Advances in Neuroimaging and Monitoring to Defend Cerebral Perfusion in Noncardiac Surgery

The 2019 Detection and Neurological Impact of Cerebrovascular Events in Noncardiac Surgery Patients: A Cohort Evaluation study (NeuroVISION) reported an alarmingly high incidence of brain injury after noncardiac surgery, with 7% of patients older than 65 yr experiencing postoperative covert stroke.⁷  Perioperative neurocognitive disorders are even more frequent, with early postoperative cognitive dysfunction and postoperative delirium reported in 10 to 15% and 20 to 30% of patients after noncardiac surgery, respectively.^12–14  These figures are likely to be conservative as there are methodologic limitations that confound the detection and assessment of these disorders.

Perioperative neurologic complications have a major effect on patient outcomes. Covert stroke is associated with both early postoperative cognitive dysfunction and a 13% increase in cognitive decline at 1 yr.^7,15,16  Overt perioperative stroke conveys a 13.3% risk of 30-day mortality¹⁷  and presents a substantial healthcare burden in the form of lifelong neurologic sequelae in 69% of cases.¹⁸  Postoperative cognitive dysfunction and delirium both delay patient recovery. Data show these conditions double the average length of hospital stay and increase both short- and long-term patient mortality.^19,20  The burden of these complications is worsened by their often covert nature. Only 40 to 50% of patients with delirium are diagnosed and treated, and the majority of ischemic brain lesions fail to present with classic stroke symptoms, yet markedly impair long-term cognitive recovery.^5,19,20 

Anesthesia and surgery may disrupt cerebral perfusion and lead to imbalance between cerebral oxygen supply and demand. Systemic factors such as hypotension or autonomic dysfunction can precipitate hypoperfusion. These hemodynamic disturbances have been a popular proposed treatment target, as cerebral perfusion can be measured, and systemic physiology may be optimized to maintain perfusion—for example, optimizing blood pressure, oxygenation, ventilation (carbon dioxide), and depth of anesthesia. Despite these known physiologic insults, and considerable clinical and research focus, to date, no study has demonstrated conclusive benefit from individualized perioperative hemodynamic optimization.⁴  One possible reason for the lack of conclusive data is that these studies are dependent on development of sensitive objective outcome measures for neurologic dysfunction. Another is that, as in the case of brain injury, other mechanisms such as inflammation may dominate.

Advances in neuroimaging and monitoring hold the key to identifying covert, previously undetectable perioperative organ injury and examining the role of optimizing cerebral perfusion in the perioperative setting. Neuroimaging and monitoring can assess the brain both as the primary site of injury and as an indicator of adequate organ perfusion. The brain is the most metabolically active organ, and its vascular resistance adapts to match blood flow with metabolic requirements in response to both systemic perfusion pressure changes (myogenic cerebral autoregulation) and local neuronal activity (metabolic cerebral autoregulation or neurovascular coupling). These characteristics potentially make the brain a viable index organ for monitoring the effects of systemic physiology on organ microcirculation, cellular metabolism, and oxygenation, in addition to primary brain pathophysiology.²¹ 

There is currently no ideal universal neuroimaging or neuromonitoring modality available to reliably characterize cerebral well-being in the perioperative setting. Monitoring is essential to decipher perioperative changes in cerebral hemodynamics, oxygenation metabolism, and function, as these may often be clinically silent. Furthermore, specific tools are needed to identify the functional or structural phenomena that correlate with subtle or unexplained clinical manifestations, as well as clinical syndromes such as delirium, which represent a spectrum of pathophysiology. Historically, such imaging has included positron emission tomography and single photon emission computed tomography; however, as these techniques use ionizing radiation or radioactive tracers, their use is inherently restricted to limit radiation exposure. Continuous assessment with nonionizing, noninvasive techniques such as optical techniques and ultrasonography provide alternative options for cerebral hemodynamic monitoring in the perioperative clinical environment, as does the application of serial neuroimaging methods like magnetic resonance imaging. Advances in the past 5 yr have enabled unique assessment of both neurologic and cerebrovascular function, ranging from direct measurement of cellular metabolism to global functional and brain network imaging via connectomics.²² 

All the available modalities have trade-offs between spatial and temporal resolution, and constraints to deliver at the bedside, particularly in the operating room. For example, magnetic resonance imaging creates a snapshot with high spatial resolution but can miss transient or important temporal phenomena, which occur frequently in the perioperative period. In addition, monitoring in the operating room is not generally feasible outside of a dedicated image-guided operating theater or cardiac suite due to safety and practical constraints. Conversely, optical techniques, such as near-infrared spectroscopy, can be performed at the bedside in the various patient environments, including operating rooms and hospital wards. However, spatial resolution is limited, and large, cumbersome monitoring arrays can be needed for data collection.

In the following sections, we outline indices of neurologic and cerebrovascular function that are relevant to perioperative management. This review will focus on emerging techniques that have the capacity to perform multisite measurement/imaging of biomarkers of neurologic injury used to optimize cerebral perfusion: optical, ultrasonographic, and magnetic resonance techniques. We have not included electroencephalography as this has been reviewed extensively elsewhere and does not directly reflect a representation of regional perfusion.²³  A complete summary of the imaging modalities, along with their respective strengths, weaknesses, and limitations, can be found in table 1.

Optical techniques have the capability to measure a range of properties of superficial cortical tissue continuously and completely noninvasively at the bedside. Via these techniques, it is possible to measure surrogates of cerebral blood flow, as well as cerebral oxygenation, and even cellular oxygenation and/or metabolism. Due to their noninvasive nature and relative portability, optical modalities are particularly useful for continuous monitoring in the intraoperative and/or postoperative environment.

Optical brain monitoring exploits the physical properties of biologic tissues. For example, human tissue is relatively transparent in the near-infrared light spectrum (from 700 to 1,000 nm), which permits the interrogation of superficial cerebral structures. Introduced light diffuses through superficial cerebral structures and is subject to scattering and absorption interactions. Photon-tissue absorption and scattering interactions can be used to identify specific tissue chromophores (hemoglobin and cytochrome oxidase) and physical properties (flow) in the superficial cerebral cortex.

Near-infrared spectroscopy is a noninvasive continuous optical technique that can measure the concentration of different tissue chromophores at the bedside. In clinical practice, near-infrared spectroscopy is used to predict the balance between oxygen supply and demand via regional hemoglobin−oxygen saturation. As the concentration of specific tissue chromophores is directly proportional to light attenuation, the concentration of chromophores, such as oxyhemoglobin (HbO₂) and deoxyhemoglobin (HHb), is derived using modifications of the Beer−Lambert law. Absolute regional cerebral hemoglobin oxygen saturation ([HbO₂]/[HbO₂ + HHb]) is the main parameter reported by clinical devices and is employed as a surrogate reflecting the balance between tissue oxygen supply and demand. Regional cerebral hemoglobin oxygen saturation is derived from the slope of attenuation between two spaced detectors, exploiting simplifications to the diffusion approximation of light transport in tissues.

The obvious caveat of this modality is the inability to discern underlying tissue architecture, which includes a mix of venous, arterial, and capillary blood and a variation of light scattering. Consequently, regional cerebral hemoglobin oxygen saturation values are typically much lower than arterial or pulse oximetry−derived values and higher than jugular bulb venous oxygenation when measured with an invasive catheter. Methods such as time and frequency domain spectroscopy report additional tissue characteristics and have been developed to overcome some of these constraints.²⁴  Jugular venous oximetry is an obvious comparator to near-infrared spectroscopy offering direct measurement of oxygen concentration in cerebral venous blood.³⁹  Such a measurement affords insight into cerebral oxygen extraction fraction but requires invasive insertion of a central cannula, and reflects oxygen extraction approaching a global/hemispheric cerebral level.⁴⁰  Near-infrared spectroscopy contrasts to this being noninvasive and capable of measurements over multiple regions, or even reconstruction into images.

There are multiple near-infrared spectroscopy techniques available to clinicians. Each technique differs with regards to the specific wavelengths used, the number of wavelengths, the number of light detectors and their distance from the emitter, and the type of modulation of emitted light (frequency and time domains).⁴¹  These variations were developed to address two of the main disadvantages of near-infrared spectroscopy: the contamination of the signal by extracerebral tissue and inhomogeneity of light scattering within cerebral tissue.⁴²  Spatially resolved spectroscopy is commonly employed in clinical devices and typically comprises a single light source (two to five wavelengths) and two distal detectors. Oxyhemoglobin and deoxyhemoglobin concentrations are obtained, and regional cerebral hemoglobin oxygen saturation is calculated by the gradient of light attenuation measured between detectors, which also serves to attenuate (albeit not abolish) contamination of the signal by extracerebral tissues (fig. 1). This technique assumes that scattering properties are static and homogenous and conform to a population average.

Frequency-domain near-infrared spectroscopy and time-resolved near-infrared spectroscopy are advanced techniques that characterize both absorption and scattering. Frequency-domain near-infrared spectroscopy employs a pulsed high-frequency light source, with distortion of the frequency of these pulses at the detector related to the scattering properties of tissues. Time-resolved spectroscopy introduces a picosecond light pulse into tissue and measures the temporal point spread function of photon arrival time at a distal detector.⁴²  This function can be fitted to the Patterson equation, an approximation of light diffusion, to calculate absorption and scattering coefficients and thus absolute regional cerebral hemoglobin oxygen saturation.⁴² 

While hemoglobin species can be resolved using as few as two wavelengths of light, multiple wavelengths can be combined (in broadband spectroscopy) to resolve other tissue chromophores. Cytochrome c oxidase (complex IV in the mitochondrial respiratory chain) is of particular interest, as it reflects cellular oxygenation and metabolism.⁴³  The redox status of the copper complex within the cytochrome c oxidase molecular structure can be identified (and separated from hemoglobin) using broadband near-infrared spectroscopy and an adjusted Beer−Lambert technique between 780 and 900 nm.

In addition to assessing the amount of light that is absorbed by a tissue, the degree of scattering can be measured to identify other tissue properties with similar optode detector arrangement. Diffuse correlation spectroscopy introduces a long coherence laser light source to observe small changes in light intensity that relate to scattering from movement of red blood cells.⁴⁴  An autocorrelation function of the measured fluctuations can be converted into a blood flow index and provide an estimate of regional cerebral blood flow. In contrast to positron emission tomography and magnetic resonance imaging, diffuse correlation spectroscopy offers real-time assessments of regional blood flow and has been shown to strongly correlate with cerebral blood flow values measured using transcranial Doppler ultrasound, xenon-enhanced computed tomography, and arterial spin labeling magnetic resonance imaging.^45– This assessment can be augmented with indocyanine green dye, which can calibrate the diffuse correlation spectroscopy to absolute cerebral blood flow values.⁴⁸  The volume of tissue that is assessed is subject to the same limitations as near-infrared spectroscopy, reflecting intracranial and extracranial components; however, recent developments, such as time domain diffuse correlation spectroscopy, can be used to minimize extracerebral contamination.²⁵  Other approaches, including an epidural diffuse correlation spectroscopy optode to measure spinal cord blood flow, have been described to evaluate deep tissues.²⁶ 

Scattering of light can also be induced using ultrasound directed at cerebral tissue, known as the acousto-optic effect.²⁷  By tagging near-infrared light using ultrasound, users can make a depth-resolved estimate of both blood flow and oxygenation.²⁸ 

Diffuse optical tomography uses large arrays of near-infrared spectroscopy light sources and detectors, using either continuous-wave, time-resolved spectroscopy, or even diffuse correlation spectroscopy to create topographical maps of measurements. Diffuse optical tomography can predict changes in brain oxygenation and cerebral blood flow in regional maps, rather than in a single channel. This approach is used as a functional neuroimaging modality in research and can monitor brain physiology in real time.²⁶  Advances in system miniaturization and analysis software are increasing the potential usefulness of diffuse optical tomography in the perioperative setting.²⁷ 

Transcranial Doppler ultrasound is a well-established technique for evaluating cerebral perfusion intraoperatively. It uses Doppler ultrasound (2 MHz) to insonate basal cerebral vessels. Doppler ultrasound reports a flow velocity, which allows for the detection of emboli through characteristic features on the frequency spectrogram. Further ultrasonographic iterations include color M-mode, which reports velocity and depth; B-mode color Doppler, which reports a two-dimensional image and Doppler flow; and three-dimensional Doppler. The Doppler frequency shift between emitted and received ultrasound relates to the speed of blood flow in the region. Assuming a constant vessel radius (one theoretical limitation to this technique), the measured changes in flow velocity are proportional to cerebral blood flow.⁴¹  This technique relies on identifying a suitable transcranial window, which is an area where acoustic signal can move adequately through tissue and insonation of the vessels of interest can be achieved. Transcranial Doppler ultrasound is frequently performed over the temporal bone, which enables the assessment of the middle, anterior, and posterior cerebral arteries (fig. 2A). Insonation of other major vessels is achievable using submandibular, transorbital, or suboccipital probe placements (fig. 2B). Transcranial Doppler ultrasound measures flow velocity through a given vessel in centimeters per second (fig. 2C).²⁹  Transcranial Doppler ultrasound is a powerful clinical tool, but its use is associated with several theoretical limitations. First, prolonged monitoring requires a trained and experienced transcranial Doppler ultrasound operator due to some challenges with maintaining probe position and window visualization even with the aid of commercially available fixation devices. Second, due to artefacts created from diathermy and movement intraoperatively, it can be difficult to maintain insonation for the continuous measurement of flow and identification of markers of dysfunction or injury. Vessel identification using concurrent B-mode ultrasound and tracking using robotic transcranial Doppler ultrasound might assist with this challenge.

Magnetic resonance imaging can be used to diagnose perioperative stroke in the early stages and, due to its exquisite spatial resolution, is excellent for assessing regional blood flow in the brain.³²  While structural magnetic resonance imaging scans can discern tissue types based on the intrinsic constraints on hydrogen atoms within tissue, advanced sequences allow for identification and measurement of regions of the brain where hydrogen atoms within blood are flowing. The two main imaging sequences for acquisition of cerebral blood flow data using magnetic resonance imaging are dynamic-susceptibility contrast perfusion and arterial spin-labeling perfusion. Dynamic-susceptibility contrast perfusion depends on administering a contrast agent, with the transit time of a bolus of known volume through cerebral tissue providing an estimate of cerebral blood flow, whereas arterial spin-labeling perfusion measures the transit time of magnetically labeled water in blood.³²  Both magnetic resonance imaging and positron emission tomography cerebral blood flow measurements offer spatial information regarding cerebral perfusion, such that regional comparisons of cerebral blood flow can be obtained. Magnetic resonance imaging can also be used to assess flow through large vessels in the form of phase-contrast magnetic resonance imaging, which provides an estimate of flow velocity.³¹  While magnetic resonance imaging has tremendous utility in the clinic, an important drawback in the perioperative setting is the limited scope to acquire real-time point-of-care information.

Cerebral blood flow can be assessed globally in the large basal cerebral blood vessels or regionally in the microcirculation using the techniques described in the Perioperative Neuroimaging and Neuromonitoring section. Combining measures of cerebral blood flow with blood pressure or oxygenation data allow for the evaluation of autoregulation and cerebral metabolic rate, respectively. Functional connectivity between neuronal networks can be measured by correlating simultaneous monitoring of cerebral blood flow across multiple brain regions.

The cerebral metabolic rate of oxygen can be derived using a Fick model by measuring cerebral blood flow and the oxygen saturation difference between arterial and venous blood.⁵²  Cerebral metabolic rate of oxygen is an estimate of the metabolic demands of cerebral tissue and remains relatively constant during physiologic conditions. As cerebral metabolic rate of oxygen is the amount of oxygen extracted by tissue, it depends on both cerebral blood flow and the oxygen extraction fraction. An assessment of cerebral metabolic rate of oxygen in the perioperative period can indicate whether cerebral tissue is being optimally perfused.

Cerebral metabolic rate of oxygen can be assessed at the bedside using time-domain near-infrared spectroscopy concurrently with transcranial Doppler ultrasound or diffuse correlation spectroscopy⁵³ via a Fick model. This utility provides clinicians with the ability to monitor cerebral metabolic rate of oxygen and hemodynamic adequacy in the intraoperative period. Of note, however, near-infrared spectroscopy limits assessments to a small region of the cerebral cortex, which might not accurately reflect whole-brain cerebral metabolic rate of oxygen and is not a direct measure of oxygen consumption.

Accurate estimates of global cerebral metabolic rate of oxygen can be obtained using magnetic resonance imaging. Two emerging methods for assessing cerebral metabolic rate of oxygen using magnetic resonance imaging are T2 relaxation under spin tagging and susceptometry-based oximetry. T2 relaxation under spin tagging, which uses the same spin-labeling technique as arterial spin-labeling perfusion, can be used to determine the oxygen content of venous blood within the venous sinuses of the cranium and estimate deoxyhemoglobin concentration based on the T2 relaxation time.⁵⁴  In contrast, susceptometry-based oximetry determines the magnetic susceptibility of blood within the venous sinuses to estimate deoxyhemoglobin concentration.⁵⁵  It is possible to estimate cerebral metabolic rate of oxygen by pairing either T2 relaxation under spin tagging or susceptometry-based oximetry sequences with flow estimates from phase-contrast magnetic resonance imaging.⁵⁶ 

In addition to assessing deoxyhemoglobin and oxyhemoglobin, there is growing interest in imaging other chromophores related to cerebral oxygenation by way of broadband near-infrared spectroscopy. One of these chromophores is cytochrome c oxidase, which is the terminal electron acceptor in the mitochondrial electron transport chain. Previous studies have shown that cytochrome c oxidase concentrations increase when oxygen levels in cerebral tissue are elevated and decrease during periods of hypoxia−ischemia.⁵⁷  Changes in oxidized cytochrome c oxidase concentration convey unique benefits over deoxyhemoglobin and oxyhemoglobin, as these changes are more specific for cerebral metabolism and less susceptible to contamination by extracranial tissue.⁵⁸  The change in oxidized cytochrome c oxidase concentration index has been validated in both animal models and healthy adults. ^57,59,60 

Cerebral autoregulation is the mechanism by which blood vessels in the brain dilate and constrict during fluctuations in perfusion pressure to adjust cerebrovascular resistance to maintain a constant cerebral blood flow. Autoregulation functions over a range of perfusion pressures, outside of which these vascular mechanisms fail, and tissue becomes susceptible to injury.^61,62 

Cerebral autoregulation is often conceptualized in terms of a static model. Within this model, blood flow remains stable when the mean arterial pressure (MAP) is between 50 and 150 mm Hg; blood pressure within this range results in stable cerebral blood flow,⁶³  whereas pressure below this range leads to hypoperfusion, and pressure over this range leads to hyperperfusion. Static autoregulation analysis is only possible when a range of MAP values have either been induced or have occurred physiologically so that consequent cerebral blood flow changes can be identified at a steady state (fig. 3). Ideally, modeling this curve for each patient would enable identification of the upper and lower limits of blood pressure beyond which cerebral blood flow is variable.

In contrast to static models, dynamic models of autoregulation incorporate the temporal component of vascular reactivity to enable less invasive and more rapidly calculable indices of autoregulation. Rapid changes in cerebral perfusion pressure invariably result in changes in cerebral blood flow until the cerebral vasculature can adequately adjust to the altered conditions, typically within 5 to 15 s.⁶²  However, gradual changes to cerebral perfusion pressure are more readily compensated for by the vasculature and should not lead to detectable changes in cerebral blood flow.

The main method for assessing cerebrovascular autoregulation involves simultaneously monitoring arterial blood pressure or a surrogate for cerebral blood flow, for example via transcranial Doppler ultrasound−measured flow velocity, near-infrared spectroscopy monitored regional cerebral hemoglobin oxygen saturation, or invasively monitored brain tissue oxygenation. Autoregulation acts as a high-pass filter in the relationship between arterial blood pressure and cerebral blood flow, and its effectiveness can be defined by the phase and strength of the relationship (known as transfer function). Continuous monitoring is desirable clinically but can be challenging to conduct over long periods with ultrasonography due to the window, probe fixation, and sources of artefact. Other surrogates for cerebral blood volume and cerebral blood flow can be used in a similar fashion. Most notably, intracranial pressure (ICP) has been used as a surrogate for the assessment of cerebral blood volume and/or cerebrovascular reactivity. Although not used electively in the perioperative setting, this signal processing approach has been generalized to other modalities. When autoregulation is intact, slow waves (0.003 to 0.05 Hz) in ICP and MAP are not correlated, which suggests that slow changes in perfusion pressure are being adequately accommodated.⁶⁴  When autoregulation is compromised, these slow waves are correlated, indicating that local ICP is coupled to MAP without compensating vascular mechanisms. This relationship is called the pressure reactivity index, which is the Pearson correlation coefficient between MAP and ICP values.⁶⁵  Pressure reactivity index values range from –1 to 1, with negative and lower values indicating better performance of autoregulatory mechanisms. Pressure reactivity index values greater than 0.3 indicate that autoregulation is not intact.

After obtaining multiple ICP and MAP measurements, a curve can be constructed that demonstrates the relationship between cerebral perfusion pressure and pressure reactivity index, as shown in figure 4. This curve is U-shaped, with the theoretically optimal cerebral perfusion pressure values resulting in the lowest possible pressure reactivity index values, and pressure reactivity index values increasing as cerebral perfusion pressure deviates from the range of optimal values (reflecting the cerebral perfusion pressure range of cerebral autoregulation).⁶⁶  As such, repeated calculation of pressure reactivity index at different cerebral perfusion pressure values allows estimation of the optimum cerebral perfusion pressure, and instantaneous determination of pressure reactivity index at a specific cerebral perfusion pressure may indicate whether the upper or lower limit of autoregulation has been surpassed.

While invasive monitoring of ICP is not feasible during routine noncardiac surgery, similar metrics have been developed for neuroimaging. For transcranial Doppler ultrasound, simultaneously measuring middle cerebral artery flow velocity and systemic arterial blood pressure permits determination of the mean velocity index, with Pearson correlation coefficient assessing slow waves in ICP and arterial blood pressure waveforms. With near-infrared spectroscopy, the tissue oxygen index is derived from slow waves in regional cerebral hemoglobin oxygen saturation and arterial blood pressure waveforms, and the total hemoglobin index is derived from total hemoglobin and arterial blood pressure.⁶⁷  As diffuse correlation spectroscopy provides a direct estimate of cerebral blood flow, it has the potential to form the basis for a more sensitive autoregulation index than near-infrared spectroscopy.⁶⁸ 

Advances in signal processing promise to improve on the fidelity of these noninvasive techniques. For example, wavelet analysis can be used to define changes in frequency and time domains simultaneously and has the potential to provide more accurate determinations of optimum cerebral perfusion pressure (fig. 5).⁶⁹ 

An additional method for assessing autoregulation is to determine vasodilatory reserve. Vasodilatory reserve assesses the maximal capacity of a patient’s cerebral vasculature to dilate to accommodate decreases in blood pressure.³⁰  This assessment can be made preoperatively through a vasodilatory challenge using carbon dioxide or acetazolamide as measured with computed tomography or magnetic resonance imaging.^70–72  Critical closing pressure, which is the cerebral perfusion pressure where measurements of flow are modeled to drop to zero, can be derived from transcranial Doppler ultrasound or diffuse correlation spectroscopy.⁴⁶ 

The brain achieves its function by coordinated action between millions of neuronal networks. Connectomics is the study of brain connectivity between these networks, and involves the analysis of synchronous activity within the brain as detected by changes in regional blood flow or through imaging of underlying tissue architecture (fig. 6).²²  The pattern of neural activity in different regions of the brain under a given state is referred to as a connectome.⁷³  This technique reveals the underlying structural and functional connectivity of the brain, which differs between healthy and disease states and might offer novel avenues for detecting subclinical or overt injuries early in their development.

Functional magnetic resonance imaging has become the definitive standard for assessing functional brain connectivity.³⁷  Blood oxygenation level−dependent contrast-enhanced imaging is the predominant functional magnetic resonance imaging method. Local increases in metabolic activity result in transient accumulation of paramagnetic deoxyhemoglobin, but because of neurovascular coupling, there is a reflex increase in cerebral blood flow, leading to rapid clearance of deoxyhemoglobin. This decrease in deoxyhemoglobin concentration produces a blood oxygenation level−dependent signal and acts as a proxy for neurologic activation.³⁸  Regional activation patterns in the cerebral cortex are analyzed topologically, allowing networks of interconnected regions to be identified; these networks similarly connect to one another, adding layers of complexity to brain connectivity.⁷³ 

Connectivity can also be assessed structurally by generating tractograms using diffusion magnetic resonance imaging.³⁴  Diffusion-weighted magnetic resonance imaging detects restriction of diffusion by cellular compartments within the brain. Diffusion-weighted magnetic resonance imaging will detect an increased signal where water is more tightly restricted, whereas it will detect a decreased signal where water is diffusing more freely. The signal received is related to the isotropy of a volume of water (anisotropy is a term indicating that diffusion is restricted).⁷⁴  Tractography methods, such as diffusion tensor imaging, take advantage of the strong anisotropy of white matter tracts to generate measurements reflecting axonal connections between brain regions.⁷⁴ 

The most basic diffusion magnetic resonance imaging tractography technique is diffusion tensor imaging, which uses six diffusion directions acquired in three perpendicular axes to derive tracts from mathematical tensors.³⁵  More robust tractograms can be derived though high-angular resolution diffusion imaging, which involves the acquisition of diffusion images in many directions (often hundreds). This approach allows for the construction of complex orientation distribution functions that provide additional information regarding fiber orientation and density.^75,76 

Further information regarding the composition of tissue can be acquired using diffusion kurtosis imaging. This technique involves the acquisition of images at multiple magnetic field strengths to determine how diffusion of water in a given region of the brain differs from a normal Gaussian distribution^36,77  (fig. 7). Due to its high sensitivity, diffusion kurtosis imaging can also be used to identify alterations to the cellular and parenchymal architecture. This utility means diffusion kurtosis imaging can be used in the diagnosis of neurodegenerative conditions such as Parkinson disease and Alzheimer disease.^78,79 

While magnetic resonance imaging is useful in the preoperative and postoperative settings, it is only used intraoperatively during neurosurgery. Near-infrared spectroscopy is a viable alternative to magnetic resonance imaging as it can achieve a similar level of functional mapping. Functional near-infrared spectroscopy is a form of cerebral oximetry that uses multiple optodes arranged in a wearable cap to capture measurements of cerebral oxygenation across the entire cortex, instead of just a sample of tissue (as in bifrontal near-infrared spectroscopy).⁸⁰  Signals from these optodes can be modeled through diffuse optical tomography,²⁷  which allows for oxyhemoglobin and deoxyhemoglobin concentrations to be mapped to the cerebral cortex with high spatial and temporal resolution.

Clinicians have access to a myriad of tools to help them estimate the risk of overt, clinically apparent surgical complications based on demography and clinical factors. There is, however, a lack of tools available for the practical individualized assessment and estimation of subtle neurologic impairment.^81,82  Advances in technology have the potential to optimize current preoperative assessment practices and further minimize the risk of surgical complications. One potential strategy that could benefit patient outcomes and improve resource allocation is to use neuroimaging to conduct a detailed preoperative assessment of brain frailty.

Neuroimaging methods have achieved mainstream utility in the context of neurosurgery for functional and structural brain mapping (discussed in more detail in the Postoperative Diagnostic Imaging Assessment section).⁸³  Outside of neurosurgery, however, little evidence exists for the usefulness of preoperative neuroimaging methods; consequently, they are rarely used.⁸⁴  There has been growing focus on objective quantification of patient vulnerability beyond basic features pertaining to medical history and demographics, including the use of preoperative assessments of cognitive function, which enable risk stratification and specific measurement of cognitive decline. This practice is particularly true for the assessment of cognitive vulnerability, a major issue in older patients who have a high risk of postoperative cognitive decline and delirium.^85,86  This section highlights specific applications of neuroimaging in the preoperative period toward this same end.

Preoperative assessment of vasodilatory reserve aims to assess the need for surgical intervention for conditions involving the major blood vessels of the head and neck, as well as enabling risk stratification, such as in moyamoya disease.⁸⁷  For carotid endarterectomy, preoperative assessment of the vasodilatory reserve of the cerebral vasculature has been shown to aid in predicting adverse cerebrovascular events.⁸⁸  Carotid endarterectomy is associated with a high risk of transient hypoperfusion and embolization, as well as the possibility of cranial hyperperfusion syndrome immediately after correction of the stenosis.⁸⁹  As such, vasodilatory reserve is essential for maintaining adequate cerebral perfusion during and immediately after this surgery.

Across several small clinical studies, impaired vasodilatory reserve has been associated with an increased risk of postoperative stroke or transient ischemic attack post−carotid endarterectomy.^30,71,72  Several studies used transcranial Doppler ultrasound or 133-xenon emission tomography to monitor blood flow before and after the administration of acetazolamide (a potent vasodilator). An increase in middle cerebral artery flow velocity of 40% or more, as measured by transcranial Doppler ultrasound, was found to be protective against postoperative cerebrovascular events, such as stroke or transient ischemic attack, whereas an absent or less than 40% increase in cerebral blood flow was associated with an increased risk.³⁰  Similarly, an increase in cerebral blood flow of more than 20% after inspiration of 8% CO₂ was associated with a decreased risk of cerebrovascular events.⁷²  In a meta-analysis of 991 patients with carotid stenosis or occlusion, reduced vasodilatory reserve was associated with a four-fold increased risk of stroke or transient ischemic attack.⁶⁴ 

While these studies indicate that preoperative assessment of cerebrovascular function, specifically vasodilatory reserve, might aid in predicting injury associated with carotid endarterectomy, no studies have validated the use of preoperative neuroimaging for risk assessment in patients undergoing other procedures. An assessment of indices of cerebrovascular function before noncardiac surgery is an enticing avenue of future research. This approach could be used as a marker of autoregulation, aid with the prediction of the optimal perioperative MAP target, and improve patient risk stratification. Such markers could guide strategies for monitoring, intraoperative management, and postoperative diagnostic testing, although data are lacking to support these applications.

In the context of major cardiac surgery, one study assessed the utility of preoperative regional cerebral hemoglobin oxygen saturation assessment. In 1,178 patients undergoing procedures requiring cardiopulmonary bypass, a minimum preoperative absolute regional cerebral hemoglobin oxygen saturation score recorded by bifrontal near-infrared spectroscopy of less than 50% was associated with a significantly increased risk of both mortality and morbidity.⁹⁰  However, a regional cerebral hemoglobin oxygen saturation less than 50% was also associated with a higher burden of comorbidities, including decreased left ventricular ejection fraction and heart failure. Thus, while it is unclear whether including regional cerebral hemoglobin oxygen saturation itself in preoperative screening improves the prediction of patient risk associated with cardiac surgery, further investigation in noncardiac surgical settings is warranted.

Application of neuroimaging methods in the intraoperative period has the potential to protect patients from injury through enabling early detection of organ hypoperfusion and avoidance of tissue oxygen supply−demand mismatch. Use of available modalities are obviously limited to those that can be applied noninvasively with minimal imposition on the surgical field and are safe for continuous real-time monitoring. As such, most intraoperative applications in noncardiac, nonneurologic surgery involve the use of optical and ultrasonographic methods for monitoring cerebral perfusion.

Intraoperative monitoring of cerebral perfusion has primarily been validated for major cardiac procedures requiring cardiopulmonary bypass. In this setting, decreases in cerebral blood flow or cerebral oxygenation during the operative period are usually directly correlated with surgical or anesthetic factors (for example, large-vessel manipulation and/or clamping, pump flow alterations, blood volume loss) and are thus readily corrected. Currently, however, intraoperative neuromonitoring for noncardiac surgery is still an emerging field.

Our understanding of the clinical relevance of cerebral regional cerebral hemoglobin oxygen saturation or hemoglobin concentrations is still incomplete in the intraoperative setting. A systematic review of the literature reported mixed findings, with insufficient evidence yet to support or refute an outcome benefit.⁹¹  The utility of any monitoring technique relates to the effectiveness of clinical intervention. For example, interventions in response to regional cerebral hemoglobin oxygen saturation focus on increasing cerebral oxygen delivery by manipulating several physiologic processes, including blood pressure, ventilation and carbon dioxide, inspired oxygen fraction, transfusion, and reduction of neuronal activity by increasing depth of anesthesia, to reduce cerebral metabolic rate of oxygen. Many critics highlight intrinsic limitations of widely used clinical near-infrared spectroscopy monitors, including the inability to discern differences in underlying tissue composition within the sampled near-infrared spectroscopy region.²⁴  Therefore, the regional cerebral hemoglobin oxygen saturation value depends on not only tissue oxygenation but also tissue composition, which cannot be readily standardized between patients or tests. This point was demonstrated by Bickler et al., who compared multiple oximetry devices and found high variability dependent on underlying tissue architecture.⁹²  As such, baseline regional cerebral hemoglobin oxygen saturation values in healthy adults vary considerably, with studies reporting average baseline regional cerebral hemoglobin oxygen saturation values between 55 and 80%. However, studies often focus on changes in cerebral oxygen saturation rather than absolute measurements. While alterations in regional cerebral hemoglobin oxygen saturation in the intraoperative period have been associated with adverse patient outcomes during some major cardiac and vascular procedures, the definitions of significant decreases in regional cerebral hemoglobin oxygen saturation have varied considerably.

It is possible that time-domain or frequency-domain near-infrared spectroscopy might reduce the effects of heterogeneity in tissue composition and allow the detection of subtle changes in oxyhemoglobin or deoxyhemoglobin concentrations. Variations in total hemoglobin content might prove to be more accurate than regional cerebral hemoglobin oxygen saturation changes for detecting intraoperative hemodynamic perturbations. Future research should focus on emerging optical imaging methods within this area to help resolve these unanswered questions.

Multiple studies have evaluated associations between intraoperative neuromonitoring variables and the incidence of delirium after surgery. These studies are confounded by the constraints and variability underlying near-infrared spectroscopy, which have been explained in detail, as well as evolution in the appropriate use of screening and diagnostic tools for neurologic outcome. A meta-analysis from 2020 found that intraoperative regional cerebral hemoglobin oxygen saturation monitoring was not associated with a decreased rate of postoperative delirium, as defined by standardized clinical screening tools including the Confusion Assessment Method, the Confusion Assessment Method−Intensive Care Unit, the Diagnostic Statistical Manual of Mental Disorders (4th or 5th edition) criteria, or the Mini Mental Status Examination.⁹³  Of note, the Mini Mental Status Examination is not considered an accurate diagnostic test for delirium, suggesting that some studies employ inappropriate assessment tools, and highlights a need for objective measures of postoperative neurocognitive disorders.⁹⁴ 

However, some noncardiac studies have detected associations between specific near-infrared spectroscopy metrics and postoperative delirium. A 2018 study investigated the utility of near-infrared spectroscopy during urgent surgery for hip fracture in 40 patients over 65 yr of age.⁹⁵  Investigators commenced bifrontal near-infrared spectroscopy monitoring from the moment patients were admitted to the hospital, such that cerebral oxygenation was monitored during resuscitation and surgery. The results showed that lower nadir regional cerebral hemoglobin oxygen saturation values were associated with increased risks of postoperative delirium, as well as 30-day mortality. While both studies identified statistically significant associations between regional cerebral hemoglobin oxygen saturation decreases and adverse outcomes, the reported differences were of little clinical significance, and the studies included small cohorts of high-risk patients. Thus, these findings provide limited support for the use of intraoperative regional cerebral hemoglobin oxygen saturation monitoring to reduce the likelihood of delirium after noncardiac, nonneurologic surgery.

Recognizing the 2018 change in nomenclature to the preferred “neurocognitive dysfunction” and “neurocognitive disorder,” most literature has historically reported postoperative cognitive dysfunction.⁶  Postoperative cognitive dysfunction has been associated with multiple near-infrared spectroscopy and transcranial Doppler ultrasound values in several noncardiac cohorts. A meta-analysis detected a statistically significant decrease in postoperative cognitive dysfunction incidence when intraoperative regional cerebral hemoglobin oxygen saturation monitoring was used (odds ratio, 0.53 [95% CI, 0.39 to 0.73]; n = 1,380).⁹³  Among the studies included in this meta-analysis, postoperative cognitive dysfunction definitions were primarily based on decreased postoperative performance on the Mini Mental Status Examination or Montreal Cognitive Assessment, recognizing the perioperative limitations and time constraints of the definitive standard of neuropsychiatric assessment batteries. The efficacy of regional cerebral hemoglobin oxygen saturation monitoring for preventing postoperative cognitive dysfunction has been examined in patients undergoing cardiac surgery or major vascular surgery and major abdominal surgery.^96,97 

For major cardiac and vascular surgery, a decrease in absolute regional cerebral hemoglobin oxygen saturation to less than 50% or a decrease of more than 20% in regional cerebral hemoglobin oxygen saturation below baseline has been associated with postoperative cognitive dysfunction.⁹⁸  Additionally, an absolute regional cerebral hemoglobin oxygen saturation nadir less than 35% or a transient decrease in absolute regional cerebral hemoglobin oxygen saturation to less than 40% persisting for more than 10 min has been associated with postoperative neurologic dysfunction.⁹⁷ 

In a retrospective study of 125 older adults (older than 65 yr) undergoing knee replacement surgery with spinal anesthesia, bifrontal near-infrared spectroscopy was used to monitor decreases in cerebral oxygen saturation during surgery.⁹⁹  Interhemispheric differences in regional cerebral hemoglobin oxygen saturation throughout the procedure were associated with memory decline at 3 months postoperatively.

In 122 patients undergoing major abdominal surgery, regional cerebral hemoglobin oxygen saturation−guided care was found to decrease the incidence of postoperative cognitive dysfunction when assessed by serial Mini Mental Status Examinations.⁹⁶  In a retrospective study of 46 older patients (greater than 55 yr of age) undergoing major abdominal surgery, cerebral near-infrared spectroscopy was used to assess the safety of specific aspects of surgery. A statistically significant regional cerebral hemoglobin oxygen saturation decrease (greater than or equal to 20% decrease below baseline) occurred in 11 of the 46 patients, and in 6 of these individuals, there was a clear temporal association between the decrease in regional cerebral hemoglobin oxygen saturation and intraoperative hemorrhage.¹⁰⁰  Decreases in regional cerebral hemoglobin oxygen saturation occurred despite systolic arterial blood pressure remaining normal and were corrected only by blood transfusion. Across the entire cohort, there was a statistically significant correlation between decreases in regional cerebral hemoglobin oxygen saturation and hemoglobin, correctable by blood transfusion, indicating that monitoring conventional hemodynamic parameters alone was insufficient for identifying periods of insufficient cerebral oxygen supply. Oximetry was necessary for real-time assessment of end-organ perfusion during periods of hemodynamic compromise. An exploratory study investigated the relationship between intraoperative factors during liver transplantation and cerebral oxygenation to identify which steps were associated with an increased risk of end-organ hypoperfusion.¹⁰¹  Several procedural steps, including portal vein clamping, administration of vasoactive substances, and phlebotomy, were found to significantly alter cerebral perfusion.

Near-infrared spectroscopy is commonly recommended in orthopedic operations requiring the beach chair position. Patient positioning can reduce cerebral perfusion pressure based on the relative position to the heart. This positioning can threaten cerebral perfusion, resulting in catastrophic global hypoxic injury, particularly in the setting of autonomic compromise.¹⁰²  Monitoring cerebral perfusion has shown some utility for the assessment of the safety of different patient positions during specific procedures, including shoulder surgery performed in the beach chair position. Cerebral perfusion as measured by both near-infrared spectroscopy and transcranial Doppler ultrasound is impaired in the seated position under anesthesia, which may contribute to an increased risk of neurologic injury.^103,104  Several studies have reported an increased incidence of marked cerebral desaturation events (regional cerebral hemoglobin oxygen saturation decrease) when shoulder surgery is performed with the patient in the seated versus prone position.^105–109  However, only one of these studies correlated this increased risk of cerebral desaturation with decreased performance on postoperative neurocognitive testing.¹⁰⁵  By contrast, prone positioning does not seem to decrease regional cerebral hemoglobin oxygen saturation values during orthopedic surgery.¹¹⁰ 

Intraoperative failure of cerebral autoregulation might increase the likelihood of perioperative complications. Periods of cerebral autoregulation failure have been identified in patients undergoing noncardiac surgery. One study that included 140 patients who had major noncardiac surgery measured the optimal near-infrared spectroscopy−derived autoregulation index (tissue oxygen index) for each patient and reported that a lower optimal tissue oxygen index (i.e., able to maintain autoregulation at a lower cerebral perfusion pressure) was associated with an improved rate of cognitive recovery on postoperative day 3. A higher optimal tissue oxygen index, however, was associated with reduced cognitive recovery and an increased rate of major adverse events.¹¹¹  In a study of 66 patients who had neurosurgery, tissue oxygen index values were retrospectively calculated using intraoperative blood pressure and bifrontal near-infrared spectroscopy readings.¹¹²  Based on intraoperative tissue oxygen index values, the authors estimated the optimal perfusion pressure and compared this with the actual intraoperative arterial blood pressure. In 30 patients, the intraoperative blood pressure was found to be lower than optimal, although clinical outcomes were not evaluated.

Determination of intraoperative autoregulation has the potential to enhance perioperative management. Rather than generalizing therapeutic targets for each patient based on large-scale clinical trials, each patient can be assessed individually to determine which MAP values are likely to provide optimal perfusion of end-organs and decrease the risk of injury (that is, the optimum cerebral perfusion pressure). This approach was proven to be feasible in the setting of traumatic brain injury in the CPPopt [optimum cerebral perfusion pressure] Guided Therapy: Assessment of Target Effectiveness trial (COGiTATE), with trials in the perioperative setting underway.¹¹³ 

There is a fundamental lack of objective and reproducible testing for perioperative neurologic injury. Large clinical trials have shown that perioperative stroke and cognitive dysfunction are frequently diagnosed beyond the immediate postoperative period.¹¹⁴  This finding is largely attributed to difficulties differentiating between the acute effects of surgery and anesthesia, and the complex neurologic sequelae. However, advances in neuroimaging have offered new insights into the structural basis of perioperative neurologic injury and novel avenues for early diagnosis.

Diffusion-weighted magnetic resonance imaging is the definitive standard for identifying acute brain infarction, with a higher sensitivity than conventional magnetic resonance imaging sequences or computed tomography.³³  Lesions are identified by the presence of cytotoxic edema, which occur within 4 h of infarction and persists for up to 2 weeks after surgery.¹¹⁵  As cytotoxic edema results in intracellular swelling, the infarcted area is composed of a large volume of water, which is inherently diffusion-restricted. This direct relationship between pathology and imaging contrast is what makes diffusion-weighted magnetic resonance imaging the most sensitive and specific tool for identifying infarct in the early postoperative period.

The majority of brain infarctions are silent or covert and remain undiagnosed clinically or by neurocognitive testing, hence why the use of diffusion-weighted magnetic resonance imaging is necessary for definitive identification and quantification of these lesions. In the NeuroVISION study, 7% of 1,116 adults aged 65 yr and older and undergoing noncardiac surgery developed new ischemic lesions. These lesions were identified only by postoperative diffusion-weighted magnetic resonance imaging.⁷  None of these patients were clinically diagnosed with a stroke, yet 42% experienced postoperative cognitive dysfunction in the first year after surgery. Patients with occult stroke were more likely to be diagnosed with delirium within the first 3 postoperative days and more likely to experience a subsequent stroke or transient ischemic attack than patients without new diffusion-weighted magnetic resonance imaging lesions.

Cognitive and neuropsychologic assessments are the mainstay for diagnosing postoperative cognitive dysfunction in the research setting, and additional evidence of clinical impact is essential for the diagnosis of neurocognitive disorder. However, as the definitive standard, neuropsychologic tests, are not suitable for the application in the postoperative setting, more than 60% of postoperative cognitive dysfunction and delirium cases remain undiagnosed.^19,20  Decreased performance in specific neurocognitive domains after surgery is a downstream effect of damage to the underlying neural architecture. Thus, there is growing interest in direct postoperative assessment of neurocognitive function through advanced neuroimaging.

Connectomics has clinical utility in some perioperative contexts. Notable changes in brain network topology based on diffusion tensor imaging data were identified in a population of infants undergoing surgical repair of dextrotransposition of the great vessels, and these changes were associated with subsequent reductions in academic achievement, learning, and memory during adolescence.¹¹⁶  Connectomics has also been used to study cardiovascular disease more broadly in the context of metabolic syndrome and type 2 diabetes.^117,118  In patients with type 2 diabetes, functional magnetic resonance imaging revealed decreased activity in the parietal and temporal lobes, which correlated with decreases in neurocognitive function.¹¹⁹  In another study, functional magnetic resonance imaging changes were noted before structural changes, indicating that developing characteristic functional magnetic resonance imaging signatures might enable improved and earlier diagnosis of cognitive dysfunction.¹¹⁸  However, replicable brain signatures of postoperative cognitive dysfunction and their association with (and the ultimate impact of) magnetic resonance imaging−defined brain infarction have not yet been established.

A 2019 study identified white matter structural features that underly cognitive and emotional function in healthy adults using diffusion magnetic resonance imaging (fig. 7).⁷⁷  An assessment of cerebral tissue with diffusion kurtosis imaging revealed that emotional dysregulation was associated with increased diffusivity and reduced cellular complexity in specific regions of the temporal and parietal lobes, as well as the adjoining white matter tracts. Emotional dysregulation is a factor contributing to morbidity in postoperative cognitive dysfunction,¹²⁰  and using diffusion kurtosis imaging in the context of perioperative medicine will likely provide novel insights into the structural basis of this disease. For example, diffusion kurtosis imaging has previously revealed sensitive and specific white matter changes underlying depression, concussion, and motor neuron disease.^121–123  Such understanding, and the development of replicable, robust, and objective markers of postoperative cognitive dysfunction, will help identify at-risk individuals, facilitate the development of preventive therapies, enable early intervention, and overcome the insensitivity and subjectivity of the current definitive standard psychometric testing.

Neurologic injury after noncardiac surgery is associated with marked morbidity and mortality and is responsible for an immense global healthcare burden. Research and expert consensus statements continue to highlight the desire for improved hemodynamic optimization during surgery, but the role of this strategy to improve outcome is uncertain. Novel neuroimaging modalities have the potential to revolutionize perioperative care by providing real-time, noninvasive measurements of end-organ perfusion with potential for early detection of insult and prevention of perioperative injury. However, advanced neuroimaging methods have predominantly been developed and evaluated in critical care settings, particularly in the context of neurotrauma, and there is a paucity of studies that assess application of these methods in the perioperative setting. Further research using novel neuroimaging techniques and derived biomarkers in perioperative practice is essential to outline both the pathophysiology of organ injury and the utility of brain monitoring–directed therapy.

Support was provided solely from institutional and/or departmental sources.

The authors declare no competing interests.