1.1.

Need for Greater Precision in Human Cortical Mapping

Precise information about structural and functional anatomy is essential in performing maximally effective and safe neurosurgical interventions. Functional brain mapping is necessary for identifying and preserving essential cortex during these interventions. A number of functional brain mapping methods have been adapted for patient care, including electrical stimulation mapping (ESM), electrophysiological recordings, and functional magnetic resonance imaging (fMRI). Other newer techniques are establishing their clinical role in functional brain mapping, such as functional near-infrared spectroscopy (fNIRS). Each has its own strengths and weaknesses. ESM is used to identify essential language cortex; however, it is unable to delineate these sites or to identify secondary language sites, and requires direct stimulation of cortex, raising concerns about triggering epileptic activity. Noninvasive electrophysiological recordings like electroencephalography (EEG) can be used to detect epileptic activity and with the recordings of sensory evoked potentials (EPs) allow for the identification of somatosensory and visual cortex. These methods offer high temporal resolution; however, intraoperatively they are recorded from grid electrodes with poor (typically 1 cm) spatial resolution. FMRI provides functional identification of targeted cortical areas. Importantly, it is noninvasive and can investigate the deep structures that ISOI cannot. As such, fMRI will continue to have an important role in preoperative planning and scientific research. However, alignment of preoperative images with intraoperative views is complicated by brain shift and swelling. fNIRS is a noninvasive optical brain imaging method that detects cortical activation, and though it has excellent temporal resolution (1 ms), it has poor (1 cm) spatial resolution. A functional mapping technique that offers higher spatial precision and is applied intraoperatively would provide more accurate information and lead to more informed intraoperative decision-making and greater patient benefit. ISOI has the advantage of potential intraoperative use allowing direct visualization of functional areas and improved surgical intervention.

ISOI is a functional imaging technique that uses a charge-coupled device (CCD) camera to indirectly measure neuronal activity by detecting hemodynamic changes related to neurovascular coupling. It requires craniotomy and durotomy, in other words, it requires direct exposure of the cortex to obtain its high spatial resolution functional maps of activity. It has been used extensively in animal research^1–5 and shows promise for clinical applications.⁶ ISOI offers high spatial ($\sim 100\mu \mathrm{m}$) and temporal (100 ms) resolution and can be tuned to detect different physiological elements such as blood flow and oxygen consumption.⁷ This methodology has led to breakthroughs in our understanding of the functional organization, physiology, and pathophysiology of the brain.⁷ It has been used extensively in animals to study auditory,^8–10 somatosensory,^11–13 visual,^14–18 and motor¹⁹ cortices in animals. The enormous advances in our understanding of cortical function resulting from ISOI studies led to concerted efforts to develop ISOI for clinical use.

1.2.

Previous Intraoperative Studies

In 1992, using intraoperative ISOI, Haglund et al.²⁰ first described their groundbreaking findings in human language cortex. This led to a surge of intraoperative ISOI studies in human language^21,22 and somatosensory cortex (SI)^23–26 as well as studies of cortical hemodynamics,^27–30 cortical response to electrical stimulation,³¹ epileptiform activity,^20,32,33 and lesion delineation.^22,34,35 Within a decade and a half, its capacity as a precise functional mapping tool in humans was established. The next, and perhaps the most important, step was to establish its use as a clinical tool to improve patient care.

Despite the number of demonstrated applications, human ISOI failed to spread from the few centers where this method had been pioneered. A number of factors have contributed to this. First, until such technology becomes routine, a strong collaborative effort between clinicians and researchers is needed. This requires both researchers experienced in ISOI and clinicians willing to devote effort to its development in the OR. Second, many of the early experiment designs, which were adapted from animal studies, utilized episodic stimulus presentation designs that required many trials and thus long ($>30\min$) experiment times, making it less compatible for the human OR. The early studies thus had only a small number of subjects, resulting in insufficient pilot data for clinical trials. Third, partly due to the limited number of patient studies, a standard ISOI system has yet to be specifically designed for clinical use; consequently, optical imaging setups utilized in the laboratory have been adapted for the OR (e.g., ad hoc attachment of the ISOI camera to an operative microscope), introducing another source of variability. Fourth, the operative setting poses additional uncontrolled conditions that lead to unfavorable signal to noise ratios. Such conditions include variability in ambient light due to the screens to monitor patient vital signs as well as large patient/camera vibrations that may be caused by the presence of clinical tools such as sequential compressive devices used during operative procedures. Thus, as of today, there is no standard ISOI method available for intraoperative use.

Recently, there has been a resurgence in efforts to improve and standardize the application of ISOI in the OR. Several groups have addressed these limitations by implementing better experimental setups³⁶ and new methods of data acquisition and analysis^37,38 that shorten the time needed to conduct experiments. Such improvements have led to larger patient studies, and the largest study to date had 41 patients, Sobottka et al.,³⁹ and increased confidence in this approach. Moreover, the field of ISOI in animals continues to evolve; its use in conjunction with new focal stimulation methodologies such as optogenetics,⁴⁰ microelectrical stimulation,⁴¹ and laser stimulation⁴² ushers in a new age of functional tract tracing. Thus, as the field develops, its potential for use in humans also reaches exciting new heights and points toward more widespread clinical and scientific impact.

2.1.

Intrinsic Signal

The “intrinsic signal” referred to in this review is the early hemodynamic responses in brain tissue related to neuronal activity.^1–6,26 In brief, these include local changes in the oxy- to deoxy-hemoglobin concentrations as well as a local increase in cerebral blood flow. These hemodynamic responses lead to changes in the absorption spectrum of the surrounding tissue that can be detected by a sensitive camera. These changes in light reflectance are small, on the order of a 1% signal change, with timecourses that peak a few seconds after the onset of neural activity. These signals are thought to correspond to the so-called “initial dip” in fMRI studies and occur prior to the blood oxygenation level-dependent (BOLD) signal. There is typically a several (5 to 10) second refractory period before the reflectance change returns to baseline.

There are many components in tissue that lead to reflectance changes. Different components are better detected at different wavelengths of illuminant. Three of these components are: (1) total hemoglobin (HbT), used to infer cerebral blood volume (CBV), is the dominant source of signal with green-yellow light ($\sim 500$ to 599 nm), (2) deoxyhemoglobin (HbR) is the dominant signal source with red light ($\sim 600$ to 699 nm), and (3) changes in cellular swelling are the dominant signal source in the near-infrared spectrum ($\sim 700$ to 800 nm).⁶ These hemodynamic changes are induced by both neuronal spike activity as well as subthreshhold synaptic activity and represent summed activity of a population of neurons.

2.2.

Basics of Intrinsic Signal Optical Imaging

In practice, changes in light reflectance are detected by a CCD camera positioned over cerebral cortex. The relative change in reflectance ($\mathrm{d}R/R$) is measured relative to a baseline control level of reflectance. Because the signal size of intrinsic signals is small, the sensitivity of the imaging camera is a critical variable. Current turnkey ISOI systems (e.g., Optical Imaging, Ltd. Imager 3001/S, SciMedia MiCAM02, Redshirt Imaging LLC NeuroPlex system with a 14-bit NeuroCCD-SM256 camera) come with high frame rate, high pixel count, and low noise 12- to 14-bit cameras. Other requirements for ISOI include selecting the camera optics that determines focal depth and the field of view. For human ISOI, the typical optics used are those provided by the operating microscope upon which the camera is attached via a side port. Imaging wise, this approach might not be ideal due to light losses, the increased depth of field, and the relative motion noise due to vibrations in the OR and between the brain and operating microscope. Illumination is typically provided by operating microscope or halogen light source which is subsequently band-pass filtered to select a wavelength to collect the desired hemodynamic responses. As brain motion noise due to pulsation and respiration is large in human ISOI, the brain is stabilized with a sterile glass footplate (but see Ref. 39). To further increase the signal to noise ratio, multiple (e.g., 6 to 20) trial stimulus presentations are collected, image alignment algorithms are used to correct for any apparent brain motion, and for image analysis spatial and/or temporal filtering is used. There currently is no standard human ISOI approach that has been proven to provide the sensitivity observed in animal studies. Recent advances in imaging cameras and analysis methods may make human ISOI more feasible.⁴³ Toward that end, Sobottka et al.³⁹ imaged the largest sample of patients ($N=41$) using three different camera setups and two analysis approaches to increase the signal to noise of the ISOI data. By increasing the signal-to-noise ratio, these methods can provide functional maps with spatial and temporal resolution on the order of tens of micrometers and tens of milliseconds, respectively.^1–6 As cerebral cortex in animals and in humans is characterized by functional domains a few hundred microns in size, ISOI has been highly instrumental in revealing the characteristic functional organizations of different cortices.

2.3.

Intrinsic Signal in Human Studies

Toga et al.⁴⁴ published one of the first studies of intraoperative ISOI to look at the temporal and spatial evolution of optical signals. In seven patients undergoing tumor resection, ISOI maps in SI were directly compared with somatosensory evoked potentials (SEPs). Consistent with studies in animals, observable signals appeared within 1 to 2 s, peaked at 3 s, and disappeared by 9 s. In response to local stimulation of skin, these signals colocalized with the largest SEPs in sensorimotor areas and were topographically appropriate. Cannestra et al.²⁸ provided a subsequent report on the time course of the signal with similar findings. These studies introduced the feasibility of using ISOI as an intraoperative mapping method.

A further study on the topographic specificity of ISOI underscored the spatiotemporal specificity of the intrinsic signal. In the OR environment, tactile or electrocutaneous stimuli used for mapping are suprathreshhold, meaning they can induce responses not only from the stimulated skin site but can also activate adjacent sites. Consequently, the responsive cortical area can then be larger and appear less topographically specific. Cannestra et al.²⁷ examined this question in a study of nine patients where separate ISOI maps were obtained when different skin areas were stimulated individually. Maps of ISOI peak responses showed unique, nonoverlapping activation patterns; whereas areal responses showed regions of overlap. Significant specificity in regard to timing of signal onset was observed during the early phase of the optical response (500 to 1750 ms) while later responses were nonspecific. Areas of peak ISOI signal corresponded to regions identified with SEPs.

The timecourse of the intrinsic signal is also relevant for stimulus timing considerations. Due to the relatively slow timecourse of the hemodynamic responses, on the order of seconds, repetitive and continuous stimulation paradigms can result in reduction and loss of responses to subsequent stimulus presentations. Thus, unlike cortical EPs which have a higher temporal resolution, hemodynamic responses are relatively slow.³⁰ Thus, for event related stimulation paradigms, it is better to incorporate sufficient interstimulus intervals (e.g., 6 to 10 s) to permit recovery to baseline prior to the next stimulus presentation.

It is important to note that the temporal and spatial specificity of optical signals differs from that of other functional mapping methods. In one study,²⁹ eight patients underwent mapping with fMRI, SEPs, and ISOI. Each modality provided unique spatial and temporal profiles. While SEPs were the most temporally specific, SEPs and ISOI signals were detected with similar spatial distributions. When fMRI and ISOI were compared, in colocalized activation regions, the temporal profile of fMRI showed an initial decrease (the initial dip) consistent with the intrinsic signal.⁴⁵ In a second study,⁴⁶ ISOI and fMRI data were collected in five patients. This study found that in response to somatosensory stimulation, the late positive phase of the optical signal correlated in time and space with the BOLD fMRI signal, though not precisely. The mismatch between BOLD and ISOI mapping was likely largely due to the different vascular sources of the hemodynamic signals. Early ISOI signals derive from microcapillaries and are a consequence of a decrease in oxyhemoglobin after neuronal activation. The BOLD signal, on the other hand, is dominated by large vessels and is related to the local increase in oxyhemoglobin that over-compensates for the initial decrease in blood oxygen. This underscores the greater spatial precision of ISOI compared to fMRI.³

4.1.

Somatosensory Cortex

SI contains a highly complex network of neurons that work together to synthesize information about position, texture, pressure, temperature, etc. of an object in contact with the skin. Precise information about the location of SI and organization allows for its preservation by clinicians and in the future may serve as a blueprint for interfaces integrating sensory information from prosthetic devices. Many of the early studies using ISOI in humans looked at the functional organization of SI with median nerve stimulation, digit stimulation, and trigeminal nerve/facial stimulation. In five patients Sato et al.^23,24 and in one patient Shoham et al.²⁶ found electrical stimulation of the median nerve resulted in hemodynamic changes in the median nerve territory of SI as confirmed with electrocorticography (ECOG). In a study designed to investigate the specificity and reliability of ISOI, Sato et al.^23,24 reported on his observations from stimulating individual fingers. Figure 2 exemplifies these findings by showing the reliability of the response to digit simulation over multiple trials.²³ Electrocutanous stimulation of digit I and V revealed separate cortical representations along the crown of the postcentral gyrus near the central sulcus. This was one of the first studies to confirm the hand homunculus in humans with functional imaging. Furthermore, in four of the cases, the following pattern was observed: stimulation of D1 and D5 introduced neural responses in two different areas. The first was located near the central sulcus, where D1 and D5 were separately represented. The second response area was near the postcentral sulcus, where D1 and D5 had overlapping representations. In addition to showing that intraoperative, ISOI is a highly effective imaging technique to monitor cortical activity during neurosurgery, with the high spatial resolution of ISOI this was the first study to functionally distinguish Brodmann subdivisions of SI within the same gyrus in humans. In a subsequent study by the group,²⁴ stimulation of digits, D1 to D5, was conducted and SI was imaged in six patients yielding confirmatory findings.

Fig. 2

Intrinsic optical responses induced by digit I and V stimulation. (a) Intrinsic optical images recorded from the left SI of a 63-year old patient (case 8 in Table 1 of Ref. 23). Right digits I and V were individually stimulated and the detected optical responses are illustrated by pseudocolor images. The black square in the right panel represents the detected area, and the yellow line indicates the central sulcus. (b) Traces of the optical response areas induced by three repetitive trials. The traces are superimposed for each trial. (c) Equivalent current dipoles (ECDs) are superimposed on a 3-D MR image. The red closed circle is the ECD to digit I stimulation, and the green closed circle is the ECD to digit V stimulation.²³ Figure and legend reproduced from Ref. 23, with permission from Oxford University Press. Imaging methods: an Imager 2001 system (Optical Imaging Ltd.) was used to collect the images. The camera was fitted onto the operating microscope. Xenon light generated by the surgical microscope illuminated cortex that was filtered at 605 nm. A sterile glass footplate was used minimize brain movement artifacts. Stimulation of digits I and V consisted of 10 mA DC electrocutaneous pulses ($n=10$) presented at 5 Hz for 2 s.

Sato et al.^23,24 also studied ISOI activation in the facial distribution of SI. In two patients, stimulation of the supraorbital (a branch of V1) and mental (a branch of V3) nerves revealed an activation pattern of overlapping and nonoverlapping areas likely indicating different subdivisions within the facial representation of SI (the nonoverlapping representation occurring in Brodmann area 3b and the overlapping representation occurring in Brodmann area 1). In another patient, Schwartz et al.²⁵ identified the face region with ECOG and then imaged this area during cutaneous stimulation. The area of cortical activation during stimulation of the upper face was located medial to that of the lower face. They also found significant overlap ($\sim 30\%$) between these activation areas.

These findings build on knowledge of SI organization derived from animal studies and human studies using other mapping modalities. For example, recent evidence indicates that human digit representation in SI is similar to the digit representation of nonhuman primates, with each digit having multiple areas of representations; in some areas (e.g., area 3b), representations of different digits are relatively distinct, while in others they appear to have more overlap.⁴⁷

Intraoperative ISOI has the potential to provide clinicians with highly detailed somatotopic maps of S1 that, because they are collected intraoperatively, are not subject to brain shift and other changes to cerebral architecture introduced by the operative environment as opposed to preoperative fMRI. These spatially specific, intraoperative maps could someday serve as the interface for prosthetic devices carrying sensory information to the brain.

4.2.

Visual Cortex

ISOI has been used extensively to study animal visual cortex. However, as the occipital lobe cortex is rarely exposed during neurosurgical procedures, there are very few studies of ISOI in human visual cortex. Sobottka et al.⁴⁸ optically imaged the occipital lobe while visually stimulating both eyes with strobe-light flashes in one patient. Visual evoked potentials (VEPs) were recorded at four different locations across the occipital lobe for comparison. The locations of ISOI responses were consistent with the electrophysiological VEP findings and corresponded to the anatomical location of visual cortex. This proof of principle report opens up the exciting possibility of mapping cortical columns in human visual cortex and to evaluate whether organizational principles established in studies of visual cortex in animals applies to humans.^16,49,50

4.3.

Language Cortex

The potential for using ISOI to map language areas in humans is one of its most exciting prospects and has important practical advantages in neurosurgery. Moreover, the greater spatial resolution of ISOI over fMRI mapping potentially offers fundamental insights into the cortical organization underlying speech production and comprehension; these functions are unique to humans and can be directly studied only in humans.

In the first study of language with optical imaging,²⁰ areas of activation were observed in the posterior inferior frontal gyrus (IFG) (the region traditionally known as Broca’s area) during a naming exercise in a single patient. Interestingly, these areas were not the same as those found to be essential language sites (sites that caused speech arrest during ESM). On the contrary, in the area that caused speech arrest, optical changes were observed in the opposite direction showing a relative increase in HbR during a naming exercise. This distinction was not observed, in three patients where temporal cortex, an area traditionally referred to as Wernicke’s area, was investigated. Imaging revealed areas of activation in two sites that were found to be essential language sites during ESM with additional areas of activation in three other temporal sites, likely secondary language sites. ESM is limited because it reliably identifies only essential language sites; however, secondary language sites are important in language comprehension and can be identified with ISOI. Extending beyond location of language cortices, Cannestra et al.²¹ studied language cortex with ISOI in 10 patients using both object naming and word discrimination tasks. Depending on the task, different activation patterns were observed. Object naming activated both anterior and posterior regions of the IFG. Auditory responsive naming activated a more posterior language area of the superior temporal gyrus (STG) and word discrimination paradigms activated only the posterior region of the STG. The authors suggest that anterior activations in the IFG reflect semantic processing, while phonological processing is the source of posterior activations. In the STG, they suggest more anterior/superior activations arise from phonological processing and more posterior/inferior activations are the result of semantic processing. Additionally, a case of ISOI of bilingual cortical representations during a visual object naming task was reported.⁵¹ Imaging revealed areas of activation common to both languages as well as language-specific activation sites in the supramarginal (Spanish) and precentral (English) gyri. The authors concluded there are both distinct and overlapping components in bilingual language representation. In both of these reports, findings from ISOI expanded beyond standard functional mapping to better describe language cortex itself.

Use of ISOI to study language cortex is especially important as its use clinically to delineate eloquent cortex with high resolution would allow for more optimal resection while preserving language function. Compared to fMRI, ISOI has the advantage of being conducted intraoperatively and has been proven useful when used in conjunction with ESM (see Sec. 6). Moreover, its use could provide meaningful new insight into our understanding of how language cortex is organized.

4.4.

Hemodynamic Oscillatory Activity

In multiple species, including humans, large amplitude $\sim 0.1\mathrm{Hz}$ oscillations in cortical hemodynamics known as slow sinusoidal hemodynamic oscillations (SSHOs) have been observed. Rayshubskiy et al.⁵² presented two cases, one in which SSHOs were observed using intraoperative ISOI and another where they were not. In the case where SSHOs were present, they were spatially localized and exhibited wave-like propagation. Moreover, they were localized to specific pial arterioles, but not others, indicating that this was not the result of systemic blood pressure oscillations. For the case where SSHOs were observed, preoperative fMRI data collected 4 days prior to optical imaging data also demonstrated $\sim 0.1\mathrm{Hz}$ oscillations in the BOLD signal around the same region. The observation of SSHOs is useful in that it could serve as a biomarker for neurological disease and should be taken into account as a potential confounder for fMRI.