Assessment of Commercially Available In-plane Bismuth Breast Shields for Clinical Use in Patients Undergoing Thoracic Computed Tomography

Hong Kong J Radiol 2021 Jun;24(2):108-15

 

 

¹ Student Research Committee, Dezful University of Medical Sciences, Dezful, Iran

² Department of Medical Engineering, Islamic Azad University, Dezful Branch, Iran

³ Department of Medical Physics, Lorestan University of Medical Sciences, Khorramabad, Iran

 

  Full Article in PDF

Computed tomography (CT) is an indispensable diagnostic tool, providing cross-sectional and 3-dimensional images of anatomical structures with exquisite anatomical detail.[1] During the last few years, the number of CT examinations performed has steadily increased.[2] [3] In 1980, three million CT scans were performed in the United States.[4] This figure expanded to 57 million in 2000,[5] 62 million in 2007,[4] and 85 million in 2011.[6] A United States study in 2009 found that although CT accounts for 11% of all radiological examinations, it is responsible for 75.4% of the overall population dose.[7] Concerns over exposure and overutilisation of CT have resulted in several publications on the potential risk of detrimental health effects.[8] [9] These studies have linked CT with increasing the risk of radiation-induced carcinogenesis, especially when radiosensitive tissues are within the scan field.

 

Thoracic CT is a common examination that contributes to radiation exposure to the breasts.[10] During thoracic CT, the breasts receive approximately 17 to 22 mGy of radiation.[11] [12] This is particularly due to their superficial position that allows the breasts to be exposed to low-energy scattered photons.[13] The breast tissue is highly radiosensitive. Delivery of as little as 10 mGy to a young woman is reported to increase the risk of radiationinduced breast cancer by 13.6%.[3] Thoracic CT is intended to evaluate the lung parenchyma and mediastinum, and often the breast is not under diagnostic evaluation.[3] [10] Therefore, it is necessary that the radiation dose be kept as low as reasonably achievable.[14]

 

The in-plane bismuth breast shield has shown to be effective at reducing radiation exposure to the breasts during thoracic CT.[3] [15] [16] Dose reductions of 26% to 61% have been reported in both phantom and clinical studies using these shields.[5] [10] [16] [17] However, some drawbacks, such as introduction of image noise, streak and beam-hardening artefacts, and changes in CT number (in Hounsfield units [HU]) of the images have been concerns.[15] [18] [19] [20] It is suggested that placement of a 1-cm thickness of foam or cotton between the shield and the breast surface can reduce image noise.[21] Despite variations in the dose reduction levels, there is agreement in the literature on the potential radiation dose reduction of bismuth shields, but their effect on image quality remains controversial. Moreover, insufficient data exist regarding the influence of the shield-to-breast distance on image quality during thoracic CT.

 

The first aim of this study was to assess quantitatively the effects on image noise and CT number changes in image data following placement of 0-, 1-, 2-, and 3-cm shield-to-phantom spacers in a homogenous body phantom. The second aim was to assess dose reduction achieved by the shield and a qualitative assessment of image quality based on image criteria adopted from the European guidelines on quality criteria for thoracic CT.

 

A 32-cm cylindrical body water phantom (GE Healthcare, Milwaukee [WI], US) was positioned at the isocentre of a 16-slice GE CT scanner (BrightSpeed, GE Healthcare) and a scanogram was obtained to plan the scanning range. The phantom was scanned using a standard protocol routinely used for adult thoracic CT (120 kVp, 100 effective mAs, 6-mm section thickness, 0.5 s/gantry rotation, 27-mm table increment per rotation). A commercially available 0.06-mm lead equivalent radioprotective bismuth breast shield (F&L Medical Products Co., Vandergrift [PA], US) was placed directly upon the anterior surface of the phantom and a second scan was acquired. Subsequent scans were acquired after placing a 1-, 2-, and 3-cm polyurethane foam spacer between the shield and the anterior surface of the phantom (Figure 1). Identical scan parameters were used for all acquisitions (Figure 2).

 

Figure 1. Bismuth breast shield (arrow) and interval spacer (arrow head) in the homogenous cylindrical body water phantom (dashed arrow).

 

Figure 2. Axial scans of a 32-cm cylindrical body water phantom in mediastinal (top row) and lung (bottom row) windows. The use of the shield significantly increased image noise and streak artefacts, when directly positioned upon the phantom surface or when there was 1-cm shield-to-phantom distance (arrows). 3-cm shield-to-phantom distance resulted in lower image noise and artefact by shifting the artefacts outside the phantom surface.

 

To determine the influence of the shield and foam spacers on image noise and HU variation, in each session, we applied region of interest (ROI) methodology on three consecutive axial slices. To obtain reliable results, three sets of identical scans were obtained. Five circular ROIs with areas of 1.5 cm² were applied to each axial section at the 12, 3, 6, and 9 o’clock positions plus one ROI in the centre. The standard deviation of the density of ROIs was considered as a quantitative analysis of image noise. The mean HUs were recorded to determine variations caused by the shield.

 

Following approval from the university ethics committee (approval number: IR.DUMS.REC.1397.053), 180 female patients (≥18 years) scheduled to undergo a thoracic CT at our institution were recruited into the study. Patients were considered eligible for inclusion if they could follow the requirements of the study for standard positioning (supine and arms above the head) and had signed an informed consent form to participate in the study as assigned by the ethics board. All emergency studies and patients with unilateral or bilateral mastectomies were excluded from the study. Based upon our quantitative assessment of image noise in the phantom study, we found that by using a 3-cm shield-to-phantom distance, there was no significant increase in noise or HU of the phantom images (except the anterior part of the phantom) as compared to the case where no shielding is applied. Therefore, we followed this strategy in the patient study.

 

Thermoluminescent dosimeters (TLDs) [GR-200, Hangzhou Freqcontrol Electronic Technology Ltd., China] were used for dose measurements. Before the study, the TLDs were calibrated. Initially, all TLDs were simultaneously irradiated with the same dose of Cobalt-60 and then read out by a Harshaw 3500 TLD reader (Harshaw, Solon [OH], US) and their element correction coefficients were calculated. TLDs were divided into 15 batches and exposed together with a 3-cm³ Radcal ionisation chamber (Radcal Corp., Monrovia, [CA], US) at different doses in a diagnostic X-ray unit at 120 kV tube voltage. Following this, TLDs were read out again and a calibration curve was generated to convert the TLD charge in nanocoulombs to absorbed dose in mGy. Before and after each use, TLDs were annealed with a standard annealing regime recommended by the manufacturer (245°C for 10 minutes).[22] [23] To prevent the probable physical and chemical damage during the dosimetry process, each TLD batch was placed in a thin plastic bag. Throughout the study, three TLDs were used as controls to measure background radiation. Patients were positioned at the isocentre of the CT scanner in the supine position. Images were acquired from the thoracic inlet to the adrenal glands. To mark the approximate adrenal region, the shadow of the kidneys was used as a guide.[16] Following this, four fresh TLDs were carefully placed on each breast (around the nipple since it is relatively flat in supine position). A radioprotective bismuth breast shield with a 3-cm shield-to-breast foam spacer was placed over the left breast so that it covered the entirety of the breast and the TLDs. The right breast remained non-shielded. The craniocaudal scan was performed on the basis of the scanogram. The same scanner and identical scan parameters were used in both the phantom and clinical studies.

 

Three expert radiologists with a mean experience of 6 ± 2.3 years visually assessed the quality of patient images based on image criteria adopted from the European guidelines on quality criteria for thoracic CT (Table 1).[24] Initially, the picture archiving and communication system of the hospital was retrospectively investigated to identify a reference thoracic CT (non-shielded) in which each image quality criterion was consistent with the European guidelines described in Table 1. After obtaining one thoracic image as reference, all 180 thoracic image datasets of the clinical study were divided into left (shielded) and right (non-shielded) sections in the 2-dimensional axial view as shown in Figure 3. Each image quality criterion in each section was compared by the identical criterion of the reference image and graded as follows: (1) quality much lower than reference image and diagnostically unacceptable; (2) quality lower than reference image but diagnostically acceptable; (3) quality equal to reference image, and (4) quality better than reference image.

 

Table 1. European guidelines on quality criteria for thoracic computed tomography.

 

Figure 3. Thoracic computed tomography of a 31-year-old female patient in (a) lung and (b) mediastinal windows. The left breast was shielded using 3-cm shield-to-breast distance. The scan quality is fully diagnostic in both the shielded and non-shielded sections.

 

Data were entered into spreadsheet software (Excel, Microsoft Inc., Redmond [WA], US) and statistical analysis was performed using a statistical package SPSS (Windows version 20.0; IBM Corp, Armonk [NY], US). The normality of the data was assessed using the Kolmogorov–Smirnov test. To compare the HU and image noise (standard deviation of HUs) of the non-shielded phantom with 0-, 1-, 2-, and 3-cm shield-to-phantom distances in each location (12, 3, 6, and 9 o’clock, and centre), the Scheffe test (post hoc) was used with alpha level of 0.05 and a confidence interval of 95%. Student’s t test was used to compare the radiation dose received by the breasts. A paired t test was used to compare each image criterion of the shielded and non-shielded sections of the thoracic images in terms of image quality. A p value <0.05 was considered statistically significant.

 

The influence of the shield and the different shield-to-phantom distances on the quantitative image noise and HU at the anterior, lateral, posterior, and central regions of the phantom images is summarised in Tables 2 and 3. The increasing noise and HU of the shielded images was progressively more pronounced at the anterior portion of the phantom compared to the lateral, posterior, and central regions. Increasing shield-to-phantom distance resulted in lower image noise and HU variation. There was no statistically significant difference between shielded and non-shielded images in terms of image noise at all phantom regions when there was a 3-cm shield-to-phantom distance (all p > 0.05). The shield increased HU of the phantom images at all phantom regions (except the posterior region) without a spacer and with all spacer thicknesses. Streak artefacts were noted with no spacing and with a 1-cm shield-to-phantom distance (Figure 2).

 

Table 2. Effect of the shield-to-phantom distance on image noise (standard deviation) at the anterior (12 o’clock), lateral (3 and 9 o’clock), posterior (6 o’clock), and central regions of a 32-cm homogenous body phantom.

 

Table 3. Effect of the shield-to-phantom distance on Hounsfield Units at the anterior (12 o’clock), lateral (3 and 9 o’clock), posterior (6 o’clock), and central regions of a 32-cm homogeneous body phantom.

 

In the patient study, the patients’ age ranged from 18 to 74 years (mean, 41.5 ± 15.7). The mean absorbed dose at the surface of shielded and non-shielded breasts was 13.6 ± 3.1 mGy and 24.04 ± 4.7 mGy, resulting in a 43.4% reduction in the breast dose (Figure 4). The mean image quality scores in the shielded and non-shielded sections of the thoracic images were 2.98 and 3.02 for the mediastinal windows and 2.94 and 2.98 for the lung windows, respectively (Figure 5) [p = 0.997]. All thoracic images were interpreted as diagnostically acceptable.

 

Figure 4. Radiation dose received by the left breast (shielded) and right breast (non-shielded). Standard deviations are shown as error bars.

 

Figure 5. Vermont Golf Association (VGA) scores of the shielded and non-shielded sections of the thoracic images in both the mediastinum and parenchymal windows. Standard deviations are shown as error bars.

 

The bismuth shield is reported to be an effective tool for reducing radiation exposure of the breast during thoracic CT.[16] However, some drawbacks such as increasing noise level and HU of the images, especially in the anterior thoracic region, have been reported.[15] [18] [19] [20]

 

In our quantitative assessment of image noise, we demonstrated that a bismuth shield with no spacer significantly increased noise and HU of the images at the anterior portion of the phantom. This result is commensurate with the previous literature.[15] [25] [26] Our result, however, is in contrast with that of Fricke et al,[12] who reported no quantitative change in image noise between shielded and non-shielded regions of the lung for ≤18-year-old patients. This discordance may be due to the fact that their study was performed on paediatric patients that are typically scanned at low radiation doses and therefore higher image noise was introduced, whereas, in our study, the phantom was scanned with a standard adult CT protocol that is associated with higher radiation dose and lower image noise.[15] Consistent with previous studies,[15] [27] [28] our study demonstrated that increasing shield-to-phantom distance lowered noise in the phantom images. When there was a 3-cm shield-to-phantom distance, there was no statistically significant difference between shielded and non-shielded images at all phantom regions in terms of image noise (all p > 0.05); however, except in the posterior region, the increase in HU was statistically significant in all other phantom regions (p < 0.001). In a similar study, Kalra et al[15] reported a significant HU increase at the anterior and central portions of the images, with 0-, 1-, 2- and 6-cm shield-to-phantom distances. Kim et al[27] reported a 19% to 40% image noise increase in the anterior lung with a 1-cm shield-to-patient distance. Similarly, Vollmar and Kalender[26] reported that bismuth shielding with no spacer increased image noise up to 40%. The commercially available bismuth breast shields have 1 cm of foam or cotton as a spacer between the shield and the patient’s breast. We found that a 3-cm shield-to-breast distance lowered image noise. However, the increase in HU remains a concern.

 

According to our results, the mean radiation dose delivered to the shielded breast was 13.6 mGy, which represents a 43.4% reduction in the patients’ breast dose (24.04 mGy vs 13.6 mGy; p < 0.001). This result is consistent with that of Yilmaz et al,[16] who reported a 40.5% reduction.

 

The assessment of image quality in the patient study revealed no significant statistical difference between images in the shielded and non-shielded sections of the thoracic images in both the mediastinal and lung windows (p = 0.362). We found no image criterion reduced to the level of a diagnostically unacceptable (score 1). All images were interpreted as diagnostically acceptable. This result is commensurate with previous clinical studies.[16] [29] 3-cm shield-to-breast distance effectively shifts the artefacts arising from the shield to the outside of the patient’s body. Previous studies used phantoms and/or two separated groups of shielded and non-shielded patients to assess image quality and radiation dose reduction by the shield. We assumed that our methodology may be more reliable than those studies due to the fact that each measurement in patients had its own control (the opposite breast).

 

There are also opinions siding against the use of bismuth shielding during chest CT examinations. The American Association of Physicists in Medicine has challenged bismuth shielding, citing compromised image quality and unpredictable and undesirable results when combining with automatic exposure control (AEC).[30] Similarly, the Society of Cardiovascular Computed Tomography has avoided bismuth shielding due to its ability to influence the accuracy of coronary calcification measurements by increasing HU of the images.[31] Moreover, it has been argued that the combination of the bismuth shield with AEC would result in overestimating the patients’ attenuation and, consequently, offsetting bismuth shield efficiency. Hence, if the shield is used in conjunction with AEC, the shield should be placed after acquiring the scanogram. In this situation, the desired image quality may be slightly compromised but it does not influence patient care. Moreover, it has been argued that there are other dose reduction technologies such as organ-based tube current modulation, global tube current reduction, and iterative reconstruction techniques that do not have the aforementioned drawbacks associated with bismuth shielding but offer similar or even higher dose reduction levels.[30]

 

According to this study, combining a bismuth shield with a 3-cm shield-to-breast distance could effectively reduce radiation exposure to the breast without quantitative and/or qualitative deterioration of image quality in terms of increasing noise and streak artefacts. However, increasing the HU at the anterior, lateral and central regions of the thorax remains a valid concern. According to Kalra et al,[15] increasing the shield-to-phantom distance up to 6 cm has also failed to eliminate this drawback. The accuracy of HU is crucial for diagnosis of some specific pathologies such as coronary artery disease, which depends upon exact calcium density measurement. Therefore, if the absolute accuracy of the HU is necessary, the use of shielding should be discouraged.

 

Combining a bismuth breast shield with a 3-cm spacer significantly reduced radiation exposure to the breast without qualitative or quantitative deterioration of the image quality in terms of image noise and streak artefacts. The bismuth shield was associated with increasing HU of the images, not only in the anterior thorax but also in the lateral and central regions. Therefore, when the absolute accuracy of HU is crucial, the use of bismuth breast shields should be discouraged.