A novel hardware duo of beam modulation and shielding to reduce scatter acquisition and dose in cone-beam breast CT

Abstract

Purpose:

In cone-beam breast CT, scattered photons form a large portion of the acquired signal,adversely impacting image quality throughout the frequency response of the imaging system. Prior simulation studies provided proof of concept for utilization of a hardware solution to prevent scatter acquisition. Here,we report the design, implementation, and characterization of an auxiliary apparatus of fluence modulation and scatter shielding that does indeed lead to projections with a reduced level of scatter.

Methods:

An apparatus was designed for permanent installation within an existing cone-beam CT system. The apparatus is composed of two primary assemblies: a “Fluence Modulator” (FM) and a “Scatter Shield” (SS). The design of the assemblies enables them to operate in synchrony during image acquisition, converting the sourced x-rays into a moving narrow beam. During a projection, this narrow beam sweeps the entire fan angle coverage of the imaging system. As the two assemblies are contingent on one another, their joint implementation is described in the singular as apparatus FM–SS. The FM and the SS assemblies are each comprised a metal housing, a sensory system, and a robotic system. A controller unit handles their relative movements. A series of comparative studies were conducted to evaluate the performance of a cone-beam CT system in two “modes” of operation: with and without FM–SS installed, and to compare the results of physical implementation with those previously simulated. The dynamic range requirements of the utilized detector in the cone-beam CT imaging system were first characterized, independent of the mode of operation. We then characterized and compared the spatial resolution of the imaging system with, and without, FM–SS. A physical breast phantom, representative of an average size breast, was developed and imaged. Actual differences in signal level obtained with, versus without, FM–SS were then compared to the expected level gains based on previously reported simulations. Following these initial assessments, the scatter acquisition in each projection in both modes of operation was investigated. Finally, as an initial study of the impact of FM–SS on radiation dose in an average size breast, a series of Monte Carlo simulations were coupled with physical measurements of air kerma, with and without FM–SS.

Results:

With implementation of FM-SS, the detector’s required dynamic range was reduced by a factor of 5.5. Substantial reduction in the acquisition of the scattered rays,by a factor of 5.1 was achieved. With the implementation of FM–SS, deposited dose was reduced by 27% in the studied breast.

Conclusions:

The disclosed implementation of FM–SS, within a cone-beam breast CT system, results in reduction of scatter-components in acquired projections, reduction of dose deposit to the breast, and relaxation of requirements for the detector’s dynamic range. Controlling or correcting for patient motion occurring during image acquisition remains an open problem to be solved prior to practical clinical usage of FM–SS cone-beam breast CT.

1 |. INTRODUCTION

Efforts to advance research and development in breast CT (bCT) technologies are fundamentally driven by the need of better solutions for living patients. Critical improvements, therefore, are those designed to minimize risk to patients while advancing the technology. In a recent communication,¹ it was demonstrated via theory and simulations, that the addition of a hardware apparatus to cone-beam renditions of bCT can lead to a reduction of dose to patient, while improving image quality by avoiding the acquisition of scattered photons in raw data. Herein the design and physical implementation of that apparatus is chronicled, and a replication of those simulation studies in actuality is performed.

Acquisition of scatter in raw data is detrimental to image quality. Recent work has demonstrated these adverse impacts occur throughout the response of the imaging system—within both low and high frequency components.² Prior approaches for scatter management within bCT have typically been based on either the correction or rejection of scatter. The objective of a scatter correction approach is remediation of the negative impacts after scatter has been acquired. Scatter rejection approaches, on the other hand, aim to reduce the acquisition of scattered photons in projections. Within each of these approaches, the different techniques developed and employed have had varying success.

In scatter correction methodologies, typically either the projection domain or image domain is targeted in isolation to address the effects of scatter contamination. Those focused on the projection domain, for example, include utilization of beam stopping arrays or strips^3–8 or beam passing arrays,^9,10 to assess the scatter distribution in projections. Strategies that target the image domain include sinogram-domain image processing, Monte Carlo (MC) simulation approaches^8,11–15 or purely postprocessing image-domain techniques.^16–19

Scatter rejection approaches, alternatively, typically focus on hardware and design alterations within the bCT system. One can reduce photon scatter contamination, for example, by simply increasing the gap between the detector and the breast.²⁰ For example, a synchrotron-radiation phase-contrast breast CT system has been developed that utilizes a system geometry that extends the gap between the breast and detector. The system also has an imager with a relatively small sensitive area that results in major reductions in scatter uptake in projections.^21,22 Utilization of differing hardware constituents is another type of scatter rejection approach taken. A series of studies on utilizing detector displacement techniques, for example, have reported a significant gain in reducing scatter in projections by reducing the imager’s sensitive area through either using a small detector, or collimating a portion of, or laterally shifting, the flat-panel detector and running multiple scans to sample the entire breast anatomy.^23,24 Using anti-scatter grids is another option for reducing the scatter acquisition.^25–29 Propagation-based phase-contrast breast CT is another approach that has achieved advances by combining the information gained from refraction and absorption of x-rays in breast to mediate the deterioration of contrast resolution in traditional renditions of bCT.^30,31

Among scatter correction and scatter rejection approaches, those targeting hardware modifications are less commonly pursued as it can be challenging to design and implement changes without introducing unwanted side effects. Changing the gap between detector and breast,for example, can result in decreased breast coverage and increase focal spot blur.^8,32,33 Usage of anti-scatter grids is typically prohibitive in a low-dose CT image acquisition technique such as bCT—because of the dose penalty associated with using anti-scatter grids. Additionally, technical difficulties arise from the need to correct the grid’s septal shadow in projections. Using laterally shifted or partially covered detectors in an image acquisition may lead to partial sampling of the field of view and prolonged scan time.

The methodology implemented here employs a holistic perspective and approach that focuses on additions to hardware in cone-beam bCT such that they account for the patient and their specific breast anatomy. In bCT, the path lengths of x-ray beams vary greatly across the fan and cone angles.³⁴ For instance, the diameter of an average-sized breast at its posterior region is 14 cm³⁵ The diameter gets smaller moving anteriorly. Moreover, the distribution of the fibroglandular tissue is not homogenous throughout the breast, as the glandular distribution is higher at central regions compared to peripheral regions.³⁶ Ideally, then, different parts of the breast should be measured differently—according to the anatomical specificities of the organ. Doing so would prevent delivering excessive dose to the breast’s anterior and peripheral regions and suboptimal measurement of the central and posterior regions. Two approaches have been described in the literature to accomplish this.

Most well investigated is usage of a bowtie attenuator between the x-ray source and the breast.^37–39 This technique is commonly used in clinical whole body CT scanners, and several groups have investigated this solution in bCT with varying degrees of complexity. This methodology succeeds in both reducing radiation dose delivered to the breast as well as relaxing the dynamic range requirements of the detector.³⁷ Drawbacks arise, however, due to the attenuation property of the bowtie filters. Of particular undesirable consequence are that: (a) a pre-patient attenuator (filter) results in an increase in the scatter radiation even before beam interacts with the breast tissue, (b) the quality of the beam incident on the breast is not consistent, and (c) that this beam hardening is amplified at the peripheries of the breast. Because the attenuation pathlength is shorter at the peripheries of the breast, this region could be measured with a relatively softer beam than is used for measurement of more central regions. Using a bowtie filter, however, leads in the exact opposite effect: it results in a softer central beam where the filter’s thickness is lesser, and a harder peripheral beam where its thickness is greater. Applied to the case of the breast—the central part of the breast, where the attenuation pathlength is greater, gets measured with a softer beam than the peripheral regions where the attenuation pathlength is shorter. This scenario is problematic as it leads to higher dose deposition in the central parts of the breast and degraded contrast resolution at the periphery of the breast.^37,40

Although the primary objective of an attenuator is to equalize the attenuation path length of different parts of the beam, an alternative, recently introduced strategy takes the approach of spatially modulating photons, rather than filtering them.¹ The specific manner in which this modulation is achieved is dependent on the anatomical specificities of the breast as follows: thicker parts of breast are measured with a larger number of x-rays, thinner parts fewer. In other words, the x-ray fluence is spatially modulated according to the breast dimensions. Implementation of this technique yields the benefits of an attenuator, but not its limitations. In this methodology, an auxiliary unit called Fluence Modulator (FM)—a pre-patient structure, installed between the source and breast—is added to the CT system. FM unit installation does not result in an added scatter in projections, nor does it spatially impact the quality of the beam. Further, FM is coupled with a postpatient unit (called Scatter-Shielding (SS) unit) to reduce the scatter acquisition in projections.

Both FM and SS assemblies have been briefly introduced above and described in detail in a recent publication.¹ This work describes the approach and results of their joint implementation, herein after referred to as the joint apparatus FM–SS. In this study, we hypothesized that implementation of the novel hardware and methodology herein described would result in reduction of scatter acquisition and dose. We report on the direct effects of FM–SS implementation on radiation dose, therefore, in addition to the direct effects on scatter acquisition.

2 |. METHODS AND MATERIALS

First described is the FM–SS design and the image acquisition methodology. This is followed by report of the resultant detector dynamic range, spatial resolution, and radiation dose characterization of the system. Experimental quantification of the scatter acquisition in projections, with and without the installation of FM–SS, is then described.

2.1 |. Theory of operation

As shown in Figure 1, FM–SS comprised the pre-patient FM assembly and the postpatient SS assembly, jointly installed in a cone-beam bCT system. Both FM and SS are individually comprised of two x-ray blocking sheets atop a rotating drum. The positioning of the sheets in each assembly within the bCT system, in reference to one another, results in the formation of a “Window” relative to the entire FM–SS apparatus, through which only a narrow beam of x-rays is able to pass. The size of the narrow beam received by the flat-panel detector is given by

$$w=J\times \beta ,$$ (1)

FIGURE 1.

(a) Planar view diagram of FM-SS apparatus, (b) planar diagram of the FM unit (note that all x-rays are blocked by the FM sheets except for the ones aligned with the FM Window), (c)–(e) the beam received at the detector at three instances of time (note the change in the width of the received narrow beam at the detector as pointed out by black arrows under each subfigure)

where β is the narrow beam’s fan angle and J is the distance of the receiving part of the detector from the source. Parameter β depends on the maximum (parametrized as η) and minimum (ϑ) angular distances of the narrow beam from the central ray. As illustrated in Figure 1b, β can be calculated as

$$\beta =\eta -\vartheta -\delta .$$ (2)

In Equation (2), parameter δ is dependent on the thickness of FM’s constituent sheets (parameter t), radial distance (parameter r) of the sheets from FM’s axis of rotation, the angular size of the FM Window (parameter γ), and the distance (parameter C) between axis of rotation of the FM module and the focal spot. The structure of the FM sheets is determined during fabrication. Therefore, t, γ, and r are constant numbers. However, by adjusting C, one can change the x-ray fluence introduced to the breast depending on its size. For instance, in the case of a small breast, one can increase C to reduce the coverage of the modulated beam.

The above-described adjustments are performed prior to the image acquisition. During a scan, the only moving parts of the FM–SS apparatus are the rotations of the FM and SS units around their respective axes of rotation. As an outcome, the fluence received at the detector changes during a single projection, as shown in Figure 1c–e. At the edges of the breast, the size of the narrow beam is small (see Figure 1c,e). At the center of the breast where the attenuation pathlength is maximized, the size of the narrow beam is maximized (see Figure 1d). Given that the FM and SS units rotate at a constant speed, this varying beam width results in measurement of different parts of the breast with different number of photons, as follows. The central parts of the breast, with longer attenuation pathlength, are measured with a higher number of photons than are the peripheral parts of the breast that have shorter attenuation pathlengths. This is fundamentally different from how beam attenuators (such as bowtie filters, discussed in detail above) operate. Here, the objective is not to equalize the attenuation pathlength, but to measure different parts of the breast with different number of x-rays.

The SS sheets utilize significant curvature in their design, as shown in Figure 1a and described in detail in Appendix B. As this is a nontrivial departure from the flat sheet design outlined in a previous FM–SS concept publication,¹ the rationale for this choice is described. Ultimately, curved rather than flat sheets improve the mechanical stability and integrity of the SS. As outlined in Section 2.3, the rotational speed of the FM and SS units is several orders of magnitude faster than that of the imaging system’s gantry. Flat SS sheets, rotated at this high speed might generate vibrations on the gantry during an image acquisition, negatively impacting the quality of the outcome images. Curved sheets prevent this from occurring.

The FM–SS bCT method of imaging is pulsed fluoroscopy. Refer to Appendix A for additional detail.

2.2 |. Mechanical design

The geometry of the imaging system is shown in Figure 2. The distance between the axis of rotation of the FM assembly and the focal spot is set to be variable to accommodate different breast sizes. The axis of rotation of the SS assembly coincides with the axis of rotation of the gantry, 520 mm from the focal spot (parametrized as SIC in Figure 2). The flat panel detector is positioned 500 mm from the gantry’s axis of rotation (parametrized as SID in Figure 2). The dimensions of the FM sheets were based on the bCT imaging system’s geometry to cover the upper and lower bounds of the generated cone beam. The height and width of the FM sheets were no less than 95 mm and 50 mm, respectively. The height and width of the SS sheets were set to be no less than 200 mm and 190 mm, respectively. This figure also outlines the components constituting each assembly. The FM assembly consists of two sheets for modulating the x-ray fluence and complementary supporting structures that hold the sheets in a desired location and provide sliding and rotation motions for them. The rotation motion is generated by a motorized rotating stage. The motion is transferred to the sheets through a sheet supporting plate. The motorized rotating stage, fixed on a stage supporting plate, is connected to a positioning slide by which the location of FM with respect to the focal spot can vary in the longitudinal direction. A cantilever beam serving as a support for the FM components attaches the entire assembly to the bCT scan gantry. The mechanical stability of the beam is augmented using a wedge that supports the beam from underneath. The SS assembly similarly comprised two sheets and their supporting structures. The sheets prevent scattered photons from reaching the detector,while the complementary supporting structures both hold the sheets in the desired location and provide rotational motions for them. To initiate an exposure in each projection, the angular position of the sheets is traced by a signal. This signal is provided by a Hall effect sensor attached to the sheet supporting plate.

FIGURE 2.

Perspective views of the FM–SS apparatus, depicting the geometry of the imaging system. The geometrical details are outlined on top. The details of each assembly are shown in the bottom subfigure. Isolated exploded view of the FM assembly is shown in (a); the SS assembly in (b)

Materials for each component were selected based on the properties required for function and guided by machining constraints. The FM sheets are made of tungsten–nickel–copper alloy with greater than 90% tungsten purity (the tungsten sheets were supplied by Midwest Tungsten, Willowbrook, IL, and machined in house). Tungsten was selected for its excellent collimation property. Nickel and copper provide corrosion resistance in the final product and increase material strength. Increased strength facilitates a smoother machining process via Electrical Discharge Machining (EDM). The SS sheets had reduced collimation property requirements and were therefore made of 4 mm thick stainless steel. Stainless steel is a relatively lighter material that places lower strain on the mechanics and dynamics of the rotary stage that the SS sheets are installed on. Stainless steel is also highly machinable and corrosion resistant.

A detailed design model description of the FM and SS sheets is provided in Appendix B.

2.3 |. Image acquisition methodology

The described FM–SS apparatus was added as an auxiliary unit to a cone-beam prototype bCT system.³³ In this system, the imager is a PAXSCAN 4030CB flat panel detector (Varex Imaging, Salt Lake City, UT). The detector was set to work at its native pitch (0.194 mm) throughout this study. An M-1500 pulsed x-ray tube (Varex Imaging, Salt Lake City, UT) was mounted to a horizontally oriented rotating gantry, controlled by a Yaskawa Servo Stepping Motor (Yaskawa America Inc., Waukegan, IL). The source was set to work at 60 kV with 0.2 mm Cu additional filtration. A CMP 200 x-ray generator (CPI Inc., Ontario, Canada) was coupled with the tube. Cables for the tube and detector were routed through a spiral cable guide laid on the floor of the scanner. Two X-RSB rotary stages were used for the FM and SS assemblies (Zaber Technologies Inc., Vancouver, British Columbia, Canada). The controls were entirely developed in house (Malcova LLC, Baltimore, MD).

In each scan procedure, three hundred projections were acquired over 360°.In the two described operation modes—one with and one without FM–SS—the tube is set to work at pulsed fluoroscopy mode.

1. Without FM–SS: For each projection,each period was 133 ms long, 6 m of which are used x-ray exposure, leading to an overall scan time of 40 s. The tube set to operate at 42 mA, corresponding to a 5.6 mGy dose (refer to Section 2.7 for more details).

2. With FM-SS: The mA and exposure time values were found empirically to satisfy the primary objective in each projection given the limitations of the tube heat unit and detector readout rate.

The primary objective is as follows: the signal level at the vicinity of the central ray’s incident point on detector must stay the same in both operation modes (with and without FM-SS). Therefore, a region of interest of 2 cm × 2 cm box around the central point was selected and the mA and exposure time were adjusted to satisfy this objective.

As described in Appendix A. A, the source is in “Expose OFF” mode by default. For x-ray exposure to commence, a line of sight between the source and detector must be achieved.This occurs only when the FM window and SS window align. With line of sight achieved, the source transitions into “Expose ON” mode. A Hall effect sensor is positioned at this point, as shown in Figure 2, to initiate an exposure. This pulse is routed to the x-ray generator. In this state, x-rays are generated at the source,collimated at the FM unit to form a narrow beam of x-rays, and projected onto the FOV. The beam transitions and interacts with the breast tissue placed within the FOV. The scattered photons are absorbed by the SS sheets. The primary photons that simply transition through the breast go through the SS Window and finally absorbed by the sensitive area of the detector panel. The exposure continues until the line of sight between the source and detector through the FM and SS windows disappears, at which point the source transitions back to the Expose OFF state and the collected projection is transferred to the host controller.

In this mode, each projection was set to 250 ms. Similar to the “Without FM-SS” mode operation, 300 projections were made over 360°. Therefore, the FM and SS units rotate synchronously, in velocity and phase, at 240 revolutions per minute. The overall scan time is 75 s. The synchronization of these modules occurs at the beginning of a scan, taking less than 3 s. The exposure occurs in 15 ms of each projection at 130 mA. These parameters satisfy the primary objective as stated above.

2.4 |. Spatial resolution measurement

The exposure time is each projection is longer with FM–SS versus without. Longer exposure time may result in motion blurring and a reduction in spatial resolution of the imaging system. We hypothesized that both blurring and impacts on resolution can be mitigated by the combined approach of prolonging the scan time and reducing the gantry speed, as we described in Section 2.3. To test this hypothesis, a comparative study was conducted. The imaging system’s modulation transfer function (MTF) was measured with and without FM–SS. The previously reported methodology was used to measure system MTF.³³ Briefly, this methodology is as follows. A 30-μm tungsten wire was positioned within the scanner at a 7-cm radial distance from the gantry’s isocenter. The wire was then scanned using the techniques outlined in Section II.C. A total of 10 reconstructed slices,each with a thickness of 0.15 mm, were averaged to reduce noise. Due to the subtlety (in size and shape) of the tungsten wire, the location of the generated point spread function was subject to minor planar translations. Therefore, in each region of interest within a coronal slice that contained the wire signal, the center of mass was found, and an alignment was enforced in the slices.The integration of the averaged point spread function generates the line spread function,which in turn Fourier transformed to yield the system MTF.

2.5 |. Phantom study

A realistic breast phantom was developed for use in this study, modeled based on those described recently.⁴¹ In this (referenced) study, a large number of bCT data sets were utilized to generate realistic breast phantoms of various dimensions. The selected mathematical model utilized herein derives from the V4 phantom model in the referenced publication, representing an average size breast. A modification was made at the posterior (furthest from the nipple) part of the phantom to incorporate a handle, enabling the suspended positioning of the breast in the imaging system’s FOV. The phantom was developed using Ultra-high-molecular-weight polyethylene (UHMW) as the fabrication material as a surrogate to adipose tissue. A view of the phantom model and a photograph of the implemented phantom are shown in Figure 3.

FIGURE 3.

Design and physical implementation of the breast phantom used in Study E. Design side view (a), perspective view (b). Photograph of the developed phantom (c). Note that the length unit in (a) is mm

2.6 |. Scatter measurement setup

We measured the scatter-to-primary-ratio (SPR) using the commonly utilized aperture method outlined in literature.¹⁵ The basic components of this device are two 152 mm (6 inches) × 152 mm × 1.6 mm lead sheets, sandwiched between two transparent plastic (Acrylic) sheets each with a thickness of 1 mm. The placement of the sheets is such that a 152 mm long, 2 mm horizontal strip is formed between them. With this tool between the source and breast phantom, we made exposures and acquired projections. The acquired projection contains a strip of detected x-rays, almost entirely formed of the primary photons (note that only a small part of the breast is exposed). The scatter component of a projection can be measured by subtracting this primary component from a projection acquired without this tool in place. Therefore, SPR can be calculated by dividing the scatter component by the primary component. To suppress the noise in scatter estimation, 100 projections were acquired in each operation mode (with and without FM–SS, with and without the scatter measurement tool in place). The acquired projections were flat-fielded. The final primary + scatter (P + S) and primary (P) images were calculated by averaging the corresponding projections. SPR was calculated as

$$SPR=\frac{{\left(P+S\right)}_{\text{strip}}-{\left(P\right)}_{\text{strip}}}{{\left(P\right)}_{\text{strip}}}.$$ (3)

2.7 |. Spatial distribution and magnitude of dose

To characterize the scanner’s x-ray beam, the half value layer (HVL) was measured using type 1100 aluminum. An ionization chamber (a 0.6 cc 10 × 6-0.6CT ionization chamber coupled with an ADDM+ Accu-Dose+Digitizer, Radcal, Monrovia, CA) was placed along the central beam of the imaging system at the isocenter. Three scans were performed without the FM–SS in place, and the average value was used in the following steps. The previously reported tungsten anode spectral model using interpolating cubic splines for application in bCT⁴² (TASMICS_bCT) was then used in conjunction with the measured air kerma, the HVL value, and breast model to find the glandular dose value.

In the case of FM–SS, we measured the air kerma in air using the same strategy outlined above. The measured air kerma, however, cannot be directly used in TASMICS_bCT to find the radiation dose. Clearly, FM–SS is a form of beam modulation that spatially alters the x-ray fluence exposed on the breast. Therefore, TASMICS_bCT is not the proper model for dose calculation when FM-SS is utilized. We conducted a Monte Carlo simulation, outlined in the next paragraph,to compare the difference in spatial distribution and magnitudes of dose with and without FM-SS in place. With the exact value of dose calculated in the absence of FM–SS (following the steps outlined in the previous paragraph) and a comparative analysis of the dose deposition with and without FM-SS,we calculated the dose in the breast phantom when scanned using the FM-SS.

The Geant4 toolkit (10.04.patch-02) with the Geant4 physics package option 4 was utilized^43–45 Default cut-off values were used for all particles. Using the HVL, TASMICS_bCT was used to derive the spectral distribution of the beam. Note that FM spatially modulates photon quantity without altering beam quality. As a result, the quality of the beam incident on different parts of the breast remains consistent. This allows the same energy spectral distribution for the beam to be used for both with and without FM–SS cases. The breast was modeled as a prolate semi-ellipsoid with equatorial semi-axis of 7 cm and center to pole distance of 10 cm (corresponding to V4 breast phantom developed for this study), filled with homogenous breast tissue material with glandular density of 17% (the average breast density in general U.S. population⁴¹). For calculating the spatial distribution of dose, the breast volume was turned into a 256 × 256 × 128 matrix. In each case, 300 projections were made over 360°. In each view angle, monoenergetic photons were generated with energies in 1 keV increments from 5 keV to 60 keV. In each interaction of a photon with breast, the type of interaction was analyzed. If Compton scatter or photoelectric events occurred, the location of the interaction and deposited dose was recorded. Finally, the recorded dose images in each monoenergetic energy bin were weighted according to the bin size in the beam’s energy spectral distribution. An aggregation of the weighted dose images provided the final dose models with and without the FM–SS apparatus.

3 |. RESULTS

Views of the developed prototype and the individual components are shown in Figure 4.

FIGURE 4.

Photographs of the developed SS (a), FM (b), and the prototype (c). Note that the positioning and movements of the FM and SS assemblies ensure the alignment of the FM and SS windows, such that a narrow beam generated at the source can transition through the FM assembly, the breast, and the SS assembly, and reach the detector

3.1 |. Comparison of acquired projections with and without FM-SS

Figure 5 shows two projections in the absence of an object in the FOV, with and without the FM–SS apparatus. Without FM–SS, the entire field of view is exposed to the x-rays beam. In the case of FM–SS, however, the recorded projection is different. First, outside the prespecified fan angle coverage, the beam is attenuated to net zero (see the dark bands on the sides of Figure 5b). Second, the x-ray fluence delivered to different parts of the detector changes given the modulation generated by the FM sheets. This point is further illustrated on Figure 5c, where we show the profiles along the lines that cross over the central ray’s incident point on detector. As a reference, the theoretical (expected) profile of the detected signal is overlaid on this plot. The theoretical profile was derived during our simulation studies and was reported recently.¹

FIGURE 5.

Comparison between the air projections without (a) and with (b) the FM-SS apparatus. The profiles along the horizontal lines overlaid on the first two subfigures are shown in (c). A profile of the expected modulation—normalized to one—is shown in (c)

Similarly, examples of RAW projections captured with and without the FM–SS apparatus, when the breast phantom is positioned in the FOV, are shown in Figure 6. The phantom described in Section II.E was used in these projections. The same window-leveling settings were applied to both projections. Profiles along the horizontal line that passes through the central point are shown in Figure 6c. This figure highlights the relaxation of the detector dynamic range requirements when FM–SS is utilized. Note the high signal levels at the parts of the projection where the breast is not measured—at the left and right sides of the Figure 6a. At the center of the breast, in near alignment with the central ray of the imaging system, the detected signal is significantly less that what is detected outside or at the thinner parts of the breast. This can be appreciated by following the orange profile shown in Figure 6c. In the case of FM–SS, however, this large drop in the signal level is largely absent. If dynamic range is defined as the ratio of the intensity detected in the least and most attenuated entries in the line profiles, the dynamic range with FM–SS was decreased by a factor of three or more compared to the case without FM–SS.

FIGURE 6.

Comparison between the projections in the presence of the breast phantom. Projections made without and with the FM–SS apparatus are, respectively, shown in (a) and (b). The profiles along the horizontal lines overlaid on the first two subfigures are shown in (c). The line crosses over the incident point of the central ray on detector

3.2 |. Comparison of the system spatial resolution with and without FM–SS

The measured system MTFs, with and without FM–SS, are compared in Figure 7, highlighting the similar spatial resolution characteristic of the imaging system, regardless of using FM–SS.

FIGURE 7.

System MTF measured with and without FM–SS

3.3 |. Comparison of deposited dose with and without FM–SS

Using the technique described in Section 2.7, the HVL was measured to be 3.56 mm Al. The average recorded air kerma without FM-SS was 7.41 mGy with a standard deviation of 0.10 mGy. The recorded air kerma with FM-SS was 7.31 mGy with a standard deviation of 0.15 mGy. Assuming a 25% glandular density (the average density) for the breast phantom described in Section II.D, the average glandular dose without FM–SS is 5.60 mGy.

The spatial distribution of the dose is shown in Figure 8. The same window-level settings are used in all dose images to provide a qualitative comparison. The line profiles in Figure 8e,f provide more concrete evidence for the impact of FM-SS on dose. Two observations are in order. First, without FM–SS, there is a drop in the dose level from the center of the coronal slice to its periphery (see the orange line in Figure 8e). This profile flattens in the case of FM–SS. Using FM–SS, the energy deposition is homogenized throughout the coronal slices whereas without FM–SS, the periphery of the breast receives higher dose as it receives a higher x-ray fluence, all else being equal. Second, the magnitude of dose at the center of the coronal slices is less in the case of FM–SS. Third, a trend of increasing dose level is observed going anteriorly in both cases: with and without FM–SS. An overall tally of the dose throughout the breast reveals average glandular dose of 4.09 mGy in the case of FM-SS, a 27% reduction in dose.

FIGURE 8.

Spatial distribution of the dose with and without FM–SS. The coronal cross-sections are shown in (a) and (b), the sagittal cross sections in (c) and (d). The line profiles of the tallied does are shown in (e) and (f). The coronal cross-section corresponds to the central slice2

The trends observed in the coronal cross sections are persistent throughout the breast, as shown in Figure 9. Here, the dose profiles, in both cases of with and without FM–SS, are compared on a ridgeline plot. In this figure, each line is composed of half of the dose line profile with FM-SS and half of the dose line profile without FM-SS.

FIGURE 9.

Ridgeline plot of the deposited dose with and without FM–SS. Each horizontal line is a line profile across the coronal cross-section of the dose deposition volume. For comparative purposes, the left half of the profile is taken from the dose volume with FM–SS and appended by the profile from dose volume without FM–SS

3.4 |. Comparison of acquired scatter with and without FM-SS

The SPR along the horizontal lines shown in Figure 6 is illustrated in Figure 10. SPR was measured in both modes of operation, with and without FM–SS. Note the magnitude of difference between scatter acquisition in projections with and without FM–SS. For instance, at the part of the projection aligned with the thickest part of the breast, column 1050 in this case, there is as much scatter recorded as the primary (SPR approaches 1.0). At the periphery of the breast, the scatter portion of the received signal remains large (~0.60), constituting a nonnegligible portion of the overall received signal. In case of the FM–SS, however, the SPR at the center of the profile approaches 0.20, an ~80% reduction in scatter acquisition compared to unaltered beam case . The SPR decreases to ~0.15 at the peripheries of the breast, a ~60% reduction compared to the case without FM–SS.

FIGURE 10.

SPR in the acquired projections, with and without FM–SS, along a horizontal line that passes over the central point. A second-order trendline (black line) is overlaid on each set of SPR values for visualization purposes. Note that the SPR outside of the breast is not included in this graph as the values approach infinity in the case of FM–SS (primary signal outside of breast is zero)

4 |. DISCUSSION

Acquiring projections with minimum levels of scatter is essential to resolving one of the main limitations of cone-beam bCT—the subpar visibility of microcalcifications and fine fibroglandular tissue structures. Limitations on the visibility of small details that arise from bCT system designs with low spatial resolution has been well studied.^46–48 Less focused upon, despite being raised as a major issue a decade ago⁴⁹ has been the contribution of scattered radiation to poor visibility outcomes. In a recent study, we reported that under conditions of iso-dose and iso-spatial-resolution, a bCT design that leads to less scatter acquisition yields better microcalcification visibility.² Here, we chronicle achievements to date toward design and implementation of such a bCT system.

The results of a previously reported simulation study had suggested that the FM–SS design, as outlined in Section II.A, can dramatically reduce scatter uptake.¹ In this study, we hypothesized that (a) the FM–SS method of imaging could be developed and incorporated into a cone-beam bCT system, (b) the spatial resolution of the system would not be affected by this addition, (c) the radiation dose would reduce or at least remain the same, and most importantly, (d) scatter uptake could be reduced significantly.

In agreement with the preliminary simulation studies reported in a prior publication, the FM–SS apparatus is able to acquire projections formed of majority (more than 80%) primary photons. Implementation of this apparatus additionally leads to a reduction in the radiation dose delivered to the breast and a relaxation of the dynamic range requirements for the detector. These achievements are attained without sacrificing the spatial resolution of the imaging system.

As several studies have reported a higher than screening levels of dose in cone-beam renditions of breast CT,⁵⁰ a dose reduction of 27% achieved in an FM–SS enabled cone-beam bCT is a significant gain, one that can lower the dose to screening dose levels. In addition, the results shown in Figures 8 and 9 suggest that FM–SS exhibits a dose equalizing effect without manipulating the beam quality. This stems from a basic property of the FM structure—that the peripheries of a breast do not need to be measured with the same number of photons as its center, because the attenuation pathlength is shorter. The profiles shown in Figure 9 reveal that a similar trend—the consistent higher levels of dose in the case of imaging without FM-SS—is observed throughout the breast.

Of note is that the gains achieved with implementation of the FM–SS apparatus are independent of the detection technology. The projections and scatter analysis presented in this work were all made utilizing an energy integrating flat panel detector. The same methodology can be similarly effective in systems that utilize photon-counting detectors in cone beam or spiral CT geometries.

The results of the scatter measurement study—that the maximum SPR level is less than 0.2 (see Figure 10)—confirm a primary hypothesis of this study:with implementation of FM–SS in cone-beam bCT acquired scatter in a projection can be significantly reduced. Several studies have reported SPR levels of cone-beam bCT applications that extend well beyond the range reported here—falling between 0.8 and 1.6.^17,20 The current spiral CT rendition of bCT has been shown to have SPR of 0.4–0.6.^2,50 This system has limited cone-beam coverage that results in the reduction of the scatter uptake. However, the magnitude of the SPR is higher than the current implementation of FM–SS, which is a compound effect of shorter distance between the gantry’s axis of rotation and detector (~20 cm in the spiral bCT vs. ~50 cm in this study), and the softer beam quality (with an effective energy of 30 keV in spiral bCT vs. 36 keV in this study). Refer to Appendix C for a comparative analysis of SPR in five different designs for bCT.

In future studies, there is further room for improvement above and beyond the implementation described here. The fluence modulation strategy of this study, for example, leads only to modulation of fluence in one direction—coronally. In bCT, a pendent breast positioned at the gantry’s isocenter can be modeled as a prolate half spheroid. The path length of the photons transitioning through the peripheral regions of a pendant breast is shorter than the ones transitioning through the central regions. In an optimal setting, the x-ray fluence should be modulated not only coronally (as described herein), but also laterally.

Although herein is chronicled a proof-of-concept implementation of FM–SS, this study is limited in its scope. Before deploying this system for clinical use, for example, there are several challenges to be addressed.

In designing a system for clinical use, scan time is a critical factor. The scan time in FM–SS mode of operation is substantially longer than in routine cone-beam mode of operation. There are two main contributing factors to the longer scan time. First, FM–SS at its core is a robotic solution to the limitations of cone-beam bCT. In the current implementation, we operate the FM and SS units at 240 RPM. In planning for a faster rotational velocity, modifications to the structure of these units need to be made to ensure safety and mechanical integrity of the apparatus. Second, the longer exposure time may cause excessive heat loading on the x-ray tube target, shortening the lifespan of the source. In this study, a 10 kW x-ray tube was utilized. Though excessive loading was not experienced (running the source at FM-SS mode of operation for 10 consecutive scans in 60 min using the technique factors outlined in Section II.C leads to a heating of 15% HU), this selection of the x-ray source may need to be modified to ensure the longevity of the tube in clinical settings.

When image acquisition is performed with a patient, the 40-second scan time in the case of FM-SS will likely result in images impacted by involuntary patient motion arising from breathing. Requiring every patient to hold her breath for this amount of time is an impractical proposition. It is necessary, therefore, and a subject for future work, to develop an effective strategy for mitigation of the involuntary breathing motion of a patient during a scan. Fortunately, several research groups have focused on the general need for motion mitigation in bCT systems and proposed methods to do so have been reported in literature. The use of a breast holders as a means to immobilize the breast against the involuntary breathing motion, for example, has been investigated in both a breast micro-CT,⁵¹ and in a cone-beam bCT.⁵² Another alternative option to physical immobilization is tracking motion of the patient for later correction—a coupled technology and approach that is commonly used in radiation oncology applications.⁵³ The implications of both the desired number of projections (300 projections in this work) as well as the optimization of acquisition mode as stated above, demonstrate important decisions and trade-offs that need to be made in future studies. Investigating the use of breast immobilization to reduce motion artifacts, or an algorithmic process to track and mitigate motion blurring are approaches that would facilitate practical implementation of FM-SS in the clinic.

The clinical implications of the outlined technology are to be noted. The FM–SS is an auxiliary apparatus designed for cone-beam CT. Utilization of this apparatus leads to a reduction of both radiation dose and scatter uptake. FM–SS can be incorporated into a system without the need to change routine imaging protocols.

Two key hypotheses follow from these results achieved here. First, whether the reduced scatter uptake resulting from implementation of FM–SS leads to increased visibility of microcalfications.² Therefore, a clinical evaluation of microcalcification visibility is in order. Second, the scatter and dose reductions make way for bilateral imaging of breasts of a patient without a need for repositioning on the scanner, although the current realization of FM–SS must be combined with a thorough beam optimization effort to make bCT a viable imaging modality capable of simultaneous bilateral breast imaging. These and other relevant considerations are the subjects of our future studies.

5 |. CONCLUSION

The implementation of a combined dynamic fluence modulator and scatter shielding apparatus is presented and described. The FM–SS method of imaging provides significant reductions in both scatter uptake in projections and radiation dose to breast. These improvements are achieved without the need to change the energy spectral distribution of the sourced beam and are independent of the utilized x-ray detection technology.

APPENDIX

Appendix A

In this section, the timing of exposure during a projection is described.

The FM–SS bCT operates in pulsed fluoroscopic mode, as demonstrated in Figure A1. From the point of view of the focal spot, when a narrow beam is generated, it must pass through the windows created jointly by FM–SS. This means that FM and SS units should be synchronized such that they rotate at the same angular velocity and phase. In addition, the timing of initiating an x-ray exposure should be carefully controlled, to avoid either partial coverage of the breast, or overheating the tube. Let us consider a single projection. In the pulsed fluoroscopy method, the source is in “X-ray OFF” mode by default, as shown in Figure A1a. The placement of the sensory system must be determined such that the beginning of the exposure pulse coincides with the positioning of the FM and SS Window at the periphery of the breast (see Figure A1b). The length of the exposure should be adjusted such that the narrow beam sweeps the entire breast. As soon as the beam Window is broken (when the line of sight between the source and detector through the FM and SS Windows disappears), the exposure stops and the source defaults back to X-ray OFF state. This cycle continues until a desired number of projections is acquired.

Appendix B

In this section, the detailed description of the mechanical design of FM and SS sheets is provided.

The design constraints that have predominant roles in the functionality of the FM–SS apparatus are dictated by the characteristics of the FM sheets. This is elucidated in the detailed design model for the FM and SS sheets, shown in Figure A2.

The mathematical modeling and simulations of the FM–SS bCT method of imaging suggest that an inner radius of 30 mm, thickness of 15 mm, and window size of 10° are the optimal values of the FM sheets for imaging an average size breast of 14 cm in diameter and 10 cm in length (an average size breast). Hence, these values are adopted as the design constraints of the FM sheets (marked with blue lines in Figure A2a). The main requirements of the design were (i) the mass imbalance of sheets around the center of rotation, and (ii) the weight of the ensued design due to high density of Tungsten (density of 19300 kg/m³). To overcome these requirements, five design variables (θ₁, θ₂, θ₃, R₁, and R₂) were considered to define the remaining boundaries of the FM sheet profile (b_n,n = 1 to 5).The final values of the design parameters were determined by the following criteria:

• Design variable θ₁ defining the boundary θ₁ should be minimum to append least amount of material to the sheet, as this portion of the sheet is the primary source of imbalance. The value 20.25° added minimum amount of material necessary for providing mechanical integrity as well as room for placing a fastener.

• Design variable θ₂ defining the boundary b₂ should be 90°, as adding material below this threshold increases the imbalance.

• Design variable θ₃ defining the boundary b₃ should be 130° or less to leave sufficient space for the x-ray fluence for entering the space between sheets.

• Design variable R₁ defining the boundary b₄ should be minimum to append least amount of material to the sheet, as adding material to this section of the part increases the imbalance. The value 4 mm is selected to guarantee the mechanical integrity of this portion of the sheet as well as the ability of the sheets to block x-ray fluence.

FIGURE A1.

Cross-sections of the FM–SS apparatus at key moments in the acquisition of a projection—beginning with (a) and ending with (d)—outlining a pulsed fluoroscopic mode of imaging

FIGURE A2.

The design of the FM and SS sheets are, respectively, shown in top row (a–c) and bottom row (d–f). The design constraints and variables are introduced in (b and e). The terminology used in describing the sheets and the final values for the parameters introduced in (b) and (e) are shown in (c) and (f). The length unit is mm and the angular unit is degrees

Design variable R₂ defining the boundary b₅ should be large enough to balance the material added to the other section of the sheet.

Appendix C

We recenly reported the results of a comparative simulation study between the SPR in projections across five fully simulated bCT designs.² The findings, such as the one shown in Figure A3, highlights the varying levels of scatter uptakes in projections. This, in turn, results from the different choices made in the image acquisition geometries, utilized beam, and image acquisition protocols.

FIGURE A3.

Scatter buildup in projections in five fully simulated bCT systems. The SPR images are formed by recording the incident primary and scattered photons separately and dividing the resulting scatter-only projections by the primary-only projections. The profiles along the horizontal lines in each SPR image are compared in the above plot

DATA AVAILABILITY STATEMENT

Data available on request from the authors.