2.1.

Instrumentation

The customized CW-NIRS system [Fig. 1(a)] was composed of illumination and detection components to acquire transient light intensity changes from the mouse. The sources of the CW-NIRS systems consisted of six pairs of dual-wavelength (785 and 850 nm) laser diodes with collimation tubes and optics (LT240P-B, Thorlabs). Each pair of collimated lights was combined into a long and light-weight glass optical fiber ($105\mu \mathrm{m}$ in core diameters) with a multimode coupler to reduce the number of optical fibers. For greater sensitivity, four avalanche photodiodes (APDs) (APD120A, Thorlabs), also connected to optical fibers, were used with a customized amplifier and low-pass filter circuit to remove high-frequency noise. Each source switching and detecting sequence was controlled by a 16-bit microcontroller unit (MCU) (Atxmega128A1, AVR). The sampling rate was 5 Hz and the system communicated via serial communication through MATLAB (MathWorks, Inc., Natick, MA, USA). Figure 1(b) shows the overall schematic diagram of our implementation.

Fig. 1

Custom-made CW-NIRS system and experimental setup: (a) customized CW-NIRS system consisting of an MCU, laser diode, coupler, APD, and optical fiber connector and (b) schematic diagram of the experiment. CW-NIRS, continuous-wave near-infrared spectroscopy; MCU, microcontroller unit; APD, avalanche photodiodes; and PC, personal computer.

2.2.

Animal Preparation

In total, 16 male C57BL/6 mice (age 9 weeks, Orient Bio Inc., Seongnam, Republic of Korea), weighing 25 to 27 g, were used in this experiment. This experiment was approved by the Korea University Institutional Animal Care and Use Committee (IACUC-2018-49) and carried out in accordance with the Korea University guidelines. Before any surgical procedure, 2% to 5% isoflurane was administered to the mouse until it was anesthetized; then the mouse was placed in a stereotaxic frame. To maintain the mouse’s body temperature, a heating pad (maintained at $37°\mathrm{C}\pm 1°\mathrm{C}$) was placed under the body, and the head was tightly fixed using a nose cone and ear bars. Stereotaxic coordinates were determined with respect to the bregma. The scalp was first shaved, disinfected with ethanol, and then removed with spring scissors. The flaps were removed to make the skull visible. The cranial surface was carefully cleaned with 4% alcohol and scraped to expose $\sim 1.5\times 1.0\mathrm{cm}$ of the skull. The duration of the surgical procedure was $\sim 15\min$. Thereafter, the customized 3D head cap printing process was immediately started.

2.3.

3D-Printed, Customized Head Cap

To ensure stable attachment of the optical fibers onto the mouse skull, we developed a novel 3D printing technique to accommodate the skull’s size and shape. After surgery, to prepare a 3D customized head cap, the shape of the skull’s surface was acquired using a customized molding frame made of alginate and plaster [Fig. 2(a)]. The molded object was scanned using a 3D scanner (EinScan Pro, Hustem Co., Ltd., Gangnam-gu, Seoul) to acquire the 3D mesh. The mesh was refined using a 3D design program (Rhino 3D, Rhinoceros, Robert McNeel & Associates, Seattle, WA), including placement of the probe holes (distanced $\sim 2.5\mathrm{mm}$ apart) [Fig. 2(b)], and then it was 3D printed. For customized 3D head cap printing, we used a stereolithography apparatus 3D printer using a biocompatible ultraviolet curable polymer (3DMaterials). The head cap was bonded to the skull using biocompatible dental glue (Super Bond C&B, Sun Medical Co., Ltd., Shiga, Japan) [Fig. 2(c)]. After hardening, optical fibers were inserted into the holders and fixed. Ferrules and sleeves were used to prevent twisting and minimize damage to the optical fiber by biting or grooming. By separating ferrule and sleeve, the animals could perform repeated experiments [Fig. 2(d)]. The total weight of the printed head cap and the long optical fibers was $\sim 1\mathrm{g}$.

Fig. 2

Customized 3D head cap printing technique: (a) a cast of the skull was formed using alginate and plaster; (b) the mold was converted to a 3D mesh; (c) the 3D-printed head cap with customized channel layout was placed on the mouse skull (Video 1, MP4, 8.4 MB [URL: https://doi.org/10.1117/1.NPh.8.2.025013.1]; Video 2, MP4, 5.7 MB [URL: https://doi.org/10.1117/1.NPh.8.2.025013.2]); and (d) a ferrule-sleeve system was used for repeated experiments.

Since the 3D rendering software determines probe position, the optical fiber layout can easily be modified depending on the experiment’s purpose. This study fixed the optical fibers [six sources and four detectors, Fig. 3(a)] to target the visual and somatosensory cortex in the mouse brain [Fig. 3(b)]. This layout resulted in a total of 13 NIRS channels [Fig. 3(c)].

Fig. 3

Customized head cap and estimated channel position on a model brain and experimental protocol: (a) estimated probe position overlayed onto a 2D model of the mouse’s brain; (b) estimated channel position according to the source-detector layout; (c) regions of the mouse cortex; and (d) experimental protocol.

Preparing a 3D head cap took approximately an hour for each mouse. It took about an hour and a half from surgery until head cap attachment. Thereafter, the anesthesia group immediately started the experiment, but the freely moving group started the experiment the next day after awakening from the anesthesia, adaptation, and stabilization.

2.4.

Visual Stimuli

A visual stimulation task was chosen to validate our system by monitoring the mice’s hemodynamic response. A stimulation kit was designed to verify the customized CW-NIRS system. The stimulation kit consisted of a rectangular-shaped box ($590\mathrm{mm}\times 590\mathrm{mm}\times 610\mathrm{mm}$) made of black acrylic. In total, 16 mice were divided into two arbitrary groups for the task, including the freely moving group ($N=8$) and anesthesia group ($N=8$). The experiments were arranged in a block paradigm, consisting of four rest periods (300 s) and four stimulation periods (30 s) [Fig. 3(d)]. On the inside of each wall, green LEDs (490 to 510 nm, LXZ1-PE01, Lumileds LLC, San Jose, CA) were attached to use as light sources for the visual stimulus. The optical fibers reach into the stimulation kit through a hole (diameter 80 mm) in the upper side of the box, allowing them to move following the mouse. The fiber length was 1 m, long enough to allow a mouse to freely move within the stimulation kit area ($590\mathrm{mm}\times 590\mathrm{mm}$), which is sufficient for visual stimuli tasks.

In the anesthesia group, each mouse was anesthetized using 2% to 5% isoflurane and 100% ${\mathrm{O}}_2$ inhaled through anesthesia tubes during the task, and they were placed in the middle of a stimulation kit. Experiments typically lasted 1 h and were followed by euthanasia. For experiments in awake mice, we placed the mouse on the middle of the customized stimulation kit, and experiments started immediately after placing them on the box. The mice were allowed to move freely in the stimulation kit while the green LEDs presented the visual stimulus. The unanesthetized mouse was allowed to move freely without any barriers in the closed box. All tasks were monitored and recorded by an infrared camera (5-MP Night Vision Camera for Raspberry Pi, SEN0184, DFRobot, Shanghai, China) placed on the top of the kit (Videos 1 and 2).

2.5.

Data Processing and Analysis

All analyses were performed using custom scripts written in MATLAB. Two time-series light intensity changes were acquired at each wavelength to calculate oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) concentration changes from optical density according to the modified Beer–Lambert law.²⁹ The hemodynamic changes were time-filtered using a 0.009 to 0.08 Hz first-order Butterworth band-pass filter to remove high-frequency noise that could affect the analysis while also maintaining important spontaneous hemodynamic fluctuations.^30–32 Brain activation map images were spatially smoothed with a Gaussian filter using a sigma of 1.25 mm. The grand block average of HbO and HbR concentration changes was obtained by averaging over all mice with a stimulation-related response in each group. To compare cortex activity, we used the mean values of the concentration changes in HbO and HbR (meand ± SD) obtained from all channels in each group. The mean HbO and HbR values were compared within channels using a paired $t$-test to determine the differences in HbO and HbR changes between corresponding channels. To verify the significance of differences between freely moving and anesthesia groups, the mean HbO and HbR values were compared across groups using a two-sample $t$-test. The significance level was set at $p<0.05$.

3.1.

Cortical Hemodynamic Responses

The grand average waveforms of HbO and HbR concentration changes during the visual stimulation task in both groups are shown in Fig. 4. The grand average of the HbO in the freely moving group slightly increased in the visual cortex (channels 9, 11,12, and 13) during the task period, whereas that in the anesthesia group did not show substantial changes. In the freely moving group, the HbO responses for movements could not be observed in the motor cortex (channels 1, 2, 3, and 5) because the hemodynamic responses following movement were filtered to clearly identify those in the visual cortex (see the Supplementary Material). In both groups, HbR signals had a higher hemodynamic variability during the task, which corresponds to random noise but would not removed by filtering.

Fig. 4

Time course of changes in HbO and HbR concentration during the visual stimulation task in each channel. The horizontal axis represents time course. The vertical axis represents changes in HbO and HbR concentration. The solid line (HbO: red; HbR: blue) shows hemodynamic responses. The black solid line shows the stimulation start time. Yellow box indicates the stimulation period. Differences in HbO response between mice in a freely moving state or under anesthesia during visual stimulation: (a) freely moving group and (b) anesthesia group. HbO, oxyhemoglobin; HbR, deoxyhemoglobin; and CH, channel.

3.2.

Imaging of Hemodynamic Changes

Figure 5 shows cortical activation maps and group averaged HbO for the visual stimulation task in freely moving and anesthetized mice. The HbO response increased in the visual cortex (channels 9, 11, 12, and 13) when visual stimulation was performed for freely moving animals [Fig. 5(a)]. Under anesthesia, the HbO response increased in channels 9; however, there were no significant changes in the HbO response for channels 11, 12, and 13 [Fig. 5(b)]. The mean values of HbO concentration changes were compared between the freely moving state and anesthesia [Fig. 5(c)]. Visual stimulation tasks induced significantly higher activation in the visual cortex (channels 9, 11, 12, and 13) in the freely moving group than in the anesthesia group. Table 1 shows the mean HbO values and statistical difference between the freely moving and anesthesia groups. All channels in the visual cortex had a significant difference for HbO ($p<0.001$). The $p$-value confirmed the visual analysis of results reported in Figs. 5(a) and 5(b). The averaged hemodynamic responses in channels 11, 12, and 13 in the anesthesia group had a very small amplitude as compared with those in the freely moving group, indicating no visual stimulation was received under anesthesia.

Fig. 5

Differences in HbO response between mice in a freely moving state or under anesthesia during visual stimulation: (a) 2D HbO image of freely moving mice during visual stimulation; (b) 2D HbO image of anesthetized mice during visual stimulation; and (c) statistical results of mice in a freely moving state or under anesthesia during visual stimulation. HbO, oxyhemoglobin and CH, channel.

Table 1

Average activation in brain regions during visual tasks in HbO.