Recent advances in optical imaging have provided a highly effective multi-scale imaging tool for documenting various brain activities. In addition to the well-known advantages including capability for capturing microscopic images at the cellular or sub-cellular level, a forte of continuous monitoring of biological events with a high spatiotemporal resolution, particularly neurobiological cascades garnered high popularity among researchers of the related fields. As the image contrast, resolution, and specificity further improved with bioengineering techniques such as fluorescent proteins [1-4] and optogenetic tools [5], light-based imaging approaches are becoming increasingly higher in demand as methods for understanding neurodynamics. Additionally, the multifaceted diversity of visualization techniques such as longitudinal time-lapse imaging of the synaptic structure and function and cell-type specific imaging has also promoted the neuroscience aspect of optical imaging applications with their non-invasive nature and ease of integration with other techniques [6, 7].

In order to accomplish optical recordings of the neural, particularly central nervous system (CNS), experimental innovation is an a priori necessity to enable the visualization of neural activities through the layers of surrounding structures. In general, the brain and spinal cord are covered with three meninges: dura mater, arachnoid mater, and pia mater while the skull and spine reside beneath the outer layer-scalp, skin, and connective tissue [8-11]. Due to their intrinsic opacity, these multiple protective layers render difficulties in the optical observation of the CNS in its natural form [12-14]. Thus, an artificial window preparation technique was developed, in which the skull is thinned down, or removed and secured with a glass coverslip [15-17]. Upon introduction, initial preparation of the cranial window (CW) was performed in large animals such as cats, rabbits, and monkeys [16, 18-20]. However, recent emergence of ethical issues and advances in optical technique have led smaller animals (e.g., rodents) as more appropriate experimental subjects. Furthering the technical improvements, the latest research shows the establishment of a spinal cord chamber window (SCCW), a surgical preparation technique modified from the dorsal skinfold chamber to observe the spinal cord optically [21-23]. In this Review, the term “central nervous system (CNS) window” will be used to refer to both the cranial window (CW) and the spinal cord chamber window (SCCW). Recently, there is a considerable demand for integrating the optical neuroimaging system with other recording modalities in conjunction with simultaneous behavioral assessments. In this regard, this Review addresses various types of the cranial and spinal windows used in rodent model studies and associated behavior tests in combination with experimental setups of multiple optical neuroimaging techniques and other recording modalities.

Compared with recent reviews related to the cranial window [24, 25], this Review introduces various structural and functional in vivo imaging technologies utilizing the cranial and spinal window and the various types of cranial and spinal windows in rodent models by categorizing them into ROIs, surgical preparation approaches, and window material with extensive Tables. In addition, we describe practical considerations when using the cranial and spinal window with rodent models in conjunction with other neuroscience methodologies or behavior tests with various examples.

Optical imaging approaches have been used to investigate structural and functional brain connections in rodents. For example, the optical neuroimaging methods such as two- or multi-photon excitation (2PE or MPE) imaging [26-29], calcium-sensitive dye imaging (calcium imaging, CaSDI) [30, 31], voltage-sensitive dye imaging (VSDI) [32-34], laser speckle contrast imaging (LSCI) [35-38], optical intrinsic signal imaging (OISI) [35, 39-42], and ultrasound imaging (USI) combined with light [43, 44] have been widely used. These techniques enabled the investigation of vascular and cellular structures, neuronal activation, blood flow, blood pressure, and oxygen saturation. In the following section, we introduce the characteristics of neuroimaging technologies.

2PE imaging became popular in neuroscience as a nonlinear laser-scanning fluorescence microscopy technique for deep-tissue cellular imaging of 500 μm~1 mm in depth and sub-micrometer spatial resolution. The two-photon excitation wavelength lies in the near-infrared range, nearly twice longer than the usual wavelengths for confocal or epifluorescence excitation. 2PE has been utilized for structural fluorescence imaging or CaSDI and VSDI for neuronal activity in wide FOV [45-48]. Recent advancement in MPE, such as 3PE, allows even deeper functional imaging beneath 1 mm depth [49].

OISI has been used to infer neural activity based on cortical reflectance change originating from hemodynamic response. OISI is used for label-free imaging of cortical functional architecture and local microcirculation with ~100 μm spatial and 1~2 sec temporal resolutions. In previous research, OISI was applied to analyze the functional connectivity in large-area of the brain cortex and even onto the mouse’s exposed skull [50-52]. The oxy-/deoxy-hemoglobin concentration for specific brain regions can be estimated with multi-color light illumination [53-55].

A laser speckle is a random interference pattern from the scattering of coherent light within tissue. The fluctuations of speckle pattern caused by dynamic movements of scatters (i.e., red blood cells) within the living tissue enable a two-dimensional blood flow or tissue perfusion, called LSCI. LSCI provides a simple and powerful tool for full-field semi-quantitative functional blood flow. The spatial resolution (10 μm) and temporal resolution (10 msec to 10 sec) of LSCI can be customized depending on the specific application. Because of the limited light penetration depth, LSCI only provides superficial blood flow mapping [56, 57].

CaSDI and VSDI utilize special dyes sensitive to neuronal activity among optical neuronal imaging techniques. These approaches use dyes that generate fluorescence in response to an action potential to monitor neuronal activity in cellular or sub-cellular measurements. CaSDI and VSDI provide spatial and temporal resolutions of the μm and msec ranges, respectively. Using these methods, it is possible to simultaneously measure the neuronal activity of multiple populations within a field-of-view (FOV) [58, 59].

From the traditional optical imaging method with superficial images limited to the cortical area of an animal under anesthesia, we are witnessing deep-tissue imaging into the CNS below the cortex of awake animals performing behavioral tasks. This advancement has been feasible with diverse cranial and spinal window models, which follow below.

Region of interest (ROI)-based classification

The type of cranial and spinal window can be decided based on the region of interest (ROI) to be examined with the microscope. Here we describe various types of cranial windows such as the olfactory bulb [18, 60-64], somatosensory cortex [34, 35, 65, 66], visual cortex [31, 44, 67], hippocampus [68-70], cerebellum [71, 72], medial entorhinal cortex [73], and the spinal cord chamber window (SCCW) [21-23]. Fig. 1 shows the schematic diagrams of cranial and spinal windows of four representative ROIs: olfactory bulb, cerebral cortex, cerebellar cortex, and spinal cord.

Olfactory bulb window

The olfactory bulb (OB) in mammals is the initial station of the olfactory nerve pathway in the CNS [74]. The olfactory system is the only sensory organ that carries peripheral information directly to the cortex bypassing the thalamus [75]. As a result, the OB shows the combined function of the peripheral nervous system (PNS) and the thalamus within the CNS [76]. Both thinned-skull or open-skull windows can be implanted for OB imaging for studies on olfactory information processing. However, OB window preparation is challenging due to the limited frontal bone size (2.0×3.5 mm) and the location is between the eyes (Fig. 1a). Although the thinned-skull OB window enables long-term imaging, the imaging depth is further restricted due to the remaining bone and surface irregularity. In contrast, the open-skull OB window's small size necessitates repeated surgery owing to bone regrowth limiting its use only for short-term imaging. Thus, a large (~3 mm) open-skull window is suitable for long-term longitudinal imaging [62].

Cerebral cortex window

The cerebral cortex is the outermost layer of the cerebrum. The entire dorsal cerebral cortex including motor and sensory areas is optically accessible after removing outer skull and scalp with transcranial window technique. The somatosensory area and barrel cortex have been common targets for studying sensory-evoked optical responses [35, 41, 65, 77]. In stroke research, optical imaging of functional changes in the cerebral cortex and recovery utilizing well-established animal modeling approaches such as middle cerebral artery occlusion (MCAO) [28, 41, 78] or photothrombosis [37, 65, 66]. Typical cortical cranial window size varies from 2 mm up to 8 mm in diameter depending on the region of interest (ROI). Examples include synaptic and dendritic changes in retrosplenial cortex (posterior cortical area) with small-sized CW (~3 mm in diameter) by 2PE imaging [79]; OIS changes in small targeted region (visual cortex, ~4 mm) using intact skull cranial window model [80]; cerebral blood flow changes in somatosensory area (~6 mm in diameter) by LSCI and OISI [35]; large area (7 mm×8 mm) cortical neuronal connectivity analysis with VSDI [34]; and whole skull optical clearing window (SOWC) for microvessels and synaptic resolution [81].

Windows for deep brain region

Conventional transcranial windows only have a direct access to the cortical areas leaving subcortical or deep brain regions hardly accessible with optical techniques. As an example, the hippocampus, a prominent target for early detection of Alzheimer's disease (AD), is located below the cortex in rodents. Special type of cranial windows were constructed utilizing a gradient-index (GRIN) lens [68] or a triangular prism [73] for longitudinal observation of the deep brain target such as the hippocampus or cerebellum. These examples show that for longitudinal in vivo imaging, cranial window preparation is a valuable technique for observing disease progression in situ.

Spinal cord chamber window (SCCW)

The spinal cord chamber window is another cranial and spinal window type to achieve long-term imaging of the spinal cord. Spinal cord serves as an intermediate channel for nerve signals between the cerebrum and the peripheral nerves. While the cerebrum has been a predominant research subject in neuroscience, research on the spinal cord, which bridges the brain and peripheral nerve, has become increasingly important with the increased focus on peripheral nerve disorders. Previously, imaging for lumbar [37] or thoracic [36] has been majorly shown, which is straightforward to access and significant in size even for a small mouse, but lately, cervical imaging has also been tried [38]. SCCW was used for long-term monitoring of brain cancer metastases through the blood-brain barrier (BBB) [37, 38] or the response of astrocytes and microglias to ischemic brain injury [36]. Spinal cord with SCCW was observed after spinal injury through microglia inflammation and heterogenous dieback of axon stumps [69, 70]. Studies on neurorehabilitation or functional recovery of the spinal cord injury model are expected to be active when the SCCW model is combined with neuromodulation approaches.

Types of cranial and spinal windows

The early cranial and spinal window for brain imaging in small animals was a one-off procedure that involved drilling the skull down till the brain was exposed. For long-term in vivo imaging, the next generation of cranial and spinal windows covered the brain parenchyma with glass coverslip over the open-skull or thinned skull. However, these window types require further modification, including the retractable type or flexible polymer-based windows, which have been adopted in conjunction with other neuroscience techniques such as brain stimulation or electrophysiological recording [30, 39, 40, 77, 80, 82-84]. With advances in MEMS supported by nanotechnology, flexible transparent microelectrode will allow a range of previously unfeasible experiments due to the physical barrier from glass windows. Examples include functional brain mapping in response to optogenetic or ultrasound brain stimulation with freely moving mice during behavior tests. Fig. 2 shows schematic diagrams for typical types of cranial windows. We illustrate the examples of cranial and spinal windows for optical neuroimaging in vivo in Table 1, comparing animal subjects, cranial and spinal window types, optical imaging methods, and test apparatus from references.

Open-skull preparation

The open-skull preparation was used to monitor the blood-oxygen-level-dependent (BOLD) or optical intrinsic signal from the sensory, barrel, or visual cortex evoked by specific stimulation for observation under anesthesia [33]. Although this preparation was straightforward and yielded clear optical resolution, it came with the risk of inflammation during and after the imaging. This preparation method was the early form for optically observing the brain, and it has been used on small animals like mice and rats as well as medium and large mammals like cats and monkeys [85, 86]. A more recent study demonstrates a modified open-skull preparation method for ultrasound and photoacoustic mesoscopic imaging, covering the exposed cerebral cortex by suturing the scalp [43]. In the recent study, the exposed brain was coated with the hydrophilized 130 nm–thick-nanosheet of the polyethylene-oxide-coated CYTOP (PEO-CYTOP), which enabled in vivo deep brain imaging in a wide FOV, resulting in good adhesiveness to the brain surface. It also aided in the long-term control of surface bleeding and inflammation [87, 88].

Sealed cranial windows (closed cranial window)

Meanwhile, the sealed cranial window with glass coverslip [16, 19, 89], agarose [90], or silicon oil [65, 90] has been applied to protect the covering of the exposed cortical surface. Compared to the previously mentioned open skull preparation, this procedure allowed for mid or long-term observation [20, 91]. It has been used to observe the metastasis of brain tumors [92-96] and changes in brain tissue caused by a focal stroke [28, 41]. Anesthesia conditions, body temperature, and breathing should be managed for long-term and longitudinal observation. Furthermore, the choice of the glass window depends on the animal size and the duration of the experiment. Sometimes the dura mater is removed for securing image resolution with potential impairment of microenvironment physiology. This sealed window method allows mid to long-term observation, although the imaging duration can be limited due to the skull regrowth phenomenon. In addition, the imaging quality may be compromised by astrocytic gliosis and the undesired inflammatory response with microglial activation [97].

Open-close/retractable/accessible window

On the other hand, hybrid cranial windows have addressed the shortcomings of both open-skull preparation and sealed cranial windows while maximizing the benefits. The sliding-top cranial window has a chamber, thread, and slipover cap to obtain stable and simultaneous brain imaging and electrical recording. This method eliminated two significant problems that are cardiac and respiration-induced image artifacts and brain infection caused by contamination [98]. Similarly, retractable plugs on the glass coverslip were used to assess the recording electrode for the brain surface [30, 67, 84]. Meanwhile, a flexible and transparent polymer-based cranial window enabled the recording electrode or injection needle to penetrate the window surface. Other examples of hybrid windows are artificial dura [55] and accessible cranial windows [19]. Although the open-close, retractable, and accessible windows are relatively safe, there still exists a risk of infection and contamination by external assessment.

Thinned-skull preparation

The thinned-skull preparation, which involves grinding the skull until the skull becomes thin enough to be optically transparent, was invented to mitigate the shortcomings of the open-skull method, such as disrupting the local environment and potential contamination. By reporting the optical view while preventing direct exposure of the cortical surface through the skull thinning process, this technique also addresses the issues of the sealed cranial window. Although imaging resolution is limited due to the small residual skull layer and dura mater, image quality degradation caused by skull growth and gliosis can be minimized, allowing for more scientifically controlled observation [29, 78, 97, 99-101].

Skull optical clearing window (SOCW)

The optical tissue clearing window technique applies the optical clearing solution (SOCS) on the skull surface instead of the craniotomy [81, 102]. The SOCS included collagenase, EDTA disodium, and glycerol. The skull optical clearing window (SOCW) is easy to handle, and the SOCW technique is safe and reliable as no apparent inflammatory responses were reported associated with the SOCS. As a result, SOCW is well suited to studying microglia that are highly sensitive to the microenvironment. Although SOCS can make the skull almost entirely transparent, it works only for a short period. Thus, repeatedly SOCS may be necessary. Through the SOCW preparation, the minimum diameter of microvessels was measured as 14.4±0.8 µm closed to that of the exposed microvessels as 12.8±0.9 µm [102]. In addition, the imaging depth was achieved down to 250 μm below the pial surface [81]. Therefore, the SOCW could be an alternative approach for extensive thinning or removing the skull.

Material-based window classification

The previous classification was based on the different types of cranial and spinal windows; however, the window model can also be categorized based on the materials used. The original cranial window was prepared by placing a cover glass over the exposed brain surface. In contrast, the nanocrystalline material called yttria-stabilized zirconia (YSZ) was used to reinforce the glass's mechanical vulnerability. Compared to glass, YSZ has a higher hardness, and its biocompatibility is well established. The transparency is lower than that of glass when used for optical coherence tomography (OCT) imaging, but it ensures superior image quality compared to the native skull [103-107]. On the contrary, there are cases where transparent and flexible polydimethylsiloxane (PDMS) has been used to cover the cerebral cortex. Also, PDMS can be employed as the substrate of electrode fabrication. Electrode patterns were transcribed using silicon, graphite, graphene, or carbon nanotubes (CNTs) to electrically stimulate the cerebral cortex and optically record responses [39, 82, 83, 108, 109]. Furthermore, local field potential (LFP) recording was attempted with optical imaging with flexible and penetrable PDMS window [40, 77]. Meanwhile, a 35 μm collagen membrane sheet was used as an artificial dura substitute with high biocompatibility and semi-permeability to allow VSD dye to pass. Although this chemically and optically transparent collagen membrane made VSD staining easier, its utility for longitudinal investigation needs to be examined [110]. These examples highlight the proper choice of window materials based on the study purpose and observation method.

Various applications of the cranial and spinal window based on ROI, type, and material are illustrated. From the original single observation of a medium-large animal for an easy-to-access brain region, the cranial and spinal window has evolved for a mid and long-term observation of target CNS regions with small animal disease models. The practical issues will be addressed in the following section.

Combining with other neuroscience methodology

The cranial window preparation for the brain in vivo imaging has also been used to observe changes in the cerebral cortex combined with other methodologies. First of all, cranial windows have been used to see the response to brain stimulations such as optogenetic [6, 34, 111, 112], electrical [39], or ultrasound stimulation methods [113]. The window material is chosen according to brain stimulation type, considering permeability and interference. Moreover, optical brain imaging was performed simultaneously with neuroelectrical recordings such as patch-clamp [40, 80, 84], electrocorticography (ECoG) [114], and electroencephalography (EEG) [33]. As the rodents’ skull size ranges from mm to cm-scale, proper spatial arrangement is required [115, 116]. Microelectrodes have also been developed by microelectromechanical systems (MEMS) technology to record simultaneously optical and electrical neural activity [108, 114, 117]. On the other hand, chemicals such as a dye can facilitate structural observation or acquiring functional brain signals including blood flow responses following drug administration [67, 84]. Furthermore, multiple methods were combined for the observation of simultaneous electrical and optical signals from the brain [39, 40, 71, 84, 91]. Table 2 shows the advanced cranial and spinal windows to observe rodent brain combined with other neuroscience methodologies. The practical examples of cranial and spinal windows utilized for optical neuroimaging combined with other neuroscience methodologies are described in Fig. 3.

Behavioral tests with awake animal models

Many brain imaging experiments through the cranial window model have been performed with animals under anesthesia. However, in vivo imaging with an awake animal undergoing a behavioral task is gaining popularity to avoid the anesthetic effect. Furthermore, freely behaving animals have been studied with EEG recording or equipped with a miniaturized microscope; however, these methods have limitations in single-cell level registration over large-area for brain connectivity. The optical window-based in vivo imaging system can be combined with a special apparatus during behavioral testing. For example, various treadmills have been used for studying exercise-induced behavioral recovery and neuroplasticity changes, such as straight treadmills [118-121] and ball-type treadmills [30, 73, 79, 122]. Other device types include a flat-floored air-lifted platform [123] and virtual reality (VR) environments [73, 118, 124], enabling the tracking of mouse locomotor activity while imaging. In addition, learning and memory function has been studied with spatial perception based on hippocampal place cell activity. Meanwhile, an automated home-cage mesoscopic functional imaging [125] showed visually-evoked cortical maps from multiple mice with a longitudinal study over three months.

The previously described apparatuses were not typical rodent behavioral tests such as elevated maze or conditioned cages have been used for functional recovery with well-defined rodent disease models and behavioral test paradigm. However, one limitation of such studies was the behavioral test and cellular recording were performed separately, and not integratable. Thus, optical neuroimaging techniques integrated with other modalities along with behavioral experiments, allows powerful integration of results and direct explanation the functional change and recovery mechanisms at the cellular resolution with a disease model. We are observing the expansion of optical neuroimaging apparatuses together with conventional behavioral assays with specific disease model, such as stroke [126] or Alzheimer’s disease [127].

Upon construction of the cranial and spinal window, the cover material’s optical, chemical, and physical characteristics must be considered in accordance with the study’s purpose and objectives. In order to optimize the quality of target images, region of interest (ROI), field of view (FOV), and tissue depth should be specified, from which the optical imaging parameters and experimental setting can be selected. In addition to the camera sensor area, the number of pixels, and window glass material, including refractive index, and the working distance of the objective lens needs to be determined prior to the window-making procedure. Moreover, since it is applied to live animals, selecting the window's physical strength and an appropriate weight is also necessary for the long-term preservation of the window and subsequent observation. As much as physical and chemical properties, the factors concerning animal species and conditions should also be considered. In other words, materials and shapes of the cranial window need to be selected based on the species, age, size, weight, life expectancy, the type of animal disease model, location of ROI, and size of FOV.

As for species-dependent differences, the rat and mouse have a considerable difference in life expectancy and physical strength. Therefore, such phenotypic differences should be taken into account, as the window strength and durability for rats and the whole window weight for mice may become significant factors. For example, a flexible polymer-based window may require additional protection depending on the observation period and animal size. Although highly effective, typical optical imaging methods using the cranial and spinal window exhibit both benefits and limitations. While the cranial and spinal window enables efficient 2D wide-field imaging, information collected from optical neuroimaging often suffers from the constraint of superficial information. For studies of 3D functional neurodynamics, simultaneous acquisition of neuroelectric signals using electrophysiology in the targeted brain areas at variable depths can be adopted [39, 42, 63, 108, 117].

Combined electrophysiological recording and optical imaging have been attempted to reveal synaptic responses and cortical activity with high spatio-temporal resolution in large populations of cortical neurons. This multimodal approach enables electro-optic mapping of cellular neuronal activity and ultimately correlates circuit-level behavior and the connectivity among different brain regions [40, 73, 79, 117]. Of cautioning note, the insertion of electrodes would induce unavoidable damage to the tissue, and the imaging FOV would also be restricted by the electrode. Single electrode insertion to the exposed brain could be sufficient in some cases; [63, 67, 89] however, specialized design and devices (e.g., micro electric mechanical system: MEMS) and/or implementation of multi-channel recording setup may be necessary for effective assessment of the 3D neural structures [39, 42, 108, 117].

When performing imaging using a cranial and spinal window with a behaving animal, spatial arrangements of the microscope stage are crucial considering FOV, the working distance of the objective lens, anesthesia setting, and proper monitoring of the animal under experiment. In conjunction with a recording or behavioral test, sequential procedures with involved devices should also be considered in multimodal imaging. Progressive habituation protocols are necessary for awaked rodent imaging with head-fixed circumstances, to adapt the animal to the restrained condition [31, 123, 125]. Developing a behavioral test scenario is necessary while considering spatial allowance, temporal resolution limitation, and the maximum measurement duration per single imaging run. A virtual reality (VR) environment can be implemented for a freely-moving-like environment to extend to the in vivo optical imaging further using the cranial and spinal window.

Optical in vivo imaging using cranial and spinal windows can be compared to imaging with miniaturized microscopes (miniscopes). In general, using the cranial and spinal window provides more expandability to combine with other techniques and a larger optical FOV [128, 129]. Furthermore, multimodal imaging is easier for simultaneous acquisition of various data (neuronal activity, cerebral hemodynamics, structural imaging). On the other hand, miniscopes have evolved to integrate electrophysiology and calcium imaging in free-moving animals. Also, the miniscope approach is suitable for entirely free-moving or social interaction tests. Thus, the proper choice of approach depends on these considerations.

Over the past decade, traditional cranial and spinal window-based optical in vivo imaging technology has slowly paved the way for high-resolution neuroimaging of living animals. Advances in optics, MEMS technology, and biocompatible materials allow versatile cranial and spinal window technology, which can be a paradigm change in neuroscience. This Review introduced various cranial and spinal window types used in the optical neuroimaging field with rodent window models for anesthetized and awake animals. The combination of window-based optical imaging with other imaging modalities such as ultrasound and MRI can provide further depth and dynamics to the study of the nervous system, which can be furthered with electrophysiology techniques.

The evolution of cranial and spinal windows has several important implications for future neuroscience practices. The future cranial and spinal window will find applications in functional brain mapping and reveal mechanisms of brain dysfunction at the cellular resolution with various disease animal models. For example, one could envisage using the cranial and spinal window with a behaving animal for longitudinal investigation on brain connectivity during optogenetic stimulation, neural plasticity changes, or testing new drug candidates related to various neurological diseases such as Alzheimer’s disease, Parkinson’s disease, stroke, epilepsy, and chronic pain. With ongoing progression in cranial and spinal window preparation, we expect transformable, multi-functional, or flexible cranial and spinal window to become commonplace in neuroscience laboratories alongside advances in neuroimaging modalities.

The work was supported by The GIST Research Institute (GRI) research collaboration grant funded by GIST in 2022 and the 2022 Joint Research Project of Institutes of Science and Technology, a grant from the National Research Foundation of Korea (N.R.F.) funded by the Korean government (MEST) (NRF-2019R1A2C2086003), and the Korea Medical Device Development Fund grant funded by the Korea government, the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety (Project Number: 1711138096, KMDF_PR_ 220200901_0076), and the Brain Pool program funded by the Ministry of Science and ICT through the NRF of Korea (2021H1D3A2A01099707).

The cranial and spinal windows for optical neuroimaging in vivo

The advanced cranial and spinal windows applying to optical neuroimaging combined with other neurological research methods