2.1.

Navigation and Tracking Systems

Our “standard” navigation setup was based on the declipse SPECT navigation device (SurgicEye GmbH, Munich, Germany; see Fig. 1). This navigation system incorporates a Polaris Vicra OTS (${\mathrm{OTS}}^{\mathrm{vicra}}$; Northern Digital Inc., Waterloo, Canada) to allow for object position and orientation tracking of geometrically unique reference targets using a fixed pulsed NIR light source [20 Hz; center wavelength (CWL) of 863 nm; full-width-at-half-maximum (FWHM) of 49 nm].^4,9 The pulse on-time per period was about 1.8 ms, resulting in a duty cycle of roughly 4% (see Appendix A).

Next to the declipseSPECT integrated ${\mathrm{OTS}}^{\mathrm{vicra}}$, a prototype OTS (${\mathrm{OTS}}^{\mathrm{orthos}}$; Orthos prototype, Advanced Real-time Tracking GmbH, Weilheim i.OB, Germany) was used. This system allows for manual customization of the pulse frequencies in the 10 to 60 Hz range (CWL: 885 nm, FWHM: 108 nm). The pulse-on time per period was about 0.4 ms, which results in a duty cycle of 0.8% at 20 Hz. To allow for automatic triggering of the pulse frequency, either an analog video signal or digital transistor-transistor-logic signal could be connected via the Bayonet Neill–Concelman (BNC) input connector.

In addition to both these OTSs, a light-emitting diode (LED)-based illumination setup was created, simulating the light source of an OTS. This allowed for controlled variations to be made in both the tracking pulse frequency and duty cycle during the experiments. This configuration consisted of three 850 nm LEDs (CWL: 845 nm, FWHM: 33 nm; HE1-220AC, Harvatek Corp., Hsinchu City, Taiwan) connected to a multichannel universal LED controller (Mightex Systems, Pleasanton, California).

2.2.

Fluorescence Camera Systems

Two open surgery conformité européenne (CE) marked fluorescence cameras were used for this study. First, a PDE camera (Hamamatsu Photonics K.K., Hamamatsu, Japan; European specifications) using a ring of LEDs (CWL: 755 nm, FWHM: 10 nm) to excite the NIR fluorescent dye indocyanine green (ICG). The light detected by this charge-coupled device (CCD) camera is filtered by a $>820\mathrm{nm}$ long pass filter.¹⁰ The resulting video-feed was delivered as an analog 50 Hz interlaced video signal. Second, a VITOM exoscopic camera setup (IMAGE1 S H3-Z FI Three-Chip FULL HD camera head and VITOM^® II NIR/ICG Telescope 0 deg; KARL STORZ Endoskope GmbH & Co. KG, Tuttlingen, Germany; European specifications). In this setup, fluorescence excitation occurs using a prototype D-LIGHT P filtered Xenon light source (CWL: $\sim 740\mathrm{nm}$, FWHM: 120 nm).¹¹ The light collected by the CCD camera was filtered to allow light in the 390 to 670 nm and $>800\mathrm{nm}$ range to be collected.¹² The resulting video-feed was delivered as a digital 50 Hz progressive video signal. A summary of the fluorescence camera and tracking system characteristics is shown in Table 1 and Fig. 4.

Table 1

Characteristics of the fluorescent dye and devices used.

Fig. 4

Spectral environment. (a) The excitation light used by the PDE and VITOM camera and the ICG absorption spectrum. (b) The ICG spectrum and the light spectrum employed by the ${\mathrm{OTS}}^{\mathrm{vicra}}$. Spectral transmission window of the PDE and VITOM is estimated from the literature.^10,12

2.3.

Surgical Fluorescence Phantom and Human Tissue

A surgical fluorescence phantom was fabricated for the experiments. This setup consisted of a previously described, silicone-based phantom⁸ with two fluorescent beads incorporated (see Appendix B). In addition, surgically excised human tissue was used. ICG-nanocolloid (dissolved in saline) was prepared as previously described.¹³ Four 0.1 mL injections, with effectively $40\mu \mathrm{g}/\mathrm{mL}$ ICG, where used for the injection site, where 0.1 mL with 1% of the injected dose was used for the sentinel lymph node.¹³

2.4.

Characterization of the Tracking Interference

All fluorescence camera recordings were performed using an Epiphan frame grabber (DVI2PCIe, Epiphan Systems Inc., Ottawa, Ontario, Canada) integrated within the declipseSPECT navigation system. To quantify the OTS-induced interference, the acquired recordings were analyzed using MATLAB^® (The MathWorks Inc., Natick, Massachusetts). To evaluate the intensity of the interference signals in the fluorescence imaging recordings, the mean pixel intensity (MPI) per video frame was calculated

With MPI in counts, $t$ is the video frame number, $p_i$ is the individual pixel value, and $N$ is the total number of pixels in one frame. To quantify the dynamic character of these interference patterns, a “severity of dynamic interference value” (SDIV) was proposed. To include rapid changes in intensity, while excluding gradual changes, this SDIV was formulated as the standard deviation ($\sigma$) of the first derivative of MPI; i.e., a stable video illumination would result in a low SDIV while a highly dynamic flickering in the video would result in a high SDIV

With SDIV in counts/frame or counts/s, $J$ is the total number of video frames, MPI’ is the first derivative of MPI, and $\overline{{\mathrm{MPI}}^{\prime }}$ is the average thereof.

2.5.

Interference Effect for a Modular Tracking Frequency and Duty Cycle

To quantitatively measure the effect that a variable OTS pulse frequency and duty cycle have on the observed interference during fluorescence imaging, the LEDs were used. To generate light reflection, the LED setup was pointed at a flat surface using a 15 cm distance and the camera systems were placed at a 17 cm distance (Fig. 5). These distances were chosen to maximize the captive nature of the most sensitive setup while preventing saturation of the camera CCD chips. At each LED setting (frequencies: 5 to 100 Hz, duty cycles: 1% to 90%), 2 s video recordings were generated and analyzed (i.e., MPI and SDIV values) for both fluorescence cameras. These experiments were performed in duplicate.

Fig. 5

Schematic overview of the measurements with the LED setup. (a) Setup for the PDE camera and LED setup, both pointed at a black surface. (b) Analogy of the clinical usage of the navigated PDE. (c) Setup for the VITOM and LED setup, both pointed at the black surface. (d) Analogy of the clinical usage of the navigated VITOM.

2.6.

Interference in a Surgical Phantom and Human Tissue

To simulate surgical application, the two navigated fluorescence camera systems were sequentially positioned right above the surgical phantom using a distance of 15 cm. Subsequently, either the ${\mathrm{OTS}}^{\mathrm{vicra}}$ or ${\mathrm{OTS}}^{\mathrm{orthos}}$ was also placed above the surgical phantom using a distance of 130 or 30 cm. To illustrate the difference in interference in the “standard” navigation setup, video recordings were produced and analyzed (i.e., MPI and SDIV values) with the ${\mathrm{OTS}}^{\mathrm{vicra}}$ turned off and on. Furthermore, using the BNC synchronization input of the ${\mathrm{OTS}}^{\mathrm{orthos}}$, the video signal of the PDE was sampled, allowing for synchronization of the tracking frequency with the PDE acquisition frequency. To illustrate the effect of synchronized NIR optical tracking during surgery, video recordings were acquired with the ${\mathrm{OTS}}^{\mathrm{orthos}}$ turned off and on, using a tracking frequency of 20 (unsynchronized) and 25 Hz (synchronized). These recordings were also analyzed for the above-mentioned MPI and SDIV values. Furthermore, the most important measurements, i.e., PDE fluorescence imaging with “no tracking,” “${\mathrm{OTS}}^{\mathrm{vicra}}$ tracking at 130 cm” and “synchronized ${\mathrm{OTS}}^{\mathrm{orthos}}$ tracking at 130 cm,” were also recorded using the human tissue.

2.7.

Gated Tracking Pulse and Fluorescence Imaging for PDE

To create an alternating gated setup, a tracking pulse should only be sent whenever there is no fluorescence image capture (Fig. 6). We assumed the PDE camera would have a “dead-imaging-moment” long enough to send out this tracking pulse, without it being detected by the camera. This was investigated by building and programming a small electronic circuit to sample the frequency timing of the PDE video signal and trigger the ${\mathrm{OTS}}^{\mathrm{orthos}}$ tracking pulse. This circuit was built around an LM1881N video processing chip and an Arduino Uno microcontroller (Arduino AG, Italy).

Fig. 6

Gated tracking pulse and fluorescence imaging. In a sequential fashion, optical tracking pulses are only emitted when the fluorescence camera is not acquiring a fluorescence image.

3.1.

Interference in a Surgical Phantom Setup

Table 2 and Appendix C demonstrate the interference the 20 Hz ${\mathrm{OTS}}^{\mathrm{vicra}}$ induces for the different fluorescence cameras. This interference is the result of the light emitted from the OTS, overlapping with the spectral detection window for which NIR fluorescence cameras are traditionally designed ($>800\mathrm{nm}$). More specifically, the spectral emission of the OTS overlaps with the fluorescence emission of the clinically approved NIR dye ICG (emission max at 820 nm) and more experimental Cy7-analogs (emission max between 795 and 825 nm) (Fig. 4).^4,14,15 For both systems, the MPI and SDIV values rose when the ${\mathrm{OTS}}^{\mathrm{vicra}}$ was turned on (Table 2); a rise of 142.6 counts (PDE) versus 18.5 counts (VITOM) for the mean MPI values and $1757.5\mathrm{counts}/\mathrm{s}$ (PDE) versus $410.0\mathrm{counts}/\mathrm{s}$ (VITOM) for the SDIV values was recorded. As illustrated in Appendix C, the interference in the PDE setup was so severe that it inhibited the differentiation of the fluorescent beads with respect to their surroundings. Using the VITOM setup, the dynamic ${\mathrm{OTS}}^{\mathrm{vicra}}$ light was also clearly visible; however, it did not limit the delineation of the fluorescent beads.

Table 2

Quantitative analysis of the surgical fluorescence phantom measurements with the OTSvicra and OTSorthos. Mean MPI and SDIV values calculated for the PDE and VITOM, with and without the OTS turned on at different settings and distances.

3.2.

Interference Effect for a Modular Tracking Frequency and Duty Cycle

Next to the spectral overlap between the light emitted by the OTS and the spectral detection window of the surgical fluorescence cameras, exposure to the variations in LED light pulse frequencies and duty cycles influenced the interferences (Fig. 5). Plotting of the MPI and SDIV revealed the LED settings affected the two cameras differently (Fig. 7 and Appendix D). Of both cameras, the PDE proved most sensitive to the NIR LED light, yielding substantially higher MPI values at identical settings; e.g., a mean MPI of 59.3 counts for the PDE and 26.9 counts for the VITOM at a frequency of 20 Hz and duty cycle of 3.2%. As can be seen in Appendix D, the intensity of the MPI decreased with decreasing duty cycles (i.e., duty cycles approaching 0%). Figure 7 reveals the duty cycle settings also influenced the SDIV values, showing a decrease in the intensity of the dynamic interference when duty cycle values approached either 0% or 100%. For both cameras, minimal dynamic interference (i.e., low SDIV value) was observed at tracking pulse frequencies of 25, 50, and 75 Hz, which is in line with recording frame rates of 25 frames per second. Unfortunately, a stable interfering background illumination remained (i.e., stable nonzero MPI values). These results indicate that the mismatch between the frequency of the camera recording frame rate and the OTS tracking light pulse cause the dynamic character (flickering) of the interference.

Fig. 7

Visualization of the severity of dynamic interference for the fluorescence cameras using the LED setup. (a) The SDIV results for the PDE in 3-D-view (left) or top-view (right). (b) The SDIV results for the VITOM in 3-D-view (left) or top-view (right).

3.3.

Synchronized Tracking Frequency for the PDE in Both Phantom and Human Tissue

Given the above-reported findings, we reasoned that using an alternative OTS with controllable tracking frequency could help solve or reduce the severe flickering effects. Using the surgical phantom setup, Table 2 and Appendix C illustrate that a 20-Hz ${\mathrm{OTS}}^{\mathrm{orthos}}$ already reduced the interference intensity fourfold compared to what was observed for the ${\mathrm{OTS}}^{\mathrm{vicra}}$: mean MPI values of 42.2 counts versus 181.3 counts, respectively. This improvement is most likely a combination of a shorter pulse-on time (i.e., a lower duty cycle) and a lower light intensity. When the ${\mathrm{OTS}}^{\mathrm{orthos}}$ pulse frequency was matched with the PDE recording frequency (at 25 Hz), SDIV values approached zero and the dynamic character of the interference was completely eliminated (Table 2 and Appendix C), thus providing comparable SDIV values with the ${\mathrm{OTS}}^{\mathrm{orthos}}$ turned on and off. However, mean MPI values differed between 54.8 counts with the ${\mathrm{OTS}}^{\mathrm{orthos}}$ turned on and 32.6 counts with the ${\mathrm{OTS}}^{\mathrm{orthos}}$ turned off (distance of 130 cm). Despite this stable background, differentiation between fluorescent beads and background in the phantom setup was possible. Figure 8 clearly shows that similar results are found when this concept is applied to human tissue; where fluorescence imaging was rendered unfeasible during tracking with the ${\mathrm{OTS}}^{\mathrm{vicra}}$ while fluorescence imaging became possible during tracking with the synchronized ${\mathrm{OTS}}^{\mathrm{orthos}}$.

Fig. 8

Evaluation of optical tracking during fluorescence imaging for human tissue. (a) Excised human tissue with the PDE camera situated 15 cm above the tissue. (b) Fluorescence imaging without any tracking. Both injection site (IS) and sentinel lymph node (LN) are visible. (c, d) Example video frames for fluorescence imaging during tracking with the (c) ${\mathrm{OTS}}^{\mathrm{vicra}}$ and (d) synchronized ${\mathrm{OTS}}^{\mathrm{orthos}}$.

3.4.

Gated Tracking Pulse and Fluorescence Imaging for PDE

In an attempt to completely remove the interference during fluorescence imaging, including the background illumination, an alternating gated tracking and fluorescence imaging setup was pursued. Unfortunately, it was not possible to identify a “dead-imaging-moment” capable of harboring the ${\mathrm{OTS}}^{\mathrm{orthos}}$ tracking pulse. Consequently, we were not able to get rid of the stable background illumination.