1.

Introduction

As the neuroscience community strives to gain more detailed insight into cellular and vascular mechanisms underlying brain function, techniques and practices for measuring brain activity in awake animals have become indispensable.^1,2 The confounding effects of anesthesia vary notably with different anesthetics and heterogeneously impact different regions of the brain and over different timescales, and their influence on energy metabolism, hemodynamic signaling, and neuronal circuitry can be especially problematic for investigations of task-evoked neurovascular coupling or microvascular dysfunction.^3,4 A wide range of custom head- and body-fixation and habituation protocols have been engineered for imaging studies in awake animals.^5,6 Imaging brain activity in awake animals avoids the complications of anesthesia and enables a more accurate correlation between observations of brain pathology with cognitive and behavioral assessments. However, experimental readouts from awake subjects are highly susceptible to perturbations associated with spontaneous, naturally occurring behaviors. Awake animals constantly engage in whisking, grooming, and other spontaneous “fidgeting” motions, and they rapidly fluctuate among different states of alertness. These frequent, transient events can notably alter neuronal and hemodynamic activity or introduce motion artifacts in the experimental readouts,^7–10 warranting a large number of experimental trials. This can be particularly problematic for longitudinal studies and investigations exploring task-evoked activation.

Although designing firmer head fixation methods or motion-correction algorithms helps to mitigate their impact, spontaneous fidgeting and variations in alertness notably disrupt optical measurements of brain function and impair robust interpretation of results. These behaviors are naturally unavoidable and warrant the urgent need for robust monitoring methods. In practice, investigators utilize an array of peripheral sensors and cameras to monitor an animal subject’s behavior during imaging experiments. The use of multiple, peripheral sensors can often be cumbersome, particularly if the sensors are bulky or require careful positioning within an optical imaging system’s often-limited sample space. Precise positioning and testing of multiple sensors can prolong experimental setup time. In some labs, each peripheral sensor is recorded with a different acquisition software, often asynchronously from the primary brain imaging measurements, which confounds data processing. Because the duration of the head restraint should be limited to $\sim 2\mathrm{h}$ for awake mouse subjects each day,¹¹ a lengthy setup and testing process for multiple sensors significantly reduces the available time for viable data acquisition. In addition, the assessment of animal behavior data must usually be performed offline after completing the full imaging experiment. The inability to review measurements immediately after acquisition often results in an insufficient number of valid measurements per experiment, significantly increasing the number of experiments and animal subjects required to complete a study.

Recognition continues to grow for the wide range of factors contributing to fluctuations in brain activity, both under resting conditions and in response to a functional stimulus.^7–9 The growing number of physiological and behavioral parameters for monitoring significantly increases the complexity of experimental setups and warrants techniques for centralizing these complementary measurements into a convenient all-in-one system. Zhang et al.¹² provided detailed guidance to monitor multiple behavioral and physiological metrics concurrently with brain activity measurements using custom data acquisition software, illustrating the indispensable importance of behavioral monitoring and the notable convenience of a robust and centralized acquisition system. In this work, we demonstrate a low-cost platform to streamline multiple readouts of animal behavior from compact peripheral sensors to a single, portable custom acquisition device. The total cost of this platform is under $500, and all data acquisition and processing are performed using a portable microcomputer. The customized, compact sensors can be quickly and easily positioned, greatly reducing imaging preparation time and enabling simple integration with custom and commercial imaging systems. Measurements from our device can be immediately reviewed after completing a sequence of experimental trials, and the recordings can be easily synchronized with readouts from multiple brain imaging modalities such as two-photon microscopy or optical coherence tomography. This work constitutes our pilot efforts to develop a user-friendly, affordable device that easily and efficiently couples with intravital imaging systems to monitor critical physiological and behavioral readouts for awake brain imaging in rodents. To validate its performance, we coupled the device with our custom wide-field optical imaging system (WFOI) for functional activation experiments. Specifically, we used our device to monitor awake mouse behavior while measuring stimulus-induced cortical hemodynamic and neural activity with WFOI in Thy1-GCaMP6f mice. The results confirm that our integrated platform efficiently detects the animal’s spontaneous motion during experiments and the dynamic fluctuations in pupil diameter, demonstrating its convenience and ease of use for physiological monitoring during awake brain imaging experiments.

2.

Results

We designed a user-friendly and low-cost device to monitor animal physiological signals and motion for awake imaging experiments, and we validated its performance with functional stimulation experiments. The general design is shown in Fig. 1(a). It requires minimal setup time and facilitates our neuroimaging measurements by providing real-time feedback on animal behavior.

Fig. 1

(a) Schematic of the integrated platform setup on an awake mouse with a cranial window. (b) The integrated platform design consists of a data acquisition system and an animal fixation frame. b(i)–b(v) Components of the data acquisition system: (i) BMA250 accelerometer, (ii) Jetson Nano microcomputer, (iii) body camera, (iv) pupil camera, and (v) 16-bit A/D converter. The system’s commercial components are provided in Table S1 in the Supplementary Material.

2.2.

Software Design

2.2.1.

GUI on microcomputer

A custom graphical user interface (GUI) was developed in Python 3.6.9 with the Tkinter toolbox, specifically for the Jetson Nano microcomputer. The GUI incorporates several features, including (1) real-time display from body and pupil cameras, (2) automatic data acquisition initiated by a 5 V trigger signal, and (3) real-time display of acceleration during data acquisition [Fig. 2(a)]. Immediately after each experimental run, the camera and accelerometer data can be easily reviewed to quickly screen for spontaneous behaviors.

Fig. 2

(a) Custom GUI with real-time display of acceleration and frame-wise feedback from two cameras. (b) Stepwise images of the pupillometry algorithm.

2.2.2.

Pupillometry analysis

To monitor dynamic changes in pupil size, we embedded an algorithm developed with the OpenCV library in our custom data acquisition software.¹⁰ The analysis procedure is developed in Python 3.6.9, and the step-by-step procedure is illustrated in Fig. 2(b). Briefly, after acquiring the raw grayscale image, the pixel value is inverted. Thresholding is then applied to binarize the image. After thresholding, small areas and regions with low circularity are removed to isolate the pupil region. In select frames, momentary whisker twitching will result in the pupil being blocked and inadvertently segmented into two parts. To correct this, the divided pupil region is reconnected and fitted as an ellipse. The pupil diameter is calculated by the number of pixels spanning the minor axis of the ellipse. Pupil diameter is low-pass filtered below 1 Hz with a fourth-order Butterworth filter.

2.3.

Testing and Validation Via Functional Activation Experiments

2.3.1.

Rapid screening of motion artifacts

We tested our integrated device’s utility to quickly detect motion artifacts during functional stimulation experiments. The somatosensory cortices of awake Thy-1 GCaMP6f mice were imaged with our WFOI setup while the integrated device monitored their behavior and measured pupil diameter. Imaging sessions were performed to monitor stimulus-induced changes (pneumatic whisker deflection, 300-ms airpuff, 3-Hz stimulus train, 3-s duration per trial) in blood volume by optical intrinsic signal (OIS) (reflectance measurements $\lambda =568\pm 10\mathrm{nm}$, $5\text{frames}/\mathrm{s}$, 30 s per trial). In separate imaging sessions, stimulus-induced changes in neuronal calcium dynamics were monitored by GCaMP fluorescence (fluorescence measurements ${\lambda }_{\mathrm{ex}}=470\pm 40\mathrm{nm}$, ${\lambda }_{\mathrm{em}}=525\pm 50\mathrm{nm}$, $20\text{frames}/\mathrm{s}$, 12.5 s per trial). To identify spontaneous motion artifacts, the acceleration recordings were screened in 0.5 s intervals. A motion artifact was defined as any instance during which the acceleration reading exceeded $0.2\mathrm{m}/{\mathrm{s}}^2$ for 200 ms or longer. Figure 3 displays representative results from individual imaging trials with and without motion artifacts. Pneumatic whisker deflection provoked dynamic, local increases in both hemodynamics and neural activity [Figs. 3(a) and 3(b)], consistent with prior studies.^17,18 A concomitant increase in pupil diameter was observed for each stimulation trial. Example trials of both viable and corrupted OIS measurements lasting 30 s are displayed in Fig. 3(b), along with GCaMP fluorescence measurements lasting 12.5-s durations, collected separately [Fig. 3(d)]. For several stimulation trials, the whisker stimulus evoked a confounding motion artifact, yielding distinct fluctuations in the accelerometer recordings. Example motion artifacts are highlighted by shaded blue regions in Figs. 3(b) and 3(d). In trials with no motion artifacts, response profiles of cerebral blood volume (CBV, calculated as normalized changes in reflectance, $\mathrm{\Delta }R/R_0$), neuronal GCaMP fluorescence ($\mathrm{\Delta }F/F_0$), and pupil diameter resembled well-established measured profiles reported previously.^3,12 When spontaneous motion artifacts occurred, unpredictable fluctuations were observed in all readouts. Pupil dilations also persisted for a longer duration in the presence of motion artifacts, reflecting the initiation of the animal’s fight-or-flight response.¹⁹

Fig. 3

Representative OIS ($\mathrm{\Delta }R/R_0$) and GCaMP ($\mathrm{\Delta }F/F_0$) measurements performed using WFOI and our integrated device. (a) OIS imaging of the functional hyperemia center (red rectangle) in the barrel cortex through a 3-mm cranial window. (b) Four representative trials showing 30-s time courses of total hemoglobin change, pupil diameter, and acceleration. (c) GCaMP imaging of the neural activity center (green rectangle) through the same cranial window. (d) Four representative trials showing 12.5-s time courses of GCaMP fluorescence, pupil diameter, and acceleration. Gray shaded bars represent the pneumatic whisker stimulus, and blue indicates spontaneous motion artifacts. The total blood dynamic change signal (red curves) was deliberately inverted to accurately reflect the increase in blood volume.

2.3.2.

Correlating pupillary response with cortical blood volume and neuronal calcium

We further validated our device’s performance by analyzing and correlating kinetic features of functional imaging data collected with WFOI and our integrated device. Specifically, we correlated stimulus-induced pupillometry changes acquired from our device with vascular and neuronal measurements from the cortex acquired with WFOI. The data were first screened to remove trials corrupted by spontaneous motion artifacts. Block-averaged time courses of OIS versus pupil diameter and GCaMP versus pupil diameter are displayed in Figs. 4(a) and 4(c). We calculated the cross-correlation between these combinations of signals as well as the onset time and rise time of each signal from Figs. 4(b) and 4(d). The pupil diameter measurement was detrended before cross-correlation. The rising time was defined as the time from 10% to 90% of the peak height after the beginning of the first stimulus, and the onset time was defined as the beginning of the first stimulus to 10% of the peak height.

Fig. 4

Trial-averaged results of (a) OIS ($\mathrm{\Delta }R/R_0$) versus pupil diameter and (c) GCaMP ($\mathrm{\Delta }F/F_0$) versus pupil diameter in response to whisker stimulation. Gray shaded bars indicate stimulus pulses. (b) Cross-correlation between changes in total hemodynamic changes and pupil diameter 5 s after stimulus onset. The red dot indicates the peak correlation value. (d) Cross-correlation between GCaMP fluorescence and pupil diameter 4 s after stimulus onset. The green dot indicates the maximum correlation location. Panel (e) and (f) are the onset time (*$p=0.0439$ and ***$p=0.0003$) and rising time (*$p=0.0268$ and ****$p<0.0001$) comparisons between pupil diameter and optical results. Statistical difference was tested using Student’s $t$-test (mean ± SD).

Figure 4(b) relates the kinetics of stimulus-induced pupillary dilation and cortical blood volume. The responses of cortical blood volume and pupil dilation were strongly correlated ($0.77\pm 0.0033$). Our onset time analysis indicates that time pupil dilation initiates nearly 300 ms before the change in cortical blood volume [pupil onset time: $0.394\pm 0.168\mathrm{s}$, $\mathrm{\Delta }R/R_0$ onset time: $0.763\pm 0.722\mathrm{s}$, Fig. 4(e)]. Pupil diameter and OIS rose from 10% to 90% of their peaks in $1.957\pm 0.928\mathrm{s}$ and $1.397\pm 0.558\mathrm{s}$, respectively [Fig. 4(f)]. The results indicate that, although pupillary dilation initiates earlier than vascular changes in the cortex, pupillary dilation takes longer than cortical blood volume to reach its stimulus-induced peak, in agreement with previous reports.^{10,14,20–22}

Our measured pupil diameter changes also correlated well with evoked increases in calcium in cortical Thy1 neurons, with a correlation value of $0.43\pm 0.014$ [Fig. 4(d)]. However, the kinetic features of GCaMP fluorescence differed considerably from pupil dilation. Neuronal calcium increased and decreased rapidly in response to each stimulus pulse, whereas the pupil dilated more gradually as the stimulus train persisted. Our calculations indicate that the onset time of neuronal GCaMP fluorescence ($0.09\pm 0.03\mathrm{s}$) occurred earlier than that of the pupil [$0.283\pm 0.15\mathrm{s}$, Fig. 4(e)]. The rise time for GCaMP ($0.133\pm 0.043\mathrm{s}$) was also considerably shorter than pupillary dilation [$1.475\pm 0.699\mathrm{s}$, Fig. 4(f)].

Taken together, our observations suggest notable differences between the kinetics of stimulus-evoked neuronal and vascular responses in the rodent cortex compared with its pupil diameter. Stimulated changes to pupil diameter appear to initiate earlier than changes in cortical blood volume but not neuronal calcium. Pupil dilation also requires longer to reach its respective stimulus-induced peak than cortical blood volume or neural activity.

3.

Discussion and Conclusion

Methods and tools for awake animal imaging have grown increasingly necessary to gain detailed insight into the intricate processes that dynamically regulate brain function.^1,2 These techniques are especially vital to investigations exploring neurovascular coupling. As a result, recognition has grown for the need to account for spontaneous perturbations to brain activity. Behavioral monitoring is an integral component of virtually all brain imaging experiments in awake subjects, and numerous approaches have been implemented to identify factors that could modulate cerebral hemodynamics and neural activity.^12,23 Our custom device easily and effectively permits real-time screening of naturally occurring, intermittent phenomena in small animals that modulate brain activity, such as spontaneous locomotion, whisking, and grooming activity, and variations in alertness. The device consists of inexpensive (<$500), portable components and streamlines measurement and analysis using a powerful microcomputer. The compact sensors can be easily configured and positioned, significantly reducing experimental setup time to under 5 min. The customized, triggerable, user-friendly software provides immediate assessment of behavioral data following acquisition and permits easy, precise synchronization with two-photon microscopes, or other intravital brain monitoring systems. Our pilot results demonstrate that the device can easily identify spontaneous motion artifacts and fluctuations in alertness during functional activation trials. In practice, during an imaging session, the software allows investigators to review behavioral readouts immediately following a sequence of imaging trials. Such rapid feedback affords opportunities to immediately determine whether more stimulus trials can or should be acquired during each experiment session, potentially improving the throughput and success rate of awake imaging studies.

We coupled our device with WFOI and precisely correlated pupil dilation dynamics with cortical blood volume dynamics and Thy1 neuronal activity. Monitoring pupil diameter has been extensively utilized in humans and animals as a robust readout of neuronal activity patterns governing alertness and arousal.^14,24,25 We observed that the onset of stimulus-evoked pupil dilation lags the associated onset of cortical neuronal activity by $\sim 190\mathrm{ms}$. Our findings agree well with established reports that measured pupil diameter and cortical neuron activity during fluctuations in alertness or spontaneous treadmill activity. The delay between cortical neural activation and pupil response reflects the complex parasympathetic and sympathetic pathways that modulate smooth muscle constriction and relaxation in the eye.^16,26 Conversely, our observations showed that stimulus-evoked pupil dilation leads to a corresponding increase in cortical blood volume by $\sim 300\mathrm{ms}$. Similarly, Turner et al.¹⁰ observed that spontaneous pupil dilation preceded fluctuations in CBV by $\sim 1\mathrm{s}$ in alert mice and mice during the REM and NREM stages of sleep. We hypothesize that smooth muscle signaling and vascular transit time largely account for the delayed CBV response relative to pupil dilation.²⁷

Irrespective of onset time, we found that stimulus-evoked pupillary dilation requires more time to reach its peak than either cortical blood volume or neural activity. Prior studies demonstrated that increased activity from both adrenergic and cholinergic neurons precedes pupillary dilation,¹⁶ whereas cortical astrocytic activity follows pupillary dilation.²⁸ These and other findings illustrate the intricate balance of precisely timed signaling between multiple cell types in different neuromodulatory systems that contribute to regulating alertness and arousal state. Although extensive investigations are required to rigorously characterize the influence of multiple neuromodulatory circuits on spontaneous or stimulus-evoked cellular and hemodynamic activity in the brain,^19,29 our device shows promise as a simple and affordable instrument to facilitate these comprehensive studies.

One limitation of our validation experiments comes from not accounting for the influence of the partial blood volume effect [Fig. 4(b)].³⁰ For each stimulation trial, the stimulus-induced increase in blood volume reduced the peak magnitudes of neuronal GCaMP fluorescence signal (from $\sim 5\%$ to $\sim 1\%$) for later stimulus pulses. However, the partial volume effect’s impact on the calcium signal’s onset and rise time was minimal and likely did not appreciably affect our kinetic analysis. Multiple methods exist to robustly account for the reduction in fluorescence signal by hemoglobin’s absorption.^31,32 Our current findings motivate more comprehensive investigations relating neuronal, vascular, and pupil kinetics using multispectral OISI/fluorescence imaging. In future experiments, partial volume correction will be implemented on all measurements by regressing the fluorescent measurement with 530-nm reflectance measurements.^31,32

Our device shows promise for extensive and easy monitoring of dynamic behavioral events in animals during awake imaging sessions. Future developments will focus on extending the device’s capabilities to increase the sampling rate and measure more peripheral physiological sensors. Specifically, to enable robust tracking of the animal’s whisker and to potentially monitor respiration and heart rate, technical improvements will focus on improving both cameras’ optical resolution and frame rates. Prior reports emphasized the strong influences of a stimulated whisker’s initial position, angle of deflection, and neighboring whisker activity on the resultant cortical response.^33–35 To explore this further, the body camera’s magnification and frame rate will be increased up to 100 to 150 fps. To investigate the potential, yet disputable, relationship between pupil fluctuations and respiration rate,^36,37 pupil camera’s frame rate and magnification will also be increased to 100 fps or higher. Additional efforts will explore modifications to utilize our device during MRI and PET experiments. Future developments will also extend our software’s capabilities to operate with more affordable and widely utilized micro-computers.

4.

Methods