2.1.

Materials

A CCD camera was used for image acquisition (C10600, Hamamatsu, Bridgewater, New Jersey) with a F1.4/12 mm lens (HR961NCN, Navitar, Rochester, New York). These components were integrated into a light-insulated luminescence imager that was designed using Solidworks CAD software and 3D printed on a commercially available large-format 3D printer (gMax $1.5+$, gCreate, Brooklyn, New York) using black polylactic acid (PLA, Hatchbox, Los Angeles, California). Several imager designs were produced for applications ranging from in vitro to large animal studies. Light isolation was accomplished using shaped compressible polyurethane (PU) and polyethylene (PE) foam (McMaster-Carr, Douglasville, Georgia).

2.2.

Device Design and Development

The new portable luminescence imager was designed to image wounds on large animal models in a room with low ambient light. Key design parameters for the portable luminescence imager included minimal size, high-luminescence sensitivity, versatility between different imaging scenarios, and complete isolation from ambient light. The device has three major components: a CCD camera, an optical lens, and a custom-designed imaging chamber. A diagram of the imager and its components is shown in Fig. 1(a).

Fig. 1

A breakdown of all optical components of the new imaging device is shown here. (a) The device utilizes a multipart design consisting of a CCD, the body of the imaging chamber, an insertable sliding white light portal, and interchangeable contoured bases. Luminescence and white light imaging can be switched between with the adjustable light portal. (b) A CAD drawing depicts the setup of various components associated with the imaging chamber.

The imaging chamber was designed and 3D printed for ease-of-use in large animal imaging scenarios. The chamber consists of four main parts, including the main body of the device, a sliding white light shutter, a detachable base designed to contour to the surface of the animal being imaged, and compressible light-isolating foam gaskets. A breakdown of these subparts is shown in Fig. 1(b). The main body of the imaging device mounts to the camera and detachable base. It also contains integrated handles and an on-board standard Video Electronics Standards Association mount to allow for hand positioning or attachment to an articulating arm for hands-free use. The white light shutter allows for rapid exchange between white light and luminescence imaging scenarios with minimal user interaction. The detachable base of the design is contoured to interface directly with the surface of the animal being imaged (Fig. 1). Two revisions of the device were designed with lenses of different working distances to showcase the device’s utility in different large animal imaging scenarios (Fig. 2). The devices allowed different view areas to be captured: $5.14\times 4\mathrm{cm}$ [Fig. 2(a)] and $4.75\times 3.5\mathrm{cm}$ [Fig. 2(b)] for a zoomed-in or zoomed-out view as desired. In both designs, a base matching the contour of a pig’s back was created. For in vitro studies of the effect of animal curvature on imaging, a curved imaging stage with the same contour was also designed and 3D printed. Finally, compressible PU or PE foam gaskets are placed at interfaces to block ambient light. The foam gaskets form the base contours to shield the camera from ambient light for the porcine wound study (Fig. 2).

Fig. 2

Two designs of a luminescent imager with different working distances. In both designs, white light illumination is provided by a mechanical light shutter. The device can be positioned on the region of interest using a pair of handles or mounted to an articulating arm. (a) An imager for imaging larger wounds ($>3\mathrm{cm}$ diameter) has a 12-cm working distance and (b) an imager for imaging smaller wounds ($<3\mathrm{cm}$ diameter) has a 3-cm working distance.

By comparing physical parameters with a representative commercial imager—the Kodak In Vivo FX Pro—we have determined that the new imager is significantly more compact (new imager: $22\times 22\times 25\mathrm{cm}$, commercial imager: $104\times 61\times 97\mathrm{cm}$) and lightweight (new imager: 5 kg, commercial imager: 142 kg) than a standard black-box commercial imaging device.

2.3.

Image Acquisition and Processing

Image acquisition was handled using software provided by the camera manufacturer (HCImageLive). Images were processed using PortableView, an in-house developed software described previously.³² Briefly, this software allows the user to generate manual, semiautomatic, and fully automatic regions of interest (ROIs). It then generates absolute and statistical information from the ROIs and allows the user to superimpose up to three images at a time. For studies involving multiple time points, this allows images to be overlaid and ROIs to be automatically extrapolated to generate ROIs over the same area measured in the initial image. The manual ROI generation tool allows the user to draw freehand or use an adjustable ellipse. The semiautomatic tool allows the user to adjust an intensity threshold within the ROI. The fully automatic tool allows the user to select a contiguous area of interest from which the algorithm will generate an ROI from a user specified iso-intensity level. PortableView utilizes a MATLAB^®—(Mathworks, Inc., Natick, Massachusetts) based interface with object-oriented C++ wrapped functions (Microsoft Visual Studios) for computational efficiency.

2.4.

Imager Characterization

The effectiveness of the light isolation strategy was evaluated by measuring background intensity while imaging in room lit with standard incandescent lighting measuring $\sim 300\mathrm{lux}$ (20 min, $8\times 8$ binning). Images taken using the same exposure parameters in total darkness were used as a control. Light background was compared with the generation 1 imager developed in our previous work and with the industry-standard Kodak In Vivo FX Pro.

The sensitivity limit of detection (LOD) of the portable imager’s ability to detect ROS-associated luminescence in vitro was compared with the Kodak using a calibration curve.⁴⁵ Two hundred $\mu \mathrm{L}$ of five concentrations of ${\mathrm{H}}_2{\mathrm{O}}_2$ (2500, 625, 156.25, 39.1, and $0\mu \mathrm{M}$) was added in triplicate for each concentration to wells of a 96-well plate. Twenty $\mu \mathrm{L}$ L-012 ($15\mathrm{mg}/\mathrm{mL}$) were added to each well. The LOD and sensitivity of the imager’s luminescence detection were analyzed. LOD was defined as the lowest concentration each machine is capable of detecting and was calculated using the formula:⁴⁶

Sensitivity was determined by the slope of the linear line of best fit of the standard curve, as previously defined.^47,48

The ability of the imager to detect different concentrations of ROS using a new ROS film was characterized by applying six different concentrations of ${\mathrm{H}}_2{\mathrm{O}}_2$ to L-012-based ROS sensing films (provided by Progenitec Inc., Arlington, Texas; $50\mu \mathrm{L}$; 0, 1, 5, 10, 25, 50, and $100\mu \mathrm{M}$). The established ROS-sensing probe L-012 was used as a control, with $20\mu \mathrm{L}$ of L-012 ($15\mathrm{mg}/\mathrm{mL}$) mixed with $200\mu \mathrm{L}$ of ${\mathrm{H}}_2{\mathrm{O}}_2$ in the wells of a 96-well plate. Sensitivity and LOD were calculated as described above.

Homogeneity was determined on surfaces of different curvatures by arranging consistent droplets of ${\mathrm{H}}_2{\mathrm{O}}_2$ (100 mM, $20\mu \mathrm{L}$) in a matrix pattern on the imaging stage and injecting L-012 in each droplet directly prior to imaging ($5\mu \mathrm{L}$, 48.3 mM). Luminescent signals were acquired for 12 min at $8\times 8$ binning, and consistency at the center and edges of the stage was quantified.

2.5.

Animal Study

The ability of the imaging modality to detect ROS in wounds in vivo was examined a porcine wound model.^9,49 This animal protocol was approved and all animals were cared for according to the standard guidelines approved by the Animal Care and Use Committee at the University of Texas Southwestern Medical Center at Dallas in accordance with the Animal Welfare Act and Guide for the Care and Use of Laboratory Animals.

The imager’s ability to visualize both vascularization and ROS distribution in healing wounds was proven using a porcine wound healing model. Vascularization was observed under white light illumination and ROS observed using luminescence. An established excisional wound model was used.⁴⁹ Briefly, female pigs with a weight between 90 and 120 lbs were used in this investigation. Aseptic technique was used to place six full-thickness wounds (3-cm diameter, 2 wounds per treatment group per animal) on the dorso-lateral surfaces of the animal. After wounding, two wounds per animal were inoculated with $\sim 2000$ colony-forming units of Pseudomonas aeruginosa (American Type Culture Collection 27853) per wound. All wounds were packed with gauze and covered with Tegaderm (3M, St. Paul, Minnesota). Dressings were changed at days 3, 7, 10, 14, 17, and 21. Before dressing changes, a circular section of ROS-sensing film was applied to the wound prior to wound imaging. The imager was then positioned over the wound and the light portal was moved to the open position. A white light image was then acquired for blood vessel visualization. Finally, the light portal was moved to the closed position and a luminescence image was acquired. Wounds images were analyzed using PortableView software as described above. The relationship between blood vessel location and ROS distribution was investigated. Additionally, the change in average intensity over time was compared between infected and uninfected wounds. Finally, the difference in infected versus uninfected wound ROS cluster number and average integrated density (defined as the sum of cluster intensity values divided by cluster area) was analyzed.

2.6.

Statistical Analysis

All values are reported as $\text{average}\pm \text{standard deviation}$. ROS levels and luminescent intensities were correlated using linear regression, with fit measured by the coefficient of determination ($R^2$). Significant difference in ROS-associated intensity between vascular- and capillary-associated areas was determined with a $z$-transformation test as described previously.⁵⁰ Briefly, a test statistic ($Z_s$) was calculated from transformed $p$-values ($Z_i$) from $k$ paired, one-tailed $t$-tests computed for each wound according to the equation:

This value was then compared to $Z_{\mathrm{crit}}$ values for a 95% confidence interval ($\pm 1.96$). The significance in the difference between average ROS levels over time between infected and uninfected wounds was determined with a paired Student’s $t$-test with an $\alpha$ of 0.05. Significant difference between treatment groups in all other studies was determined with an unpaired Student’s $t$-test with an $\alpha$ of 0.05.

3.1.

Imager Characterization

The ability of the novel imager to isolate the imaging environment from ambient light in a lit room was compared to the light isolation solution from the previous fluorescent/luminescent imaging device (blackout fabric). Several adapters were used with varying curvatures (flat versus curved with radius of 25, 20, or 15 cm) to assess the effect of a curved stage interface on light isolation. Images were taken at exposure times of 4 min and $8\times 8$ binning in a lit room and background signal was quantified (Fig. 3).

Fig. 3

Comparison of light isolation efficiency between the generation 1 imager, the new portable luminescence imager, the Kodak In-Vivo FX Pro, and contoured imager-base designs. Images were taken under ambient light with the following imaging conditions: $8\times 8$ binning, 4-min exposure. All bases were found to have acceptable background levels (below 10 a.u.) and represent a significant improvement over the previous imager’s background levels. The new luminescence imager was also found to have significantly lower background than the state-of-the-art Kodak imager.

The new luminescence imager was shown to be light-tight in a procedure room lit with typical standard incandescent overhead lighting ($\sim 300\mathrm{lux}$) for both flat and contoured surfaces, with background values equivalent to total darkness in the imaging room. This was a background intensity reduction of $\sim 30$ times from the generation I imager designed for both luminescence and fluorescence, which enabled the detection of even weak luminescence signals without interference in the new device. The new imager was also found to produce $\sim 5$ times lower background than the Kodak imager, establishing its excellent performance in background reduction even compared to commercially available devices.

The LOD and sensitivity of the portable imager’s response to luminescent signals were compared with the Kodak imager using a calibration curve in vitro (Fig. 4). Fourfold serial dilutions of ${\mathrm{H}}_2{\mathrm{O}}_2$ were added to wells of a 96-well plate in triplicate. L-012 was added to each well immediately prior to imaging ($20\mu \mathrm{L}$, $15\mathrm{mg}/\mathrm{mL}$).

Fig. 4

The sensitivity and LOD of the portable imager were compared with the Kodak imager using the luminescent ROS probe L-012. The new imaging device was found to produce luminescence sensitivity and LOD comparable to the Kodak imager.

The sensitivities of the two imaging devices were found to be $1.0941\pm 0.140\mathrm{a.u.}/\mu \mathrm{M}$ for the portable imager and $1.0996\pm 0.145\mathrm{a.u.}/\mu \mathrm{M}$ for the Kodak. The LOD for the portable imager was calculated to be $0.9\mu \mathrm{M}$ compared to the Kodak’s $18\mu \mathrm{M}$, suggesting that the portable imager can detect slightly lower concentrations of ROS in this scenario than the Kodak imager. In addition, both devices were found to produce an excellent linear relationship between ROS concentration change and luminescence. These results support that both imagers are capable of quantifying low concentrations of luminescent signal with excellent sensitivity.

The sensitivity and LOD of the relationship between ROS concentration and ROS-sensing film luminescence were characterized in vitro (Fig. 5). Several concentrations of ${\mathrm{H}}_2{\mathrm{O}}_2$ within the physiologically relevant range ($5\mu \mathrm{L}$; 0, 1, 5, 10, 25, 50, and $100\mu \mathrm{M}$) were applied to luminescing ROS-sensing film or treated with the luminescent probe L-012. Exposures were taken immediately after treatment.

Fig. 5

The ability of the luminescence imager to quantify ROS was tested in vitro using ROS-sensing film or the probe L-012. The imager was found to be capable of detecting physiologically relevant concentrations of ROS ($<5\mu \mathrm{M}$). ROS-sensing film was found to have significantly lower LOD and higher sensitivity than L-012.

The imager was found to correlate luminescent signal and ROS concentration with a robust linear relationship for both the ROS sensing film and L-012 ($R^2=0.98$). The mechanism governing the higher sensitivity and lower LOD of the ROS sensing film over those of L-012 solution has yet to be determined. It is possible that by adding L-012 into the wound dressing, the ROS-sensing film was packed with a large number of L-012 probes adjacent to each other. Such dense arrangement of L-012 probes may contribute to signal amplification. The imaging system also showed high sensitivity ($k_{\text{film}}=18.9\pm 0.4\mathrm{a.u.}/\mu \mathrm{M}$ and $k_{L\text{-}012}=3.2\pm 0.2\mathrm{a.u.}/\mu \mathrm{M}$) and was able to detect low concentrations (${\mathrm{LOD}}_{\text{film}}=0.11\mu \mathrm{M}$, ${\mathrm{LOD}}_{L\text{-}012}=1.2\mu \mathrm{M}$). The results support that the imager has excellent luminescence sensing properties sufficient for in vivo imaging.

The homogeneity of luminescent signals on surfaces with varying degrees of curvature was investigated to identify any significant decrease or variation in intensity with a consistent ROS signal across the field of view. Curvature was defined as the inscribed circular radius of the base. Droplets of ${\mathrm{H}}_2{\mathrm{O}}_2$ (100 mM, $20\mu \mathrm{L}$) were arranged in a matrix pattern on the imaging surface. Before imaging, L-012 ($5\mu \mathrm{L}$, 48.3 mM) was added to each droplet. We found that there is no significant difference in luminescent intensities detected by imagers with flat and curved bases. In addition, the extent of curvature has no significant influence on luminescent intensity at different regions from the center of the imaging area to the edge (Fig. 6).

Fig. 6

The effect of base contours on luminescence homogeneity was assessed by quantifying a matrix of ${\mathrm{H}}_2{\mathrm{O}}_2$ mixed with the luminescent ROS probe L-012 arrayed over the view area of each stage with various curvatures. We statistically analyzed the luminescent intensities between edge and center (calibrated as 100%) on different platforms. We found that curvature has no statistically significant influence on the luminescence readings by comparing the readings at the center and edge of the base plate.

These results support that the imager with curved imaging bases can be used for imaging curved objectives (such as porcine back or human arms and legs). Although a curved imaging stage is not exactly equivalent to the natural curvature found in living subjects, the lack of significant intensity variation between high and low points of a curved stage simulating biological curvature in this study provides strong evidence that this imaging device can be used in vivo without the necessity to perform additional intensity calibration.

3.2.

Porcine Wound Imaging

The ability of the portable imaging device to correlate ROS distribution via luminescence and blood vessel location via white-light imaging was investigated. Luminescent and white light imaging pairs were used to create overlaid images of ROS levels and blood vessel distribution. Wounds with clearly visible blood vessels at day 1 were analyzed over the 21-day study ($n=7$). As shown in the representative image [Fig. 7(a)] and graphical comparison [Fig. 6(b)] of the change in intensity in wound areas, we find the majority of ROS activities are found in areas near capillaries and far away from large visible vessels (0.5-mm diameter). Large vessels were enhanced in red for increased visibility.

Fig. 7

Luminescence (ROS) and white light (blood vessels) images were captured for each wound. The spatial relations between ROS activity and vasculature are shown in a representative wound. (a) Luminescent ROS images were overlaid on white light images of the wound bed to reveal the relationship between ROS activity and vasculature (shown enhanced in red). (b) Higher intensity was found to inversely correlate with large blood vessel location. This trend was found to be stable over time and statistically significant ($Z_s=0.44$ and $Z_{\text{crit}}=\pm 1.96$), with ROS intensity in capillary associated areas of the wound roughly double than areas associated with large vessels on average.

By analyzing the luminescent intensity and vasculature distribution [Fig. 7(b)], we find that ROS intensity is significantly higher in areas of the wound $\ge 2.5\mathrm{mm}$ away from large visible blood vessels ($Z_s=0.44$ and $Z_{\text{crit}}=\pm 1.96$). The results support that ROS activity is highest in capillary-dense areas of the wound bed, rather than in the vicinity of larger, visible vessels. This is consistent with expectations and current understanding of ROS production in healing wounds.

We also compared the luminescent signals between infected and uninfected wounds over 21 days. A representative image panel showing uninfected and infected wound ROS levels at 1, 3, and 7 days is shown in Fig. 8(a).

Fig. 8

The change in luminescent intensity of infected and uninfected wounds was documented for 21 days with the luminescence imaging system. (a) A representative panel of overlaid luminescence/white light images of infected and uninfected wounds over the first seven days is shown. (b) The line graph of average wound ROS-associated intensity changes over time shows that there are significantly higher ROS levels in infected wounds over the duration of the study, especially on days 7 and 14, than in control wounds ($p=0.019$).

Wounds with high bioburden were found to maintain high ROS levels longer than uninfected wounds ($p=0.019$). This effect was most prominent from day 3 to day 21, with much of the ROS intensity for uninfected wounds remaining very low after 1 week [Fig. 8(b)]. ROS levels in infected wounds were found to peak between day 3 and day 10, whereas uninfected wounds declined quickly after day 3. This finding supports established knowledge of ROS responses in normal and infected wound healing in a new two-dimensional (2-D) model.

In addition to increased ROS activity, we have also observed that infected wounds have more clustered high-intensity ROS areas (Fig. 9). We compared the ROS distribution patterns between infected and uninfected wounds, selecting one time point with the maximum ROS signal for each wound for consistency. To visualize this, surface plots were generated to compare peak height and size between infected and uninfected wounds [Fig. 9(a)]. In MATLAB, intensity clusters above a 70% maximum intensity threshold were detected and quantified for infected and uninfected wounds. The number of clusters in the wound area was then defined and compared between the two treatment groups [Fig. 9(b)]. In addition, average integrated density was compared between the two treatment groups [Fig. 9(c)].

Fig. 9

ROS distributions in infected and control wounds were compared. (a) Wound images for infected and noninfected wounds were analyzed in MATLAB and surface plots were generated. Representative surface plots for both wound categories are shown here, displaying more numerous and sharper peaks in infected versus uninfected wounds. (b) By comparing both groups of wounds, we find that infected wounds have larger numbers of high-intensity ROS activity clusters than control wounds ($p=0.04$). (c) ROS cluster areas were observed to have higher integrated density per square millimeter in infected wounds than in uninfected wounds, supporting the observed sharper ROS intensity peaks in infected wounds ($p=0.014$).

The imaging system discovered a more clustered distribution of ROS over the wound bed in infected than in uninfected wounds, with $\sim 3$ times the number of peaks occurring per ${\mathrm{cm}}^2$ in infected wounds [Fig. 8(b), $p=0.04$]. It was also observed that cluster peaks of infected wounds were significantly higher and sharper compared to the lower, broader peaks of uninfected wounds. Finally, by comparing the average integrated densities of cluster areas, we find that infected wounds have significantly higher cumulative intensity signals in ROS cluster areas than noninfected wounds ($p=0.014$). These results suggest a good relationship between ROS activity distribution, specifically cluster number and area-averaged integrated density, and bacterial colonization in wounds. These findings point to a less homogeneous and “patchier” ROS distribution in wounds with a high bacterial bioburden. Overall results suggest that the “patchier” ROS distribution can be used as an “imaging signature” for wound bioburden diagnosis.