Intraoperative cone-beam and slot-beam CT: 3D image quality and dose with a slot collimator on the O-arm imaging system

Abstract

Purpose:

To characterize the 3D imaging performance and radiation dose for a prototype slot-beam configuration on an intraoperative O-arm^™ Surgical Imaging System (Medtronic Inc., Littleton, MA) and identify potential improvements in soft-tissue image quality for surgical interventions.

Methods:

A slot collimator was integrated with the O-arm^™ system for slot-beam axial CT. The collimator can be automatically actuated to provide 1.2° slot-beam longitudinal collimation. Cone-beam and slot-beam configurations were investigated with and without an antiscatter grid (12:1 grid ratio, 60 lines/cm). Dose, scatter, image noise, and soft-tissue contrast resolution were evaluated in quantitative phantoms for head and body configurations over a range of exposure levels (beam energy and mAs), with reconstruction performed via filtered-backprojection. Qualitative imaging performance across various anatomical sites and imaging tasks was assessed with anthropomorphic head, abdomen, and pelvis phantoms.

Results:

The dose for a slot-beam scan varied from 0.02–0.06 mGy/mAs for head protocols to 0.01–0.03 mGy/mAs for body protocols, yielding dose reduction by ~1/5 to 1/3 compared to cone-beam, owing to beam collimation and reduced x-ray scatter. The slot-beam provided an ~6–7× reduction in scatter-to-primary ratio (SPR) compared to the cone-beam, yielding SPR ~20–80% for head and body without the grid and ~7–30% with the grid. Compared to cone-beam scans at equivalent dose, slot-beam images exhibited an ~2.5× increase in soft-tissue contrast-to-noise ratio (CNR) for both grid and gridless configurations. For slot-beam scans, a further ~10–30% improvement in CNR was achieved when the grid was removed. Slot-beam imaging could benefit certain interventional scenarios in which improved visualization of soft tissues is required within a fairly narrow longitudinal region of interest (±7 mm in $z$)—for example, checking the completeness of tumor resection, preservation of adjacent anatomy, or detection of complications (e.g., hemorrhage). While preserving existing capabilities for fluoroscopy and cone-beam CT, slot-beam scanning could enhance the utility of intraoperative imaging and provide a useful mode for safety and validation checks in image-guided surgery.

Conclusions:

The 3D imaging performance and dose of a prototype slot-beam CT configuration on the O-arm^™ system was investigated. Substantial improvements in soft-tissue image quality and reduction in radiation dose are evident with the slot-beam configuration due to reduced x-ray scatter.

1. INTRODUCTION

Intraoperative cone-beam CT (CBCT) systems are seeing increased utilization in minimally invasive procedures, such as spinal¹ and intracranial^2,3 neurosurgery, orthopedic trauma,⁴ otolaryngology,⁵ cranialmaxillofacial,⁶ and thoracic⁷ surgery. CBCT offers 3D visualization of anatomy and surgical instruments in the operating room (OR) as an aid to navigation and has shown to improve surgical accuracy⁸ (e.g., target localization and instrumentation placement). Image quality is conventionally limited to visualization of high-contrast structures, including surgical tools, bone, and lung. Many procedures stress the need to visualize soft-tissue structures, including the localization of tumor masses, avoidance of critical structures, and detection of potential complications (e.g., hemorrhage). Among the challenges to intraoperative CBCT imaging performance are the radiation dose (which relates to image noise) and x-ray scatter (which relates to subject contrast as well as cupping and streak artifacts).

Since the emergence of CBCT in the early 2000s, numerous efforts have been demonstrated to address the influence of x-ray scatter. Algorithms for scatter estimation and correction represent a major area of interest in combination with any or all of these scatter rejection strategies, ranging from methods that aim to measure the scatter contributions in the projection data^9,10 to analytical^11–13 or Monte Carlo^14–17 modeling and deep learning^18,19 methods. In this work, we focus on the main physical considerations that relate to reducing or rejecting x-ray scatter, including the system geometry,^20,21 use of a bowtie filter^22–24 and/or antiscatter grid,^25–27 and selection of collimation.^28,29 System geometry tends to favor larger air gap to reduce x-ray scatter reaching the detector but must be balanced against increased focal spot blur.³⁰ A bowtie filter carries a number of potential benefits,³¹ but can challenge detector calibration and introduce susceptibility to artifacts associated with patient centering. Antiscatter grids can improve subject contrast, but carry a loss in primary radiation and increase noise and/or radiation exposure to the patient.²⁵ Prepatient beam collimation (in particular, limiting the longitudinal field of view, FOV_z) reduces scatter in a dose-efficient manner. The benefit of minimizing FOV_z in a manner that still covers structures of interest is well recognized in CBCT,^28,29 and the benefit of a “slot” beam is evident in various embodiments found in digital mammography,^32,33 chest radiography,³⁴ and long-length imaging.^35,36

Motivated by the potential benefits associated with slot-beam scanning demonstrated in previous and ongoing research, the work reported below implemented and evaluated a prototype slot-beam configuration on the O-arm^™ system (Medtronic, Littleton, MA) for 3D intraoperative imaging, leveraging a slot collimator recently introduced for long-length, slot-scanning tomosynthesis.³⁶ The slot collimator reduces the beam width in the longitudinal direction, opening the possibility of slot-beam CT over a limited FOV_z with reduced x-ray scatter, reduced patient dose, and improved soft-tissue image quality. There is also the potential for future implementations in which the slot-beam configuration is used in a helical acquisition mode. The imaging performance and dose for cone-beam and slot-beam configurations with and without an antiscatter grid were quantitatively evaluated in terms of dose, scatter-to-primary ratio (SPR), image noise, and soft-tissue contrast resolution. Furthermore, the results were qualitatively demonstrated in anthropomorphic phantoms exhibiting a variety of simulated low-contrast anatomical structures pertinent to potential applications in head and body interventions.

2. MATERIALS AND METHODS

2.1. Imaging system configurations

The experimental setup is illustrated in Figure 1. As detailed in previous studies,^37,38 the O-arm^™ system features a breakable circular gantry containing a rotating anode x-ray tube (A132, Varian Medical Systems, Salt Lake City, UT) and a flat-panel detector (PaxScan 4030D, Varex, Palo Alto, CA) with a sourceto-axis distance of 64.8 cm and an axis-to-detector distance of 52.0 cm. The system geometry with the 40 × 30 cm² detector forms a 19.4° fan angle $\left({\varphi }_{fan}\right)$ and 14.6° cone angle $\left({\varphi }_{\text{cone}}\right)$ giving a 22×22×17 cm³ FOV. Nominal CBCT scan protocols involve ~745 projections (1024×384 pixels, 0.388×0.776 mm² pixel size, dualgain readout mode) acquired over 360°. A 1D antiscatter grid (Pb absorber, Al interspacer, 12:1 grid ratio, 60 lines/cm, JPI HealthCare, Seoul, Korea) is oriented with the grid septa parallel to the $x$-axis and detachable at the detector surface.

FIGURE 1.

(a) A prototype O-arm^™ system featuring a slot collimator that reduces the extent of the $x$-ray beam in the longitudinal (z) direction and is actuated within a motorized collimator assembly to enable slot-beam 3D imaging. (b) Illustration of system geometry for measurements of dose, scatter, and image quality

The system was modified to include a slot collimator described previously for long-length 2D imaging³⁶–used in this work as a basis for slot-beam 3D imaging to investigate potential benefits to x-ray scatter and 3D image quality. The collimator gives a slot-beam spanning 1.2° measurements of dose, scatter, and image quality $\left({\varphi }_{slot}\right)$ as illustrated in Figure 1b. The slot collimator is automatically positioned via motorized actuators ~7 cm below the focal spot for slot-beam imaging and retracted for cone-beam imaging and fluoroscopy. Only the central slot (of three available slots) was used in the current work, and the two edge slots can be occluded by closing the existing cone-beam collimator blades located behind the slot collimator (closer to the x-ray source) using preset collimation settings. Slot-beam projection data were obtained using the same detector binning and readout mode as in the CBCT protocol described above, and the acquisition time is identical to a CBCT scan (~15 s) except for moving the collimator (negligible).

Both cone-beam and slot-beam projection data were air-normalized using the same full-field flood images. Because the current system did not permit the collection of flood-field corrections for the slot-beam readout mode, we anticipate increased noise and reduced uniformity close to the slot-beam collimator penumbra owing to residual flood-field nonuniformity. Images were reconstructed via 3D filtered-backprojection with 0.415 mm isotropic voxel spacing, yielding 512×512×385 voxels for CBCT and 512×512×32 voxels for slot-beam CT. A smooth 2D apodization filter was used with cutoff frequency = 0.2× Nyquist frequency to yield more isotropic resolution and noise characteristics.³⁸ Lateral truncation effects were mitigated in part by extrapolating detector signal along the $u$ direction. There were no x-ray scatter or beam-hardening corrections employed in the current work so as to focus specifically on the effects associated with slot-beam geometry.

2.2. Experimental methods

2.2.1. Dosimetry

Radiation dose (air kerma) was measured cone-beam for and slot-beam scans at three x-ray tube voltages (80, 100, and 120 kV) using 16 and 32 cm diameter “head” and “body” acrylic phantoms (Gammex RMI, Middleton, WI). Three such phantoms were stacked end-to-end to give a total length of 45 cm in the longitudinal direction, sufficient to capture dose from long scatter tails. The axis of the phantom was aligned with the rotational axis of the gantry, with the central plane at $z=0$ and the peripheral dosimeter channel in the 12 o’clock position. A 0.6 cm³ Farmer ionization chamber (RadCal AccuDose, Monrovia, CA) was inserted into the phantom assembly to measure air kerma at the center $\left(K_o\right)$ and periphery $\left(K_p\right)$, and the weighted air kerma $\left(K_w\right)$ was calculated as $K_w=K_o/3+2K_p/3$. For each kV setting, measurements were performed at the mAs level (Table 1) specified in the manufacturer’s technique chart.

TABLE 1.

Air kerma measurements for cone-beam and slot-beam protocols using head and body phantoms

	Scan protocol		Air kerma (mGy)						Ratio Kw,slot/Kw,cone
			Cone-beam			Slot-beam
	kV	mAs	K o	K p	K w,cone	K o	K p	K w,cone
Head	80	200	9.4	15.1	13.2	3.0	4.3	3.8	0.29
	100	160	14.6	21.6	19.3	4.5	6.1	5.6	0.29
	120	100	14.6	20.8	18.7	4.2	5.9	5.3	0.29
Body	80	400	6.3	25.3	18.9	1.9	5.1	4.0	0.21
	100	320	10.9	29.8	23.5	3.4	7.2	5.9	0.25
	120	200	12.7	28.6	23.3	3.6	6.8	5.8	0.25

For cone-beam scans, dose measurements were performed following the protocol described in AAPM Task Group 111,³⁹ where cumulative dose at the midpoint of the scanning range (16.7 cm) was determined from a single rotation of the cone-beam system with the ionization chamber at the central plane. For slot-beam scans, although the Farmer chamber is small, its length (~2 cm) is comparable to the width of the slot-beam at isocenter (~1.4 cm). Therefore, dose measurements were obtained by scanning with gantry position intervals of 1.4 cm along the $z$-axis (without moving the Farmer chamber) such that the total longitudinal coverage matched the cone-beam configuration (16.7 cm).³⁹ The slot-beam air kerma, $K_o$ and $K_p$, were obtained by adding measurements from the (16.7/1.4) = 12 gantry positions. As a result, the central cumulative dose for a slot-beam scan can be directly compared to the cumulative dose at $z=0$ for a cone-beam scan, since the longitudinal coverage has been factored in the dose metric.³⁹

2.2.2. Scatter measurements

The magnitude of x-ray scatter at the detector was evaluated for the four-system configurations (cone- and slot-beam, with and without an antiscatter grid). The SPR was measured using the arrangement in Figure 1b, with a (1×1×1 cm³) lead blocker placed at the entrance surface of the phantom. “Scatter only” $\left(d_S\right)$ and “scatter + primary” $\left(d_{S+P}\right)$ detector signal were read from projection data (5 projections, 10×10 mm² at the center of the projection) with and without the blocker, respectively, with SPR given by $d_S/\left(d_{S+P}-d_S\right)$. For each configuration, SPR was evaluated at 80, 100, and 120 kV for the head and body acrylic phantoms.

2.2.3. Contrast-to-noise ratio

Contrast-to-noise ratio (CNR) was evaluated in 16 cm “head” and 32 cm “body” configurations of the Corgi^™ phantom⁴⁰ (The Phantom Lab, Greenwich, NY) as shown in Figure 1b. The CNR measurements focused on a low-contrast insert (polystyrene, presenting −35 HU contrast to background). Regions of interest (ROls) of 1×1 cm² size were sampled in multiple slices to evaluate the CNR:

$$\text{CNR}=\frac{\left|{\bar{\mu }}_{\mathit{\text{insert }}}-{\bar{\mu }}_{\mathit{\text{background }}}\right|}{{\sigma }_{\mathit{\text{background }}}},$$ (3)

where ${\bar{\mu }}_{\mathit{\text{insert}}}$ is the mean voxel value within the polystyrene insert, ${\bar{\mu }}_{\mathit{\text{background}}}$ is that in the adjacent background material (at the same radius), and ${\sigma }_{\mathit{\text{background}}}$ is the standard deviation of the background region. For the low-contrast insert, this formulation of CNR is equivalent to definitions that consider the RMS variations in voxel value of the background and insert (i.e., for ${\sigma }_{\mathit{\text{background}}}\approx {\sigma }_{\mathit{\text{insert}}}$). Measurements were obtained at 80, 100, and 120 kV tube voltage, with tube current ranging 10–25 mA (75–188 mAs) for cone-beam and 32–100 mA (240–750 mAs) for slot-beam to yield similar dose (air kerma) ranges for cone-beam and slot-beam scans. All evaluated protocols (maximum 120 kV and 100 mA) are within the power limit of the system (32 kW). The dose efficiency in CNR among the foursystem configurations was characterized in terms of $\text{CNR}/\sqrt{K_w}$.

2.2.4. Anthropomorphic phantom studies

Image quality characteristics of the four-system configurations were further evaluated for various anatomical sites and imaging tasks in anthropomorphic head, abdomen, and pelvis phantoms scanned using the cone-beam and slot-beam protocols described above. The head and abdomen phantoms (Kyoto Kagaku, Kyoto, Japan) presented both high-contrast bone and lowcontrast soft-tissue features––the latter more pertinent to the results herein––for example, simulated cerebral ventricles (−30 HU contrast to brain parenchyma) and noncontrast-enhanced simulated tissues in the liver. The pelvis phantom (The Phantom Laboratory, Greenwich, NY) consisted of Rando^™ plastic containing a natural human skeleton, a simulated bladder, prostate and rectum, and a QA test object (The Phantom Laboratory) placed anterior to the lumbosacral junction. The QA test object provided a semiquantitative probe within the surrounding anatomical context, with inserts presenting −1050, −150, +40, and +850 HU contrast to background. CNR was evaluated in ROIs placed in various simulated soft-tissue structures along with overall image uniformity and artifacts (scatter and streaks).

3. RESULTS

3.1. Radiation dose

Dose measurements for cone-beam and slot-beam configurations are summarized in Table 1. For cone-beam head scans, the weighted air kerma $\left(K_w\right)$ ranged ~13–19 mGy (0.07–0.19 mGy/mAs) compared to ~4–5 mGy (0.02–0.06 mGy/mAs) for slot-beam scans. Similarly, for body scans, $K_w$ for cone-beam scans ranged ~19–23 mGy (0.05–0.012 mGy/mAs) compared to ~4–6 mGy (0.01–0.03 mGy/mAs) for slot-beam scans. For both the grid and gridless configurations, therefore, the dose for slot-beam scans was approximately 1/5 to 1/3 that of cone-beam scans due to reduction of scatter dose.

3.2. Scatter-to-primary ratio

Scatter measurements for the four-system configurations at 80–100 kV are summarized in Figure 2. The SPR for the grid and gridless configurations (denoted as SPR_Grid and SPR_NoGrid, respectively) are shown as scatter plots in Figures 2a and b for head and body setups, with measurements on a particular side of the identity line (dotted gray line) signifying higher SPR for that configuration. For instance, in Figure 2a, the gridless cone-beam configuration yielded a mean SPR_NoGrid = 1.4, which is ~3× higher than the SPR for cone-beam with the grid (mean SPR_Grid = 0.4). For both head and body imaging, the slot-beam SPR was lower than the cone-beam SPR with or without a grid. For cone-beam imaging, the grid reduced SPR by a factor of ~3, yielding a mean SPR_Grid of 0.4 and 1.7 for the head and body, respectively. A similar reduction was observed for slot-beam imaging, with the grid reducing the mean SPR from 0.2 to 0.07 for the head and from 0.8 to 0.3 for the body.

FIGURE 2.

SPR measurements for four system configurations: (a and b) Scatter plot of SPR_NoGrid versus SPR_Grid for the cone-beam and slot-beam configurations and (c and d) SPR_Cone versus SPR_Slot for the grid and gridless configurations––each in the 16 and 32 cm diameter head and body phantoms. The mean SPR values are marked by bold ● with error bars marking the standard deviations. The cluster of points in each case correspond to underlying measurements at 80 kV (+), 100 kV (○), and 120 kV (×). The dotted line marks the line of identity

The measurements are shown in a scatter plot in Figures 2c and d with the SPR for the slot-beam and cone-beam configurations denoted as SPR_Slot and SPR_Cone, respectively. For the gridless configuration, the slot collimator reduced SPR from 1.4 to 0.2 for the head and from 5.5 to 0.8 for the body, yielding a reduction by a factor of ~7. The SPR reduction decreased to ~6× with the antiscatter grid, where SPR was reduced from 0.4 to 0.07 for the head and from 1.7 to 0.3 for the body.

3.3. Contrast-to-noise ratio

The contrast resolution for the four-system configurations is summarized in Figure 3. The CNR for grid and gridless configurations normalized per unit square-root dose (denoted as ${\text{CNR}}_{\text{Grid}}/\sqrt{K_w}$ and ${\text{CNR}}_{\text{NoGrid}}/\sqrt{K_w}$, respectively) are summarized in Figures 3a and b. Measurements on a particular side of the identity line signify an improvement in CNR for that configuration. While the use of a grid increased CNR for the cone-beam configuration by 56.0% and 9.0% for the head and the body, respectively, a grid provided little or no improvement for the slot-beam configuration (affecting CNR by −11.8% for the head and +1.4% for the body).

FIGURE 3.

CNR (normalized by the square root of dose, $\sqrt{K_w}$) for four-system configurations: (a and b) CNR with and without a grid and (c and d) CNR in slot- and cone-beam geometry-each measured in 16 and 32 cm diameter head and body phantoms. The mean $\text{CNR}/\sqrt{K_w}$ values are marked by bold ● with error bars marking the standard deviations along with underlying measurements at 80 kV (+), 100 kV (○), and 120 kV (×). The dotted line marks the line of identity. (e and f) Example images of the low contrast polystyrene insert in (e) the 16 cm diameter head (80 kV, ~5 mGy) and (f) the 32 cm diameter body (120 kV, ~14 mGy). In each image, the grayscale level is set to the mean voxel value, and window width is equivalent (100 HU width) to facilitate intercomparison

Figures 3c and d summarize the dose-normalized CNR for slot-beam and cone-beam configurations (denoted as ${\text{CNR}}_{\text{Slot}}/\sqrt{K_w}$ and ${\text{CNR}}_{\text{Cone}}/\sqrt{K_w}$, respectively). For head imaging, the slot-beam configuration increased CNR by a factor of 1.2–2.2 compared to the cone-beam with or without a grid, with a stronger improvement for imaging the body, where the slot-beam configuration increased CNR by a factor of 2.3–2.5 with or without a grid.

Example axial images zoomed to a 25×25 mm² region containing the low-contrast insert are shown in Figures 3e and f, further illustrating the nature of improvements quantified in the CNR measurements among the four-system configurations. Images are shown for protocols with equivalent kV and dose—$K_w$ ~5 mGy for the head (Figure 3e) and ~14 mGy for the body (Figure 3f). The benefit of an antiscatter grid was more evident in the cone-beam configuration than for the slot-beam, and image noise was elevated with the grid.

3.4. Anthropomorphic phantom images

Figure 4 shows the example images of anthropomorphic head and body phantoms scanned using the four-system configurations. For display purposes the measured linear attenuation coefficient was converted to Hounsfield units (HU) using the relationship: $\text{HU}=\left[1000\times \left(\mu -{\mu }_{\text{water}}\right)/{\mu }_{\text{water}}\right]$. As seen in Figures 4a and b, the grid improved visibility of the simulated cerebral ventricles in the cone-beam image by improving image uniformity (reduced cupping artifact) and improving contrast. The dose-normalized CNR value, however, was slightly lower with a grid due to attenuation of the primary beam and a corresponding increase in image noise. This result roughly agrees with the CNR measurements in the Corgi head phantom at 80 kV (Figure 3a), where the improvement offered by the grid was less evident (and could actually diminish CNR) compared to higher voltage settings. The slot-beam images (Figures 4c and d) exhibited greater uniformity than cone-beam images, yielding an increase in CNR by a factor of 1.5–1.9 with or without a grid. Compared to slot-beam scanning with the grid, the image noise in the gridless case was reduced (thanks to a reduction in primary attenuation), yielding a higher CNR.

FIGURE 4.

Axial and sagittal views from cone-beam and slot-beam CT images of anthropomorphic (a–d) head, (e–h) abdomen, and (i–l) pelvis phantoms acquired with the four-system configurations. The ROIs for analysis of noise within relatively uniform regions are marked by dashed cyan boxes, and ROIs for a particular foreground stimulus (analysis of contrast and CNR) are marked by dashed yellow boxes. For each image, the grayscale level was set to the mean voxel value, and window width was fixed (180 HU width for the head; 300 HU for the abdomen and pelvis) to facilitate intercomparison

Cone-beam images in the abdomen (Figures 4e and f) exhibited moderate improvement in the conspicuity of simulated vessels and soft-tissue organ boundaries when a grid was used, yielding a slight improvement in CNR (7.3%). Visibility of such low-contrast structures was improved in the slot-beam images (Figures 4g and h), and the slot-beam image acquired without a grid again yielded a higher CNR value. The improvements in uniformity, noise, artifacts, and soft-tissue visibility were also evident in the pelvis images (Figures 4i–l). For example, the posterior aspect of the QA test object was barely visible in the cone-beam configuration, whereas it could be clearly discerned in the slot-beam images. Improvement in CNR given by the slot-beam configuration was more pronounced than the improvement in the head, yielding an increase by a factor of 1.7–2.2 with or without a grid. Similar to the previous two cases, the slot-beam image acquired without a grid again reduced noise and yielded the highest CNR value compared to the other three configurations.

4. DISCUSSION

The studies investigated the influence of a slot collimator on 3D image quality and dose for the O-arm^™ imaging system. Radiation dose, x-ray scatter, and contrast resolution were evaluated for slot- and cone-beam configurations with and without an antiscatter grid for various anatomical sites and imaging protocols. The findings guide important practical considerations on how slot-beam imaging capability could be implemented for particular clinical scenarios that could benefit from higher image quality and reduced dose.

The slot collimator was found to provide stronger scatter rejection compared to the antiscatter grid. For both head and body imaging, the grid reduced SPR by ~3× compared to a gridless configuration, whereas the slotbeam configuration reduced SPR by 6–7× compared to a cone-beam. Such reduction in scatter improved the visibility of soft-tissue structures, with the slot-beam geometry yielding up to 2.5× higher soft-tissue CNR than cone-beam images at equivalent dose for both grid and gridless configurations.

The benefits and tradeoffs of the antiscatter grid exhibited some dependence on the anatomical site, whereas the slot collimator was less sensitive to such considerations. For cone-beam configurations, the grid improved CNR to a greater degree for the head than for the body (Figures 3a and b). On the other hand, improvements for the slot-beam gridless configuration were similar for the head and body (Figures 3c and d). This effect is due to the degree of scatter rejection provided by the slot collimator, which reduces scatter via prepatient collimation and does not attenuate the primary signal. In other words, there is no loss of primary photons caused by septa or interspacer material as with a grid.

The absorption of primary caused by the grid can undermine the imaging performance benefits gained with a slot-beam geometry. For instance, the grid did not improve soft-tissue CNR for either head or body imaging with the slot-beam geometry, suggesting that the increase in noise outweighed the gain in contrast offered by the grid. Increased noise was also evident in images of anthropomorphic phantoms acquired with the grid (Figure 4). Ideally, as common with some C-arm interventional/angiography systems, one might envision removing the grid for imaging with the slot-beam geometry, recognizing important flood-field calibration and clinical workflow considerations for working with a removable grid.

In view of the enhanced image quality with a slotbeam configuration compared to a cone-beam, one may envision certain clinical scenarios that could benefit from slot-beam 3D imaging, for which a high-quality image within a localized region (< 1.4 cm length in longitudinal FOV) is sufficient to address a particular task. Potential clinical applications include spinal or cranial-maxillofacial tumor resection, where improved imaging could aid in visualization of tumor margins and facilitate more complete resection without compromising surrounding tissues. The ability to assess the adequacy of tumor removal in the OR could improve outcomes and avoid the high costs associated with revision surgeries. High-quality slot-beam imaging could also be useful in intracranial neurosurgery as a means to visualize the placement of an interventional device (e.g., a needle or neuroelectrode) in the context of surrounding soft-tissues in tumor biopsy, cystectomy, or deep-brain stimulation (DBS). For instance, slot-beam imaging could be used to better verify the rotational orientation of a directional DBS lead within the low-contrast target structure, which would help to inform subsequent programming of the device. Such electrodes and associated directional markers are typically 1–2 mm long⁴¹ and are within the FOV of slot-beam 3D imaging. Thanks to the reduction in dose achieved by the slot collimator, consecutive slot-beam scans could also be acquired if the device extends beyond the FOV_z. The enhanced contrast resolution of slot-beam imaging could provide a valuable check against complications, such as hemorrhage, thrombosis, or retained foreign bodies near the surgical site, for which immediate intervention is needed to reduce morbidity or infection.

The motorized actuation of the slot collimator facilitates multi-mode fluoroscopic, cone-beam, and slotbeam imaging capability on the system, extending the potential capability of the system for 2D and 3D imaging. The collimator could potentially be implemented in combination with the motorized motion of the gantry along the z-axis as a basis for “long-length” intraoperative imaging. Such is the basis for the 2D “Long Film” capability reported by Zhang et al.³⁶ and Ladd et al.⁴² Extending such capability to helical 3D scanning could provide the improvements in image quality demonstrated above for a longer FOV_z (e.g., up to 35 cm of gantry motion) with a single longitudinal stroke, exceeding the longitudinal coverage of a cone-beam scan. Implementation and evaluation of helical slot-beam scanning is the subject of future work.

Recognizing the narrow width of the slot-beam and the importance of aligning the beam position with anatomy of interest, several strategies have been developed or can be envisioned. The current system allows a scout view using 2D full-field radiographs prior to a 3D slot scan, and the gantry position can be set based on the scout views. Such features greatly facilitate the alignment process for both cone-beam and slot-beam imaging.

The current study presents a few limitations. First, the flood-field corrections for the slot-beam scans were performed using full-field images, giving reasonable corrections about the central slice of the fairly narrow longitudinal span of the slot-beam collimators, but leading to degradation in uniformity and noise near the slotbeam penumbra. Specifically, the uniformity of signal and noise in slot-beam reconstructions was seen to degrade in regions outside ±4 mm of the central slice. The effects of the collimator penumbra could be partly mitigated by air-scan normalization with flood-fields acquired with the slot collimators in place; unfortunately, this was not possible in the current system configuration due to limitations in detector readout mode and control software for slot-beam imaging. The current study, therefore, does not investigate image quality in regions of the penumbra. Moreover, the results presented are likely a conservative view (lower bound) of the imaging performance that could be achieved in slot-beam imaging given improved air-scan flood-field normalization. Future development of slot-beam helical imaging will require not only improved flood-field corrections but also a high degree of mechanical reproducibility in slot collimator and gantry positioning to ensure nonuniformity.

Furthermore, the current work investigated a single grid configuration and orientation––specifically, a linear grid with Pb septa and Al interspacer with grid ratio 12:1. The grid orientation was fixed such that septa were parallel to the slot, and an orthogonal orientation was not possible in the current work due to mechanical constraints within the gantry. While the current orientation has been shown to be superior for cone-beam geometry (e.g., giving slightly better scatter rejection and uniformity in the scatter distribution²¹), it is likely not the superior orientation for slot-beam geometry, since it has a wider acceptance angle to scatter along the fan angle direction. This partly accounts for the modest gains observed for the grid in slot-scan acquisitions. Future work will consider variation in the grid configuration (e.g., various grid ratios) and orientation (with septa parallel to the longitudinal axis or use of a 2D grid⁴³).

5. CONCLUSIONS

The imaging performance of a prototype slot-beam CT configuration on the O-arm^™ system was investigated. The use of an antiscatter grid was shown to reduce scatter for both configurations, and while the resulting soft-tissue visibility for all anatomical sites was improved by the grid in the cone-beam geometry, its benefit was minimal for the slot-beam geometry, owing to an increase in image noise caused by primary attenuation in the grid. By comparison, the slot collimator yielded a measurably higher reduction in scatter compared to the grid. For both head and body imaging at equivalent dose, slot-beam images acquired without the grid achieved the best overall soft-tissue image quality (highest CNR for low-contrast structures) compared to the other three configurations. These warrant further investigation to validate the performance of slot-beam 3D imaging in realistic clinical scenarios, with a focus on the clinical utility of high-quality imaging with such limited longitudinal coverage, how slot-beam imaging could factor in clinical workflow, and future work that could extend slot-beam capability to helical acquisition.

DATA AVAILABILITY STATEMENT

Research data are not shared.