A Novel Bioluminescence Tomography System Compatible with CBCT-guided Small Animal Irradiators for High-Precision Preclinical Radiation Research

Abstract

Widely used cone-beam computed tomography (CBCT)-guided irradiators struggle to localize soft-tissue targets due to low imaging contrast. While bioluminescence tomography (BLT) offers a promising functional imaging solution, its adoption in pre-clinical radiotherapy research has been limited. To address this, we developed MuriGlo, a novel BLT system compatible with CBCT-guided small animal irradiators to support high-precision radiation studies. We demonstrate MuriGlo’s capabilities in supporting both in vitro and in vivo experiments. MuriGlo consists of a detachable mouse bed, thermostatic control, mirrors, filters, and a charge-coupled device (CCD) camera, enabling multi-projection and multi-spectral bioluminescence imaging (BLI). The detachable bed facilitates animal transfer between MuriGlo and an irradiator for BLT-guided radiation study. We evaluated the thermostatic control’s ability and demonstrated that it can maintain a consistent animal body temperature at 37°C throughout imaging. We also quantified detection sensitivity via signal-to-noise ratio (SNR) in detecting minimal cell quantities using glioblastoma (GL261) cells with Luc2 and AkaLuc reporters. The optical system can detect as few as 1173 GL261-Luc2 and 61 GL261-AkaLuc cells in vitro at SNR = 5. For image-guided capabilities, we present BLT-guided 5-arc, BLT-guided 2-field box, and BLI-guided single-field plans. The high conformal 5-arc plan fully covers gross tumor volume (GTV) at prescribed dose with minimal normal tissue exposure from moderate to high dose range, while the simplified, high-throughput BLT-guided 2-field box achieves 100% GTV coverage but results in larger normal tissue exposure. The choice of the planning strategy should therefore depend on the specific requirements for radiation accuracy and experimental throughput. Moreover, we compared MuriGlo’s tumor localization accuracy for widely used irradiators, SARRP and SmART+. The localization accuracy of MuriGlo for both SARRP and SmART+ irradiators is < 1 mm with GTV coverage > 97%. This universal, BLT-guided platform enables plug-and-play integration with commercial irradiators, supporting functional image guidance and enhancing high-precision preclinical radiation research.

INTRODUCTION

Pre-clinical radiotherapy research using animal models is an essential step in translating concepts from in vitro studies to clinical applications (1). X-ray cone-beam CT (CBCT)-guided small-animal irradiators are the most commonly used image-guided systems for enabling high-precision in vivo irradiation (2–6). While CBCT provides excellent guidance capability, it is less effective at localizing soft tissue targets in low-image contrast environment. Bioluminescence imaging (BLI), with its high image contrast, offers an attractive solution for soft tissue targeting and can serve as a surrogate for monitoring tumor growth and treatment response (7, 8). However, conventional two-dimensional (2D) BLI is limited in precisely localizing internal BL sources/tumors because the light propagation within tissues is susceptible to tissue optical properties and irregular animal torsos (9). To address this challenge, we and others have developed bioluminescence tomography (BLT) to reconstruct the 3D distribution of BL tumors, resulting from surface BLI (9–15). BLT has been used to guide radiation in various orthotopic tumor models, including brain (11–13), breast (14), prostate (16), and pancreas (17). However, the adoption of BLT-guided system in the broader radiation research community has been limited (18).

To advance research capabilities in the scientific community and complement the CBCT-guided irradiators (4, 5), we developed a BLI/BLT system, MuriGlo. This optical platform features an innovative mouse bed system that enables seamless transfer of animals from the BLT imaging system to commercial irradiators, allowing for image-guided irradiation without modifying existing irradiators. A major challenge in mathematical model-based BLT is that it represents an ill-posed inverse problem, meaning that limited surface BLI measurements cannot uniquely or stably determine the internal light source distribution (19). To mitigate this limitation in BLT reconstruction, we utilize multi-projection and multi-spectral imaging to improve reconstruction stability and accuracy.

In this work, we show the capabilities of MuriGlo in supporting both in vitro and in vivo studies and present planning strategies that illustrate how BLT can be utilized for image-guided radiation research. Specifically, we assess a thermostatic system implemented in the imaging chamber to maintain consistent body temperature in studied animals, crucial for preserving physiological function and ensuring stable BL spectrum and emission efficiency, as demonstrated with the Luc2 reporter (12). We also quantify the system sensitivity in detecting minimal number of cells in vitro using luciferase-labeled glioblastoma (GBM) cell lines, GL261-Luc2 and -AkaLuc (20). The Luc2 is a commonly used firefly reporter, while the AkaLuc is a recently developed, high-brightness reporter (20). Despite potential biological variability across laboratories, the in vitro results provide a useful reference for MuriGlo’s sensitivity, allowing users to assess their system’s performance in a manner that is reasonably consistent with our findings. To further demonstrate the potential of using MuriGlo as a BLI/BLT system for guiding radiation and supporting in vivo radiation research, we present planning strategies that span from complex, highly conformal plans to streamlined, high-throughput approaches. Finally, we show that MuriGlo, equipped with the innovative mouse bed system, is compatible with widely-used commercial CBCT-guided irradiators (small animal radiation research platform, SARRP and small animal radiation therapy system, SmART+) (4, 5) to localize orthotopic tumors for image-guided studies. Compared to previous work focused on characterizing MuriGlo’s imaging performance (21), the current work introduces the above new technological and translational features. These contributions move beyond previous technical characterization to establish MuriGlo as a practical and universal platform for in vitro imaging and in vivo image-guided irradiation studies for radiation society. The high precision and broad compatibility of this optical-guidance system are expected to advance radiation research by providing functional targeting beyond traditional anatomical imaging.

MATERIALS AND METHODS

System Configuration

MuriGlo consists of a rotational 3-mirror system, a fixed 45° mirror, filter wheel (Edmund Optics Inc., NJ), lens (50 mm, f/1.2, Nikkor, Nikon Inc., NY), and charge-coupled device (CCD) (iKon-M934, Andor Technology, Belfast, UK), enabling a light-tight multi-spectral imaging platform (Fig. 1). All mirrors were silver-coated with 98% reflectivity (H. L. Clausing Inc, IL). The BL signal emitted from the imaged animal passed through the 3-mirror, fixed mirror, and filter, and was captured by the CCD (Fig. 1, white line). The 3-mirror can rotate 360° around the imaged animal for multi-projection imaging. Four 20-nm band-pass filters (Chroma Technology Corp., VT) at 590, 610, 630, and 650 nm were inserted into the filter wheel for multi-spectral imaging. Four LEDs were mounted to the 3-mirror system to provide white-light illumination for animal photo image (Fig. 2) used to identify fiducial markers on the bed for BLI and CBCT registration. More details can be found in Xu et al. (21).

FIG. 1.

Workflow of applying MuriGlo/BLT for image-guided irradiation. MuriGlo captures multi-projection and multi-spectral BLIs from a mouse placed on a detachable bed with anesthesia and rubber bands for immobilization. The white line shows the optical path from the imaged object toward the CCD. The fiducial markers on the bed were employed for the CBCT and optical coordinate registration. The bed, with the anesthetized mouse, is transferred from the bed adapter to a 2nd adapter affixed to the irradiator for CBCT imaging and BLT-guided irradiation. Two types of secondary adapters were specifically designed for SARRP and SmART+.

FIG. 2.

Workflow of in vivo BLT-guided irradiation consisting of 1. data acquisition, 2. processing, and 3. irradiation. Bioluminescence GBM model is used as an example shown.

A thermostatic control was implemented in MuriGlo. The thermostatic control system consists of a type-T thermocouple (5TC-GG-T-30-36, Omega, CT), a proportion integration differentiation controller (PID, CN32Pt, Platinum Series, Omega, CT), a solid-state relay (SSRL240AC10, Omega, CT), a 24V transformer, a heater system including a heater (FCH-FGC15132R, Omega, Connecticut) and two circulation fans (CFM-6015V-254-362-20, Digi-Key Electronics, MN), a switch to control the fans, and two power distribution boards (PCB007; Evemodel, Jiangsu, China). Furthermore, two additional fans were installed in the instrument chamber to circulate hot air out and keep the CCD operating below −80°C. To measure the ambient temperature surrounding the imaged animal, a thermocouple was positioned on the rear surface of the 3rd mirror of the 3-mirror system [Supplementary Fig. S1;³ (https://doi.org/10.1667/RADE-25-00114.1.S1)]. Such a design allows the thermocouple near the imaged mouse and prevents collision between the thermocouple and the bed during the 3-mirror rotation. More details about the design of the thermostatic control system can be found in Xu et al. (21).

For the BLT-guided irradiation, the tumor-bearing mouse was placed on the bed integrated with a bed adapter and inserted into MuriGlo for BL imaging. After imaging, the bed carrying anesthetized mice can be detached from the MuriGlo adapter and connected to a secondary adapter affixed to SARRP (Xstrahl Inc.) or SmART+ (Precision X-ray Inc.) for CBCT imaging, followed by BLT reconstruction for radiation guidance. Two types of secondary adapters are available; the 1^st type can be directly attached to SARRP bed support, and the 2nd, named as universal adapter, was designed to be placed on flat beds for both irradiators (Fig. 1). Throughout the image-guided session, the anesthetized animal was immobilized with rubber bands, and the transport between MuriGlo and the irradiator was kept within 5 meter (21). The rubber bands were used to secure the four limbs and tail of the mouse through simple restraining Velcro on the imaging bed (Fig. 2). This is a routine procedure used to immobilize mice during BLI and CBCT imaging to minimize motion and ensure consistent positioning.

Evaluate Thermostatic Control System

While we have described the thermostatic control system in Xu et al. (21), this work presents a systematic evaluation of its performance. Herein, we conducted 3 tests to assess the performance of the thermostatic control in maintaining the ambient temperature of the imaging chamber and animal body temperature. In the 1^st test, we checked whether the thermocouple [Sys-TC, Supplementary Fig. S1 (https://doi.org/10.1667/RADE-25-00114.1.S1)] on the 3-mirror system accurately reflects the ambient temperature surrounding the animal. We placed a reference thermocouple (Ref-TC) at the bed’s center to record the bed’s temperature and compared the readings from both TCs.

Since our BLT system was designed for single-mouse imaging, after imaging one animal, one would remove the bed from MuriGlo and replace it with the next animal for imaging. For the 2nd test, we aimed to investigate whether the chamber temperature would drop when removing the bed and how quickly the bed could be warmed to the desired temperature after it was reinserted into the system. To mimic BLT-guided experiment workflow, we took out the bed attached with Ref-TC from MuriGlo, kept it outside for 10 min, and then reinserted it into MuriGlo. Throughout this process, we monitored the time-resolved temperature changes for both Sys- and Ref-TC.

The 3rd test evaluated whether the thermostatic control could maintain the animal body temperature at ~37°C during imaging. We used rectal temperature as a measure of mouse body temperature. C57BL/6J albino mice (n = 5, 6–8 weeks old, female; Jackson Laboratory) were used. We placed a plastic-covered thermocouple into the mouse rectum with ~2 cm depth. When the temperature reached ~36°C, the mouse was placed into MuriGlo, and its body temperature was monitored to evaluate if the thermostatic control could maintain the temperature throughout imaging.

System-specific Cell Spectrum

To achieve accurate multi-spectral BLT reconstruction, it is important to quantify the system spectral response and the emission spectrum of BL cells (12). For simplicity, we used the MuriGlo to measure the so-called system-specific spectrum, which combines the system and cell spectral response.

For the system-specific spectrum measurements, both GL261-Luc2 and -AkaLuc cells were cultured in petri dishes at 37°C with 5% CO₂ in Dulbecco’s modified eagle medium (DMEM; Gibco^™; Life Technologies Corp.) supplemented with 10% fetal bovine serum (Gemini Bio, West Sacramento, CA), reaching a confluency level >80%. GL261-Luc2 cells were mixed with D-luciferin (Perkin-Elmer, MA) at a concentration of 0.75 mg/mL in phosphate buffer solution (PBS) with pH = 7.4, and GL261-AkaLuc cells were mixed with Akalumine-HCL (Sigma-Aldrich, Darmstadt, Germany) at a concentration of 250 μM/ml in PBS with same pH level for BL imaging.

To prevent vaporization and heat loss from PBS, we used a 10-μm-thick polyvinyl chloride membrane, transparent to the spectrum of interest without attenuation, to cover the petri dish. Given the distinct spectral ranges of GL261-Luc2 and -AkaLuc cells, multispectral BLIs for GL261-Luc2 at 590, 610, 630, and 650 nm, and for GL261-AkaLuc at 610, 630, 650, and 680 nm were captured. To account for the dynamic in vitro BL signal variation throughout imaging, open-field (without optical filter) images were taken before and after each spectral BLI, and an in vitro time-resolved signal curve was generated. The intensity of multispectral BLIs taken at different time points was normalized to the time-resolved curve, and the normalized values as a function of wavelength resulted in the system-specific spectrum for the cell of interest.

Additionally, to evaluate the impact of ambient temperature possibly on varying the spectrum, we compared two conditions: 23°C and 37°C, representing the MuriGlo with the thermostatic control deactivated and activated, respectively.

Imaging Sensitivity for In Vitro Study

We investigated the capability of MuriGlo in detecting the minimal number of cells using GL261-Luc2 and -AkaLuc cells. Cells were seeded in triplicate into black 96-well plates (165305, ThermoFisher Scientific). GL261-Luc2 cells were seeded at a density of 6400 cells/well, followed by serial dilution in DMEM to achieve a range of 25 to 6400 cells/well. GL261-AkaLuc cells were seeded at a density of 400 cells/well and serially diluted to a range of 25 to 400 cells/well. The initial cell densities were determined using an automated cell counter. For BL measurements, D-luciferin was added to GL261-Luc2 wells at a final concentration of 0.75 mg/ml, and Akalumine-HCL was added to GL261-AkaLuc wells at a final concentration of 250 μM/ml. Imaging was performed after substrate addition, once the BL-cells reached their respective peak emission (approximately 10 min for GL261-Luc2 and 15 min for GL261-AkaLuc). Open-field BLIs without spectral filter at 10 × 10 binning and 60-second exposure time were taken. We used the signal-to-noise ratio (SNR) = 5 as a threshold (22) to quantify the sensitivity of MuriGlo, defined as SNR=<S>/σ, where, <S> is the average CCD counts of BLI within the region of interest (ROI) and σ is the standard variation of background.

Multispectral BLT Reconstruction

The forward model describing light propagation within tissue is given by diffusion approximation (DA) with the Robin-type boundary condition (23). The details of the reconstruction were published elsewhere (24). Briefly, the relationship between surface fluence rate ${\mathit{\varphi }}_{\mathrm{\lambda }}$ at wavelength $\mathrm{\lambda }$ and internal source $\mathit{S}$ can be expressed as:

$${\mathit{G}}_{\mathrm{\lambda }}{w}_{\mathrm{\lambda }}\mathit{S}={\mathit{\varphi }}_{\mathrm{\lambda }},$$ (1)

where ${\mathit{G}}_{\mathrm{\lambda }}\in {ℛ}^{m\times n}$ is the sensitivity matrix describing the changes of ${\mathit{\varphi }}_{\mathrm{\lambda }}\in {ℛ}^m$ related to the source $\mathit{S},w_{\mathrm{\lambda }}$ is the system-specific spectrum of the source $\mathit{S}\in {ℛ}^n$, m is the number of BLI measurements, and $n$ is the number of mesh nodes. ${\mathit{G}}_{\mathrm{\lambda }}$ is calculated from the diffusion approximation using the NIRFAST software based on finite element method (FEM) with the input of tissue optical properties (25). Due to the non-contact imaging geometry, there is inevitable deviation between the actual surface fluence rate ${\mathit{\varphi }}_{\mathrm{\lambda }}$ and BLI measurement ${\mathit{b}}_{\mathrm{\lambda }}$. The spectral derivative method utilizing the ratio of neighboring wavelengths as the input for BLT reconstruction, minimizing the deviation and eliminating complicated system modeling for non-contact imaging geometry was applied (24). The BL source distribution $\mathit{S}$ can be iteratively solved by using a compressive sensing conjugated gradient algorithm (26).

In Vivo Bioluminescence GBM Model and Imaging-guided Process

All animal procedures were conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) at the University of Texas Southwestern Medical Center. A glioblastoma (GBM) model was used as a representative case to assess the BLT system’s capability in localizing orthotopic tumors and to illustrate imaging-guided planning strategies. To establish the GBM model, $1.2\times {10}^5$ GL261-Luc2 cells were implanted into the left striatum of C57BL/6J albino female mice at a depth of 3 mm from the surgical opening. The detail of tumor implant technique can be found in the literature (27). We selected the albino strain solely to minimize light attenuation related to skin pigmentation; however, this choice is not a requirement, as BLT is well established for pigmented animals as well (12).

A summary workflow for in vivo BLT-guided irradiation process is provided in Fig. 2. Two weeks after implantation, GBM-bearing mice with fur removal underwent multi-spectral and multi-projection BLIs, followed by multi-projection photographic images. The image taken from the top of the bed was labeled as the 0° projection imaging. Photo images at $-{60}^{\circ },-{30}^{\circ },0^{\circ },{30}^{\circ }$ and ${60}^{\circ }$ projections with LEDs turned on were captured to retrieve fiducial marker positions (Fig. 1) for image registration between BLIs and CBCT images. Subsequently, the anesthetized mouse, positioned on the detachable bed, was docked into the 2nd adapter pre-installed in the irradiator for CBCT imaging. The CBCT scan serves 2 purposes: 1. providing anatomical structure to generate a tetrahedral mesh for the FEM-based BLT reconstruction, and 2. defining 3D coordinates for treatment planning and irradiation. An in-house auto-contouring algorithm was used to segment the mouse body from CBCT image. A previously published geometrical calibration method was applied to register 3D CBCT and 2D optical coordinates, using the marker positions shown in both the photo and the CBCT image. This allowed for the mapping of multi-projection BLIs onto the 3D surface mesh derived from the CBCT scan, serving as input for BLT reconstruction (21). During the entire imaging process, the mice were anesthetized with isoflurane delivered via an isoflurane vaporizer (E-Z Anesthesia; Euthanex Corp., Palmer, PA), which mixed vaporized isoflurane (Fluriso^™, VETOne^®; MWI Animal Health, Boise, ID) with 2% oxygen to provide a continuous inhalation anesthesia.

The detected in vivo BL signal at 590 nm was much lower than that at the other 3 wavelengths, and the large ratio of BL signals between 610 and 590 nm could introduce numerical errors for our spectral derivative-based reconstruction. Therefore, for the study of in vivo GBM-bearing mouse, we chose the images at 610, 630, and 650 nm for BLT reconstruction. The absorption coefficient ${\mu }_a$ 0.16, 0.08, and 0.06 mm⁻¹ and reduced scattering coefficient ${{\mu }_s}^{’}$ 1.56, 1.51, and 1.46 mm⁻¹ at 610, 630, and 650 nm, respectively, and refractive index 1.4 of tissue were used for BLT reconstruction (12, 21). Threshold of 0.5 of the maximum power density of BLT-reconstructed source was used to delineate the BLT-reconstructed gross target volume $\left({\mathrm{G}\mathrm{T}\mathrm{V}}_{\text{BLT}}\right)$ (12). ${\mathrm{T}}_2$-weighted MRI (MRS 3017, 3T, MR Solutions Ltd. GU3 1LR, UK) was utilized to define the GTV of the GBM-bearing mice, serving as the ground truth to validate the accuracy of BLT localization. The MRI and CBCT image of mouse head were registered using 3D slicer (https://www.slicer.org/) (28).

Treatment Plans for In vivo GBM

We developed three X-ray radiation plans, a BLT-guided noncoplanar 5-arc, a BLT-guided 2-field box, and a BLI-guided single-field, using SARRP MuriPlan (version 3.0.0; Xstrahl Inc.).

For the 5-arc plan, we applied a uniform 0.75 mm margin to account for the uncertainty related to BLT target localization. Following our published BLT study (12), the margin size was optimized to balance tumor coverage and normal tissue sparing, which are defined respectively as $\left({\mathrm{P}\mathrm{T}\mathrm{V}}_{\text{BLT}}\cap \mathrm{G}\mathrm{T}\mathrm{V}\right)/\mathrm{G}\mathrm{T}\mathrm{V}$ and $\left({\mathrm{P}\mathrm{T}\mathrm{V}}_{\text{BLT}}-{\mathrm{P}\mathrm{T}\mathrm{V}}_{\text{BLT}}\cap \mathrm{G}\mathrm{T}\mathrm{V}\right)/\left({\mathrm{V}}_{\text{head}}-\mathrm{G}\mathrm{T}\mathrm{V}\right)$, where ${\mathrm{V}}_{\text{head}}$ is the mouse head volume and ${\mathrm{P}\mathrm{T}\mathrm{V}}_{\text{BLT}}$ is BLT planning target volume derived by adding the margin to ${\mathrm{GTV}}_{\text{BLT}}$. The 5-arc plan was designed using a motorized variable collimator (MVC). The isocenter of each arc was positioned at the center of mass (CoM) of the ${\mathrm{G}\mathrm{T}\mathrm{V}}_{\text{BLT}}$, and the plan was optimized to deliver a prescribed dose of 5 Gy covering 95% of the ${\mathrm{PTV}}_{\text{BLT}}$. Compared to previously reported noncoplanar 7 static field plan (12), the 5-arc plan, with greater gantry-angle freedom, streamlines beam direction design and improves planning efficiency for image-guided radiation studies.

The 2-field plan was designed with 2 lateral beams, and the isocenter was placed at the CoM of GTV_BLT. Considering the overall uncertainty of BLT localization resulted from various tumor model (12, 17, 21), the collimation of each beam was opened 1 mm beyond the edge of GTV_BLT, resulting in 7 × 7 mm² as the collimator size for the in vivo case we present.

Since tumor volume would not be available when one uses surface BLI to guide irradiation, the 5 × 5 mm² collimator, used in previous studies (11, 12), was applied for the BLI-guided single field irradiation to conservatively cover the 2-week-old GBM, where tumor size is < 4 mm in diameter based on the published tumor growth curve (11). The central axis of the single field was aligned at the location where the maximum intensity of surface BLI is. The beam isocenter was placed at the inferior portion of the skull to ensure sufficient coverage of GBM within the radiation field, with the isocenter dose set to 5 Gy.

RESULTS

Effectiveness of Thermostatic Control

Considering 37°C as the desired body temperature and accounting for the distance between the Sys- and Ref-TC, we empirically set the thermostatic control to 36°C. As shown in Fig. 3a, the Sys-TC reached 36°C within 10.1 min after heater activation from room temperature, while the Ref-TC reached 37°C in 8.5 min. After 30 min, both thermocouples stabilized with only a 1°C difference, while the Ref-TC consistently showed 37°C around the bed. Thus, a 30-min warm-up period is recommended to achieve the target temperature around the animal and ensure thermal equilibrium.

FIG. 3.

Assessment of thermostatic control and system-specific cell spectrum in vitro vs. ambient temperatures. Panel a shows the temperature reading from the Sys-TC and Ref-TC. Panel b shows measured rectal temperature over time (n = 5); the red curve indicates the average, and the pink shaded area represents the standard deviation (SD). Panels c and d: present the emission spectrum of GL261-Luc2 and -AkaLuc cells, respectively, measured at 37 and 23°C (n = 6).

To simulate an imaging workflow involving multiple animals, we removed the bed with the Ref-TC from MuriGlo at the 60-min mark, kept it outside for 10 min (to simulate animal preparation), and then reinserted it into the chamber. Despite the brief chamber opening during bed removal, Sys-TC readings remained stable throughout. Upon reinsertion, the Ref-TC indicated that the bed was reheated to 37°C within 5.9 min. Figure 3b showed that the average rectal temperature increased slightly from 36 to 37°C within 60 min of system warm-up and animal placement, with fluctuations remaining within ±0.35°C [1 standard deviation (SD)]. These results confirmed the effectiveness of the thermostatic control system in maintaining stable chamber and animal body temperatures during the imaging workflow.

System-specific Cell Spectrum

Figure 3c and d show that the GL261-Luc2 and -AkaLuc cell lines have different spectra, with data collected at 37°C used as input parameters for multi-spectral BLT reconstruction. Unlike GL261-AkaLuc, the GL261-Luc2 spectrum shows a red shift as the temperature increases, which was also observed elsewhere (12).

Imaging Sensitivity for In Vitro Bioluminescence Study

To evaluate imaging sensitivity, we captured open-field BLIs of GL261-Luc2 and GL261-AkaLuc cells at various cell counts (Fig. 4a). We observe that the BL emission efficiency of GL261-AkaLuc with Akalumine-HCL was about 21-fold higher than that of GL261-Luc2 with D-Luciferin (Fig. 4a), based on the CCD-detected optical signal per cell, aligned with previous findings (20). For both cell lines, the SNR increased linearly with the increase of cell number. With a threshold SNR of 5, the MuriGlo can detect as few as 1,173 GL261-Luc2 cells and 61 GL261-AkaLuc cells, respectively (Fig. 4b).

FIG. 4.

MuriGlo sensitivity in detecting minimal numbers of GL261-Luc2 and -AkaLuc cells in vitro. Panel a shows open-field (without optical filters) BLIs vs. number of cells (n = 3). Panel b shows the corresponding SNR vs. number of cells. The green circles compare the light emission efficiency of GL261-Luc2 and -AkaLuc cells as optical signal counts per cell.

In Vivo BLT and BLI-guided Irradiation

Figure 5a shows the BLI, originating from an in vivo GBM, while Fig. 5b presents the corresponding BLI mapped onto the surface mesh of the mouse head, serving as input for BLT reconstruction. The average deviation between the CoM of GTV_BLT (black dots) and the geometric center of MRI-delineated GTV (blue contour) was 0.61 mm with a SD of 0.37 mm [n = 4, Supplementary Fig. S2 (https://doi.org/10.1667/RADE-25-00114.1.S1)]. Figure 5(c1–c3) illustrate the BLT-guided 5-arc non-coplanar plan, designed to achieve 95% PTV_BLT coverage at a prescribed dose of 5 Gy. Although the 2-field plan also achieved full GTV coverage at 5 Gy [Fig. 5(d1–d3)], it resulted in greater normal tissue involvement compared to the 5-arc plan (Fig. 5f). In contrast, the single-field plan, guided by the location of the maximum BLI signal, lacked targeting precision due to optical scattering from internal sources and variations in the animal’s torso [Fig. 5(e1–e3)], resulting in suboptimal GTV coverage 94% as shown in Fig. 5(e3 and f). In terms of normal tissue exposure, the DVH further indicates that across the 0.2–5.2 Gy range, both the single-field and 2-field box plans result in a relatively consistent normal tissue volume compared with the 5-arc plan, with the single-field plan achieving greatest overall tissue volume sparing among 3 plans. The 5-arc plan (green solid line) provides the most effective sparing at intermediate to high doses (>3 Gy), though it produces a broader low-dose bath <1 Gy. Overall, the single-field plan minimizes normal tissue exposure but provides suboptimal GTV coverage. The 5-arc plan achieves superior sparing at moderate to high doses while maintaining good tumor coverage, whereas the two-field box plan preserves tumor coverage but results in the largest normal tissue exposure.

FIG. 5.

In vivo BLT reconstruction and BLT/BLI-guided plans. Panel a shows BLI (610 nm; heat map) of a GBM-bearing mouse. Panel b shows the BLI mapped onto mouse head mesh. Panel (c1–c3) is BLT-guided 5-arc plan with 5 Gy covering 95% of PTV_BLT (pink contour) generated from a BLT-reconstructed tumor, GTV_BLT (heat map), with MRI-delineated GTV (blue contour) and GTV_BLT CoM (black point). Panel (d1–d3) is the 2-field box plan. Panel (e1–e3) is the single beam plan guided by the position of maximum BLI value (white dot, panel e3), and the isocenter was located at the skull (white cross, e1). Panel f is the dose volume histogram (DVH) of proposed plans for PTV_BLT, GTV, and normal tissue (NT).

Integrating MuriGlo with SmART+

To evaluate the feasibility of applying MuriGlo to guide SmART+ irradiation, we followed the procedures outlined in Figs. 1 and 2. Figure 6a shows a representative in vivo GBM BLT reconstruction based on a SmART+ CBCT image. The CoM deviation of the GTV_BLT and the geometric center of the GTV is 0.86 ± 0.13 mm (Fig. 6b), confirming the precision of the MuriGlo-SmART+ combination.

FIG. 6.

Comparison of in vivo BLT localization accuracy between MuriGlo-SmART+ and MuriGlo-SARRP. Panel a is a representative case of applying MuriGlo to localize in vivo GBM based on SmART+ CBCT imaging, showing overlap of MRI-labelled GTV (blue contour), ${\mathrm{G}\mathrm{T}\mathrm{V}}_{\text{BLT}}$ (BLT reconstruction, heat map), and ${\mathrm{P}\mathrm{T}\mathrm{V}}_{\text{BLT}}$ (pink contour). Panel b is the CoM deviation and tumor coverage by ${\mathrm{P}\mathrm{T}\mathrm{V}}_{\text{BLT}}$ between the MuriGlo-SARRP and MuriGlo-SmART+ $(\mathrm{n}=4)$.

Using a 0.75 mm uniform margin, the average tumor coverage by the ${\mathrm{PTV}}_{\text{BLT}}$ reached 97.5% [n = 4, Supplementary Fig. S3 (https://doi.org/10.1667/RADE-25-00114.1.S1)]. The corresponding ${\mathrm{G}\mathrm{T}\mathrm{V}}_{\text{BLT}}$ and ${\mathrm{P}\mathrm{T}\mathrm{V}}_{\text{BLT}}$ for individual animals are shown in Fig. S4. Figure 6b also demonstrates that the CoM deviation in BLT target localization for both the SARRP and SmART+ systems remained below 1 mm, and tumor coverage by the PTV_BLT consistently exceeded 97%.

DISCUSSION

While CBCT offers excellent image-guidance, it is less effective at localizing soft-tissue targets in environments with low image contrast. Leveraging the high-contrast BLI, BLT technique has been integrated into irradiators such as SARRP (9) and SmART+ (10) for image-guided radiation studies. Those on-board designs incorporate BLT and irradiator into a single unit. The optical system is often used for longitudinal studies when irradiation guidance or 3D imaging is not needed. However, such an on-board design does not allow irradiator and optical system to be operated independently, which can reduce overall utilization and experimental throughput of both systems for researchers. What is more, the mechanical design of the on-board setting is irradiator-specific, making it difficult to fit the optical system to different irradiator models. In addition, the complex workflow of BLT-guided irradiation, from data acquisition to dose delivery, limits its widespread adoption in the research community.

To meet these pressing needs, we developed a universal BLI and BLT system, MuriGlo, which is compatible with widely used irradiators SARRP and SmART+ to support precision irradiation. Building on our prior BLT-guided imaging and irradiation studies performed with an in-house prototype (9, 11, 12, 17), this work focuses on the practical in-field application of the commercial version MuriGlo. Because the commercial platform is what will be accessible to the scientific community, this work emphasizes BLT-guided workflows and application, complementing a previous study that addresses imaging performance in detail (21).

The system (Fig. 1) is a standalone unit featuring an innovative mouse bed system. This design not only ensures compatibility with current irradiators for image-guided interventions without modifying existing equipment, but also facilitates BL imaging-only applications independent of the irradiator. We have previously validated that the potential position error caused by animal transportation is within 0.2 mm when the optical system is in close proximity (<5 m) to the irradiator (29). Moreover, such positioning error will propagate to the BLI data mapped to the animal CBCT surface (Fig. 2), and ultimately to the BLT reconstruction, which has been accounted for in the radiation margin design. These conditions ensure that the use of a standalone configuration is appropriate for image-guided irradiation. Therefore, this standalone feature significantly enhances both the optical system and the irradiator use. CBCT-guided systems like SARRP and SmART+ are transformative for pre-clinical radiation research, with over 150 machines in use globally. MuriGlo’s compact standalone configuration and high compatibility with existing irradiators provide dual advantages. First, it opens opportunities for institutions looking to advance their image-guided research capability. Second, it allows the substantial existing user base of SARRP and SmART+ to effortlessly incorporate the advanced optical-guided system. We found that the target localization accuracy for the MuriGlo-SARRP and MuriGlo-SmART+ configurations were within 1 mm for the orthotopic GBM model. With the small 0.75 mm margin, we can reach similar tumor coverage >97% for both irradiators. These results confirm that the optical-guided system can universally support both user groups in utilizing the emerging image-guided technique. In addition, we streamlined the workflow of BLT-guided irradiation by incorporating automatic data acquisition, auto-contouring, data mapping, BLT reconstruction, and planning, which we expect will improve BLT acceptance within the community (Fig. 2).

In this work, we also assessed our thermostatic system and illustrated why it is important for the BL application. Maintaining normothermia in animals is crucial. The thermostatic control implemented in the optical system can successfully sustain the animal body temperature at 37°C with minimal variation (Fig. 3a and b). BL signal is the result of chemiluminescent reaction, and its emission efficency (or brightness) can be influenced by ambient temperature (12) which impacts the experiment throughput. In addition to the brightness, utilizing the correct cell spectrum is equally important for the 3D tomographic application, because the multispectral BLT reconstruction relies on the spectrum as the input parameters. Thus, it is important to maintain a consistent spectrum between the in vitro measurement and in vivo BLT application. While a red shift in the emission spectra of GL261-Luc2 with increasing temperature was previously observed using an in-house BLT system (12), the new thermostatic control system implemented in MuriGlo provides a more stable and well-characterized thermal environment. This improved control enables robust confirmation of the temperature-dependent spectral changes of GL261-Luc2 (Fig. 3c). In contrast, the emission spectrum is reporter-dependent (Fig. 3c vs d), as GL261-AkaLuc cells do not exhibit a strong temperature-dependent spectral shift (Fig. 3d) compared with the GL261-Luc2 cell line, highlighting the spectral stability of the AkaLuc reporter. Our results underscore the need for precise temperature control to ensure stable BL emission, spectrum consistency, and physiological conditions in BLI/BLT applications.

MuriGlo is a standalone system that can perform standard 2D in vitro and in vivo studies using bioluminescence reporters, similar to other 2D commercial optical imaging platforms. We adopted a straightforward in vitro assay that users can easily perform and compare their system sensitivity with ours. To evaluate sensitivity for detecting minimal cell quantities, we employed the widely used BL reporter Luc2 and the newer reporter AkaLuc. Our findings reveal that the MuriGlo system can detect as few as 61 GL261-AkaLuc cells. This observation suggests an approximately 21-fold increase in brightness compared to GL261-Luc2 (Fig. 4a). Consistent with the findings of Iwano et al. (20) which highlighted the potential of AkaLuc and Akalumine-HCL system BL imaging in vivo due to its superior brightness. The development of brighter BL reporting systems, such as AkaLuc (20) and NanoLuc (30), further enhances the potential of MuriGlo for effective deep-tissue imaging.

We show the BLT system could lead to a highly conformal plan comparable to clinical delivery (Fig. 5c) using the arc plan. The 5-arc non-coplanar plan offers a greater gantry-angle flexibility than static multi-field plan (12), which reduces planner effort in beam-angle selection for various tumor sizes and improves efficiency for image-guided radiation study. However, if higher throughput is desired while maintaining reasonable conformality, the simplified BLT-guided two-field box plan (Fig. 5d) can serve as an effective alternative. By utilizing the 3D reconstructed tumor volume and considering the uncertainty of BLT target localization ~1 mm (12, 17, 21), we can design the collimator opening to effectively cover the tumor. Although this plan does not improve normal tissue sparing compared to the single beam approach (Fig. 5f), one could leverage the reconstructed volume to devise a more conformal plan, like a 3 or 4-field box, to further reduce toxicity while at the cost of increased planning effort. This simplified BLT-guided planning approach could be considered when specific tumor margin data is unavailable, utilizing generic BLT localization uncertainties, such as 1 mm in our case. One potential concern is that variability in tumor volume may affect BLT localization accuracy, and therefore tumor coverage for radiation. In our previous study using the in-house prototype system, 10 GBM cases (one tumor per animal) were investigated (12). The tumor volume had a mean ± standard deviation of 11.2 ± 3.8 mm³. Despite this variability, BLT-guided targeting accuracy within 1 mm was consistently achieved, demonstrating the robustness of BLT-guided irradiation across a range of tumor sizes.

Recent advances in bioluminescence image-guided systems have improved radiation delivery in less toxic treatment regimens. As noted in a recent review (18), imaging abdominal tumors in small animals is very challenging, and accurate tumor delineation is typically not possible using CT imaging. Tumor motion, such as in pancreatic cancer models, increases BLT localization uncertainty and may introduce larger treatment margins compared with stationary targets like brain tumors, even under anesthesia. Despite this, Deng et al. (17) demonstrated that BLT can be applied to localize the orthotopic pancreatic tumor, allowing conformal irradiation to minimize normal tissue toxicity. Moreover, the deep learning methods under the data-driven supervised learning have been applied to BLT, achieving superior reconstruction performance with faster processing times compared to conventional iterative methods (31, 32). To address the challenge of limited experimental datasets, Deng et al. (33) introduced a self-supervised neural network that consistently reconstructs tumor size and location with high accuracy, even under strong noise conditions and in vivo GBM scenarios. In addition, while surface bioluminescence has been used as a tumor response metric due to its correlation with cell viability (34), BLT leverages the spatial distribution of surface BL signals to enable 3D functional imaging–based assessment of radiation response. Emerging techniques, including sparse orthogonal collimator and inverse planning (35–37), are also under development. The integration of these technologies, together with vendor support to streamline BLT-guided workflows, could significantly enhance the adoption of BLT within the scientific community and support high-precision functional radiation research.

CONCLUSION

The novel BLT system represents a significant innovation in pre-clinical radiation research. The platform was demonstrated to be seamlessly integrated with commercially available irradiators at comparable localization accuracy. The system can support both in vitro and in vivo studies. As we demonstrated versatile planning strategies, investigators can utilize MuriGlo to realize high-contrast optical imaging for preclinical radiation research.

Supplementary Material

Supplementary file 1

Supplementary Fig. S1. (a) shows the position of the thermocouple of the thermostatic control system located at the backside of the 3rd mirror of the 3-mirror system, and (b) shows the front view of the 3rd mirror, and the thermocouple was placed 10 mm away from its symmetry axis.

Supplementary Fig. S2. The overlaps of GTV (blue contour), GTV_BLT (heat map), and MRI for 4 mice are shown in coronal, axial, and sagittal views, respectively. A 0.75 mm margin was added to the GTV_BLT to form PTV_BLT (green contour).

Supplementary Fig. S3. Shows tumor coverage (left axis) and normal tissue coverage (right axis) versus margin expansion of 0, 0.5, 0.75, and 1 mm for 2nd-week-old GBM (n = 4).