medical physicists, it does not seem to bother most radiologists nor has it been shown to have a deleterious clinical effect.13 Shields should not be used whenever quantitative assessment of CT numbers is needed, such as when used to quantify the amount of coronary artery calcium. Bismuth shields also increase noise in the image, but again this effect is greatest near the shield and lessens quickly with distance.6 This effect has not been shown to result in reduced diagnostic effectiveness.13 A significant drawback of bismuth shields is that, while they reduce the beam intensity entering the patient, they also reduce the intensity of the beam exiting the patient. Thus, information-carrying photons are attenuated between the patient and the detectors. However, dose is reduced regionally, and radiation sensitive organs can be spared significantly. Bismuth shields are not a substitute for using low dose techniques and methodologies that reduce patient dose in CT. In summary, I would encourage the use of bismuth shielding as a methodology proven to reduce local dose that is easily implemented, inexpensive, and widely available. Bismuth shields are not perfect, they simply work. When AEC is used on Siemens scanners, the use of shields is not recommended. We agree and extend this warning to Philips systems, which also update the tube current during the scan, responding to the bismuth shield by increasing the current and, hence, countering its intended benefit. For systems without real-time tube-current adjustment (e.g., GE and Toshiba), the prescribed image quality is not achieved. Hence, shields should not be used with any AEC techniques. Turning off AEC, however, is not advised; z-axis AEC reduces dose to the breast and improves image quality, especially in the shoulders and hips. Shields should not be used whenever quantitative assessment of CT numbers is needed. In Ref. [6] cited by Dr. Gould, the CT number was increased by about 10 HU in the center of the phantom (lung and heart), even though the shield was offset by 6 cm. Also, image noise was increased across the entire thorax. Both effects represent degradation of image quality and quantitative accuracy. A significant drawback of bismuth shields: wasting dose to the patient. We completely agree. Use of bismuth shields should be replaced by globally reducing the tube current; the overall number of photons reaching the detector, and hence image noise, is the same between the two methods when the reduction in tube current is matched to the dose reduction achieved by bismuth shielding. In reducing tube current, however, dose is reduced to the entire volume, not just the anterior surface, CT numbers remain accurate, and potential streak artifacts are avoided. In conclusion, we find no compelling reason to use bismuth shields and, as Dr. Gould assists us in pointing out, many reasons to avoid them. Noise is never constant within a clinical image just as dose is not uniform. Bismuth shields do increase noise within the scan volume but not uniformly, and it is difficult within the complexity of the anatomy to detect this increase. Since use of an AEC mechanism is desirable (and shields are not a substitute), adjustments to the milliamperes proposed as an alternative can be difficult to implement and, furthermore, will result in degradation of the images globally, not regionally. When more than the chest is imaged in a single acquisition (e.g., chest and abdomen), every image will be degraded by a lowering of the AEC-determined milliamperes, not just those through the chest. As noted, bismuth shields should not be used on all scanners or in all applications. If the shields have been positioned correctly, artifacts are usually not noticeable or are minimal inside the rib cage at typical window and level settings used for viewing CT chest images. To my knowledge, there is not a single report of a missed diagnosis due to bismuth shields! I encourage readers to look at some clinical images where shields have been used and judge their quality. Positioning and care of bismuth shields is neither difficult nor time consuming. It does not interfere with or slow workflow. Training technologists in the use of shields is important but is neither difficult nor extensive. When used correctly, shields do not cause the “phototiming” algorithm to run amok. Notwithstanding the drawbacks, bismuth shielding is an easily implemented, inexpensive and effective method to reduce breast dose.