Near-Infrared Spectroscopy: A Free-Living Neuroscience Tool to Better Understand Diabetes and Obesity

The human brain is the least accessible of all organs and attempts to study it in vivo rely predominantly on neuroimaging. Functional near-infrared spectroscopy (fNIRS) allows for the study of cortical neural activity in a non-invasive manner that may resemble free-living conditions. Moreover, compared to other neuroimaging tools, fNIRS is less expensive, it does not require the use of ionizing radiation, and can be applied to all study populations (patients suffering from claustrophobia, or neonates). In this narrative review, we provide an overview of the available research performed using fNIRS in patients with diabetes and obesity. The few studies conducted to date have presented controversial results regarding patients with diabetes, some reporting a greater hemodynamic response and others reporting a reduced hemodynamic response compared to the controls, with an unclear distinction between types 1 and 2. Subjects with obesity or a binge eating disorder have reduced prefrontal activation in response to inhibitory food or non-food stimuli; however, following an intervention, such as cognitive treatment, prefrontal activation is restored. Moreover, we discuss the potential of future applications of fNIRS for a better understanding of cortical neural activity in the context of metabolic disorders.


Introduction
The adoption of a westernized lifestyle, characterized by being sedentary, coupled with the consumption of highly palatable and energy-dense foods has led to increased prevalence and incidence rates of obesity and type-2 diabetes (T2DM) in recent decades. To date, more than 1 billion people worldwide have obesity [1] and 537 million people are affected by T2DM [2].
While the pathophysiology of obesity and T2DM is not fully elucidated, the recent research shows that, in humans, the brain may directly control whole body metabolic homeostasis [3][4][5][6][7]. However, the human brain remains the least accessible of all the organs. Thus, our potential in studying it in vivo relies predominantly on neuroimaging. Positron emission tomography (PET), magnetic resonance imaging (MRI), magnetic resonance spectroscopy (MRS), functional MRI (fMRI), and magnetoencephalography have been used to study brain metabolism in the context of metabolic diseases [8][9][10][11][12]. For instance, fMRI studies have demonstrated altered connectivity patterns in patients with T2DM and in people with obesity. On the one hand, the decreased connectivity in patients with T2DM is associated with a wide range of cognitive impairments [13]; on the other hand, the dysfunctional connectivity in people with obesity is suggestive of increased reward sensitivity and decreased interoceptive awareness [14]. However, the imaging modalities

NIRS Concept-How It Works and Clinical Applications
Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical functional imaging system that is based on a near-infrared (NIR) optical window (700-900 nm), which is almost transparent to the skin, bone, and brain tissue [15]. A brain fNIRS signal is generated by shining NIR light (650-950 nm) into the head through the scalp and can reach the surface of the cerebral cortex. In its way, through the layers of the head, NIR light is subjected to "absorption" and "scattering". Scattering leads to light attenuation and is much more frequent than absorption; therefore, placing a light detector at a certain distance from the NIR light allows us to register the backscattered light and measure the changes in light attenuation. Oxygenated (OxyHb) and deoxygenated (DeoxyHb) hemoglobin are chromophores that absorb near-infrared light, and most importantly, the absorption coefficients of OxyHb and DeoxyHb are different [16]. Therefore, the scattered-back light after absorption can provide information about the amount of chromophores in that region, which can be evaluated by the modified Beer-Lambert law (mBLL). From the difference between the intensity emitting and backscattered NIR light, the relative change in the oxygen concentration is assessed. In the hardware section of fNIR devices, the infrared light-emitter diode and detectors are generally placed 3-4 cm apart, as shown in Figure 1, oriented in a defined manner in an optode band, and it allows a 2 cm penetration depth. There are three different NIRS systems: continuous-wave NIRS is the oldest and most commonly used for its suitability and cost, despite its small light-penetration depth and the impossibility to separate light attenuation from absorption and scattering; time domain NIRS is less easily transportable, but has greater accuracy and spatial resolution, as light attenuation is measured by special cameras or single-photon counters based on their arrival time; and frequency domain NIRS uses a laser diode to emit light and it measures the attenuation, phase shift, and modulation depth of the light with respect to the system's incident light [17]. When a brain area is active and involved in the execution of a certain task, the cerebral blood flow (CBF) increases to satisfy the metabolic demand of the brain. This increase in local blood flow in response to neuronal activation is called "functional hyperemia" and results from neurovascular coupling, mediated primarily by the vasoactive metabolites of arachidonic acid, nitric oxide, K+, and other mediators whose properties are partially still unknown [18]. The amount of oxygen that reaches the activated brain region is higher than the rate at which oxygen is consumed, leading to an increase in OxyHb and decrease in DeoxyHb. This is called "hemodynamic response" (HDR) and can be measured through fNIRS at multiple locations of the cerebral cortex. The HDR reaches a peak a few seconds after the stimulus onset and returns to its baseline with a certain delay (i.e., these can vary across different brain regions, task types and design, and participants' age). The stimulus events are generally presented in block designs that consist of recurring blocks of tasks and rest periods; this configuration lets the HDR increase and return to the baseline level and has the highest signal-to-noise ratio, statistical power, and maximal time efficiency. By the early 1990s, fNIRS recordings demonstrated the capability of this technique to measure hemodynamic changes in the brain in response to functional activation tasks (e.g., flashing checkerboard at different contrast levels to study visual cortical activation or the Stroop test to study frontal, temporal, and anterior cingulate cortexes) both in adults and infants [19,20]. The development of more complex configurations made possible its use in cognitive experiments, allowing the study of functional brain activity and higher cognitive functions. In recent years, fNIRS found its application in functional neuroimaging techniques, which already included EEG/MEG, fMRI, and PET. The major advantage of fNIRS compared with EEG is that it is less susceptible to moving artefacts and offers a better spatial resolution, thus allowing for the localization of brain responses to specific cortical regions; however, EEG has a better temporal resolution (0.001 s) [19]. When a brain area is active and involved in the execution of a certain task, the cerebral blood flow (CBF) increases to satisfy the metabolic demand of the brain. This increase in local blood flow in response to neuronal activation is called "functional hyperemia" and results from neurovascular coupling, mediated primarily by the vasoactive metabolites of arachidonic acid, nitric oxide, K+, and other mediators whose properties are partially still unknown [18]. The amount of oxygen that reaches the activated brain region is higher than the rate at which oxygen is consumed, leading to an increase in OxyHb and decrease in DeoxyHb. This is called "hemodynamic response" (HDR) and can be measured through fNIRS at multiple locations of the cerebral cortex. The HDR reaches a peak a few seconds after the stimulus onset and returns to its baseline with a certain delay (i.e., these can vary across different brain regions, task types and design, and participants' age). The stimulus events are generally presented in block designs that consist of recurring blocks of tasks and rest periods; this configuration lets the HDR increase and return to the baseline level and has the highest signal-to-noise ratio, statistical power, and maximal time efficiency. By the early 1990s, fNIRS recordings demonstrated the capability of this technique to measure hemodynamic changes in the brain in response to functional activation tasks (e.g., flashing checkerboard at different contrast levels to study visual cortical activation or the Stroop test to study frontal, temporal, and anterior cingulate cortexes) both in adults and infants [19,20]. The development of more complex configurations made possible its use in cognitive experiments, allowing the study of functional brain activity and higher cognitive functions. In recent years, fNIRS found its application in functional neuroimaging techniques, which already included EEG/MEG, fMRI, and PET. The major advantage of fNIRS compared with EEG is that it is less susceptible to moving artefacts and offers a better spatial resolution, thus allowing for the localization of brain responses to specific cortical regions; however, EEG has a better temporal resolution (0.001 s) [19]. In addition, compared with fMRI, fNIRS has a better temporal resolution (0.02 s) and high temporal sampling rate (50 Hz), is silent, and can measure both Oxy-and DeoxyHb chromophores providing a more complete measure of the hemodynamic response; nevertheless, fMRI has a spatial resolution of 1 mm, which is better than fNIRS that has a spatial resolution of 1 cm [20]. Importantly, the comfort and absence of physical constraints on the participant allow for fNIRS's use in a variety of settings, including infants, children, or the elderly, for long recording sessions, and resembles free-living conditions. As the optical components do not interfere with electromagnetic fields, fNIRS is ideal for multimodal imaging, for example, it can be combined with fMRI or EEG [21,22] to allow for the gathering of more complete information related to neurovascular coupling.
The fNIRS modality is being used in many areas of research, ranging from muscle physiology evaluations during exercise or pathologic conditions, such as metabolic myopathies and muscular dystrophies, chronic heart failure, and peripheral arterial disease in diabetes [21]. One of the most important and emerging clinical applications of fNIRS technology is aimed at the study of brain tissue and it has provided applications in several areas, including cerebral oximetry monitoring for adult patients undergoing cardiovascular surgery [22], neonatology to assess cerebral autoregulation in preterm infants, and the management of newborns with hypoxic ischemic encephalopathy. Moreover, fNIRS technology is also increasingly used in several neuroscience areas, including neurodegenerative diseases, such as Alzheimer's disease [23,24], schizophrenia [25], addiction [26], dyslexia [27], attention-deficit/hyperactivity disorder [28,29], stroke rehabilitation [30], autism [31], depression [32], epilepsy [33], and migraine [34]. Based on the characteristics of the methodology underlying fNIRS and brain functional information provided, fNIRS does not contribute directly to diagnoses; rather, it is an assistive device that reads the functional activities of the brain and indirectly informs us regarding abnormalities in brain functionality. This implies that, in clinical practice, neurosurgeons or neurologists utilize the feature of this modality for anesthetic-depth monitoring to confirm that the patient is in a deeply sedated state. In some other mentioned applications, such as Alzheimer's, schizophrenia, dyslexia, addiction, ADHD, epilepsy, depression, etc., fNIRS is used as a predictive modality that discriminates against the functional activity of hemodynamics with some behavioral tests.

The Complex Pathophysiology of Diabetes-Induced Alterations in the CNS
Diabetes, particularly when poorly controlled, can have significant effects on the structures and functions of the CNS leading to an increased risk for developing cognitive dysfunction and dementia [35]. Both type 1 diabetes (T1DM) and T2DM share the hyperglycemia phenotype that is known to affect the human brain and, in 2006, Mijnhout and colleagues proposed the term "diabetes-associated cognitive decline" [36]. The definition of diabetes-associated cognitive decline, however, has not yet been unanimously agreed upon [37] as the impact of diabetes on the CNS is multifaceted and can affect various aspects of neurological functions, including the alteration in structural integrity, networks connectivity and energy metabolism.
Cognitive alterations may differ between T1DM and T2DM, likely due to a relatively higher prevalence of other comorbidities in T2DM patients, such as obesity, hypertension, and lipid disorders. Indeed, subjects with T2DM presented greater neurodegeneration processes and a 26% increase in brain aging as compared to subjects without glucose tolerance abnormalities [38]. Subjects with T1DM presented lower-level performances on tests for attention, speed of information processing, and executive function [37] as compared to the control group. These alterations translate in mental slowing that is thought to be the fundamental cognitive domain impaired in T1DM, affecting both intelligence and psychomotor speed. More importantly, executive function, the components of which are working memory and response inhibition, is impaired in both adults and children with T1DM. In subjects with T2DM, the main altered domains are memory, informationprocessing speed, and executive functions [39,40], which are strictly dependent on brain cortical function. Exploring the impact of diabetes on cognitive function is extremely important for clinical reasons as cognitive dysfunction affects different behaviors that, in turn, impact on diabetes self-care, including reduced compliance in monitoring glucose, taking medication, following new instructions for diabetes management, and reducing the risk of hypoglycemia [41].
A key mechanism behind the impact of diabetes on CNS is the hyperglycemia-related damage to the blood vessels supplying the CNS, resulting in a condition known as cerebrovascular disease. Vascular damage can manifest as the thickening, narrowing, or occlusion of the blood vessels, ultimately leading to reduced blood flow, oxygen supply, neurovascular coupling disruption, brain energy metabolism alterations, and an increased risk of ischemic events in the CNS [42][43][44]. In recent years, various neuroimaging techniques have been employed to study the impact of diabetes on brain structure and function and have also been used to define the structural and functional correlates of cognitive dysfunction in diabetes providing insights into the mechanisms underlying the CNS complications of the disease [45]. In particular, the main brain structural abnormalities that have been associated with the deterioration of cognitive function are the atrophy of the cerebral gray and white matter that can be focal or generalized [38], white matter hyperintensities [46], (lacunar) infarcts [47], and microbleeds [48]. Cerebral blood flow (CBF) is responsible for the delivery of nutrients to the brain [49] and is correlated to brain activity; under normal conditions there is a coupling between metabolically active regions and CBF [50]. Studies have observed that reduced total cerebral perfusion was associated with impaired cognitive function [51]. Moreover, a recent systematic review and meta-analysis of arterial spin-labeling studies conducted on subjects with T2DM demonstrated decreased CBF in the frontal, occipital, and parietal lobes, which are involved in cognitive domains mainly altered in subjects with diabetes [52]. Moreover, the alteration in brain glucose metabolism was demonstrated in studies employing brain [ 18 F]FDG-PET. Brain glucose hypermetabolism (during conditions of insulin clamp [53]) appears as an early trait that characterizes subjects with obesity [54] and subjects at an increased risk of obesity [55] or prediabetes [56]. However, diabetes is also associated with whole brain glucose hypometabolism in multiple brain regions [57]. Taken together, these data may suggest that brain hypermetabolism may serve as a transient compensatory reaction to the initial neurodegenerative insult; however, it is progressively replaced by hypometabolism, as tissue loss becomes more severe in the chronic situation. Definitively, an efficient and healthy cognitive function is based on the correct functioning of the neurovascular unit, which presupposes an integrity cerebral cortex structure, a flow of cerebral blood for oxygen and nutrients supply, including glucose, and a correct glucose uptake.
In addition to cognitive dysfunction, individuals with T2DM are at a greater risk of developing dementia, in particular vascular dementia and neurodegenerative diseases, such as Alzheimer's and Parkinson's disease [35]. Rouch et al. [58] showed in their study on 375 older ambulatory subjects with a mild cognitive impairment that the conversion to dementia was only associated with increased arterial stiffness and not with intima-media thickness, carotid plaques, or carotid artery diameter. Recent computational simulations using a detailed multicompartmental neurovascular model [59] provided insights into how vessel stiffness determines hemodynamic oscillatory peaks in neurovascular tissue. In addition to increasing oxygen availability at sites located away from small vessels [60], these vessel oscillations support waste clearance within the brain parenchyma, specifically convective bulk flow drainage along the basement membrane of capillaries and arterial walls [61]. Therefore, changes in the blood vessel pulsatility in T2DM can impair the convective bulk flow drainage that helps prevent the accumulation of neurotoxic waste proteins associated with neurodegenerative diseases. Moreover, small-vessel oscillatory dysfunction can reduce nutrient availability at sites located at a distance from small vessels [60], where reduced glucose availability in the CNS can directly trigger behavioral deficits by promoting amyloid-beta and tau neuropathology and synaptic dysfunction. A low-frequency Fahraeus-Lindqvist-driven (not blood pressure-driven) oscillation in the small vessels is directly involved in the brain nutrient supply and can be considered a marker of vascular dysfunction.

fNIRS in Diabetes: A Promising Tool to Evaluate Central Nervous System Changes in Diabetes
fNIRS is a promising technique capable of exploring, albeit indirectly, the functioning of the neurovascular unit of the cerebral cortex, especially in the areas/lobes that underlie the main cognitive domains altered in diabetes. In a study conducted on T1DM patients, the fNIRS of the visual cortical areas was employed to explore the HDR that occurred as a reflex consequence to increased neuronal activity within the primary visual cortex in response to a standardized visual stimulation. Individuals with T1DM had a larger HDR compared to those with T2DM, and a linear relationship between HbA 1c level and HDR was observed, suggesting an autonomic nervous system dysfunction, either by the overaction of the sympathetic or under-action of the parasympathetic branches [62]. In a pilot study conducted on young adolescents with T1DM, the executive function was assessed through fNIRS during a Go/No-Go response-inhibition task. Individuals with T1DM had a similar performance task as compared to the healthy controls, but higher activations in the frontoparietal network, including the bilateral supramarginal gyri and bilateral rostrolateral prefrontal cortices. The activations in these regions were positively correlated with fewer parent-reported rule-breaking behaviors suggesting a link between this brain network and better self-control. These findings are consistent with a large fMRI study of children with T1DM based on different groups of participants [63]. Executive function is also impaired in older adults with T2DM, as it was demonstrated by the findings of greater gait variability and postural instability during walking under dual-tasking conditions, irrespective of the presence of diabetic polyneuropathy [64]. Although the impact of T2DM on CNS sensorimotor regions is understudied, there is evidence of the atrophy of the primary cortex (M1), secondary motor cortex (M2), and primary somatosensory cortex (S1), and also white matter projections between the sensorimotor cortices and subcortical structures are thought to be degenerated [65]. Van Harten et al. found a correlation between periventricular white matter hyperintensity in an MRI and poor motor speed that was also independently associated with the duration of T2DM [66]. With respect to these acknowledgements, fNIRS studies found a decline in OxyHb in the sensory and motor cortexes with simultaneous deficits in manual motor tasks in post-menopausal women with T2DM and in the prefrontal cortex (PFC) of diabetic older adults during attention-demanding locomotion tasks suggesting the role of PFC in the control of walking, notably under attention-demanding dual-task conditions [67]. These results may provide the evidence of a cortical contribution to motor dysfunction that is notoriously considered the result of peripheral nerve damage [68].
Similar to T1DM, subjects with T2DM are characterized by visuospatial dysfunctions, as demonstrated by a study exploring the impact of a pattern-reversal checkerboard stimulation on the primary visual cortex, which confirms an alteration in HDR as compared to the healthy controls [62]. These results obtained with fNIRS support the findings of our previous studies demonstrating in individuals with obesity and prediabetes an abnormal visual cortex plasticity measured by the effect of short-term monocular deprivation on binocular rivalry dynamics [69,70] and an fMRI study showing a reduction in neuronal activity in the lingual gyrus (visual cortex) that was correlated with poor performance on visuospatial tests [71].
Older T2DM individuals experience mild cognitive impairment, specifically in the domain of recall/working memory, and it may be exacerbated in older adult females, who are at the highest risk of cardiovascular decline due to diabetes. In an fNIRS study exploring OxyHb and DeoxyHb during memory-based tasks in a cross-sectional sample of postmenopausal women with T2DM, a deficit in working memory accuracy was associated with differences in OxyHb responses, altered PFC activity magnitudes, and increased functional cortical activity across the region of interests compared to the controls. These data indicate a shift of OxyHb in cortical activity patterns with memory deficits in postmenopausal T2DM, representing a novel diabetes-specific finding that is unlikely to be detected by fMRI [72]. This underscores the value of using non-MRI-based neuroimaging techniques to evaluate cortical hemodynamic function to detect early mild cognitive impairment. Recently, it was demonstrated that, in individuals with T2DM, deficits in working memory and reaction time might be improved by integrated yoga practice as compared to the control, and the improvements were associated with higher oxygenation in dorsolateral and ventrolateral PFC regions [73].
Elderly individuals with T2DM demonstrated a drop in vascular reactivity during Mini-Cog with a three-item recall test in the PFC, as compared to age-matched controls [74], suggesting that fNIRS captures cerebrovascular reactivity to cognitive load and it may provide a biomarker for cerebrovascular dysfunction in T2DM. Experimental studies have proven the neuroprotective effect of GLP-1 receptor agonists by ameliorating cognitive impairment in subjects with T2DM [75][76][77]. A recent study tried to clarify if this effect was due to a direct action on the central nervous system or the consequence of better metabolic control. Li et al. demonstrated that treatment with Liraglutide for 12 weeks improved cognitive function, as revealed by the increased brain activation assessed with fNIRS in cortex regions related to better cognitive performance, regardless of changes in blood pressure, glycemia, and body weight in patients with type-2 diabetes compared with regular hypoglycemic treatment [78]. These findings are congruent with the knowledge that GLP-1 and its receptors (GLP-1R) are also expressed in the brain, especially in the hippocampus (crucial for learning and memory), and that mice lacking GLP-1R present impaired associative contextual learning that can be reversed by hippocampal GLP-1R somatic cell gene transfer [79]. In conclusion, individuals affected by T1DM or T2DM have neural slowing, increased cortical atrophy, white matter lesions, and modified cerebral perfusion, which impair brain function differently, even if with some similarities. Certainly, further studies are needed to better understand the relationship between diabetes mellitus, cerebrovascular pathophysiology, and their implications on neurological function. The studies regarding the application of fNIRS studies on patients with diabetes are summarized in Table 1.

Studies Using Brain NIRS to Study Inhibitory Control in Patients with Obesity or Binge Eating Disorder
In westernized societies, we live within an "obesogenic" environment where our senses (visual, or olfactory) are continuously exposed to palatable foods with a high caloric content. In such an environment, maintaining a healthy diet and avoiding the overconsumption of unhealthy foods requires cognitive control over our eating behavior. Previous fMRI studies have shown that frontal and prefrontal regions are activated in volitional appetite control, and this activation is blunted in patients with obesity [80]. These data suggest that dysfunctional frontal circuitry involved in inhibitory control may contribute to the wrong eating habits in patients with obesity. An excessive activation of the brain's reward circuitry from palatable foods has also been described in patients with obesity [81].
Since the activity of the brain cortex and, in particular, the prefrontal cortex can be assessed with NIRS, studies employing NIRS have been used in studying the inhibitory control in patients with obesity and healthy lean controls. In this context, BED has also been studied, since subjects with BED episodically exhibit a loss of control of what they are consuming, leading to the consumption of a high number of calories. In both young and middle-aged overweight subjects, obesity, or BED, it has been shown that the activation of the PFC to food stimuli or the Stroop test is attenuated compared to healthy lean controls [82,83].
Veit et al. studied patients with BED and healthy controls at baseline and 3 months following a cognitive intervention program. The subjects were studied with NIRS during a food go/no-go task. In particular, subjects were shown healthy and un-healthy foods; in the first experiment, they were instructed to "go" for the healthy foods and "no-go" for the unhealthy foods, and in the second experiment, they were instructed to perform the opposite (i.e., "no-go" for healthy foods and "go" for unhealthy foods). Patients with BED had decreased right prefrontal cortex activation compared to the healthy subjects and, following impulsivity-focused cognitive treatment, they increased their right prefrontal activity during inhibition. Moreover, this increase in prefrontal activity was associated with a reduction in trait impulsivity [84].
Xu and colleagues performed a prospective weight loss study with a combination of exercise training and food restriction on 31 overweight or obese adolescents and young adults. The Stroop test was performed at baseline in conjunction with NIRS and following a 4-week fitness investigation. The authors reported a positive correlation between the hemodynamic interference due to the Stroop effect in the bilateral dorsolateral prefrontal cortex (DLPFC), left ventrolateral PFC (l-VLPFC), and left frontopolar area (l-FPA), and the amount of weight loss achieved [85]. The findings, which are in line with the previous fMRI studies [86], suggest that increased brain activation in the regions involved in executive function is associated with better weight loss outcomes.
Finally, Huang et al. studied the effect of age on brain activation during the Stroop test in 38 children, adolescents, and young adults with obesity aged 9 to 25 years old. They found that Stroop interference was not increased as a function of age [87]. This study's results contrast with the findings of a previous study on children, adolescents, and young adults aged from 7 to 29 years old with a normal weight, where brain activation during the Stroop test was shown to be positively associated with age in the DLPFC [88]. Thus, as the authors discussed, these findings could be interpreted as evidence that, in the context of obesity, the neural mechanisms that are needed to suppress the Stroop effect are not as developed as in normal-weight individuals [87].
In conclusion, most NIRS studies assessed prefrontal cortex activation in response to inhibitory food or non-food stimuli, with subjects with obesity or BED having reduced prefrontal activation in response to inhibitory food or non-food stimuli. Interestingly, following an intervention, such as cognitive treatment, prefrontal activation is restored. These findings are in line with recent intervention studies where non-invasive brain stimulation with transcranial direct-current stimulation (tDCS) over the right DLPFC induced a decrease in self-reported cravings, probably by strengthening inhibitory control [89], and tDCS targeted to bilaterally stimulate the PFC and insula decreased impulsivity and consequently BMI in patients with obesity [90]. The original fNIRS studies treating these themes are summarized in Table 2.

Future Applications-Adding Continuous Glucose Monitoring and Glucose Tolerance Tests to NIRS
Continuous glucose monitoring (CGM) allows for the assessment of intraday glycemic variability. On the other hand, the assessment of glycemic variability is known to be important for the study of brain function. In fact, it was demonstrated that both hyperglycemia and hypoglycemia can be associated with patient experiences of physical, affective, and cognitive symptoms, as well as cognitive-motor disruptions [91]. Of note, even mild or transient hypoglycemia can reduce the mental efficiency; although, the impact of this effect depends on the task the patient is dealing with [91]. In addition, both hypoglycemia and hyperglycemia have not only acute, but also chronic, effects on patients with T1DM and T2DM [91].
With regards to cognitive function, it has been demonstrated that there is a relationship between intraday glycemic variability (calculated by a common CGM metric named mean amplitude of glycemic excursions, MAGEs [92]) and cognitive performance, as assessed by cognitive tests, such as a composite score of executive and attention functioning and the mini mental status examination (MMSE) [93]. Interestingly, the association between MAGE and the impairment in cognitive function was independent from age, sex, body mass index, waist-to-hip ratio, drug intake, physical activity, mean arterial blood pressure, glycated hemoglobin, as well as the absolute glycemic values, both in fasting and post-prandial states [93]. It has also been hypothesized that glycemic variability can be responsible for cerebrovascular damage [94]. In fact, it has been shown that, in people with metabolic syndrome (although without T2DM), glycemic variability is associated with reduced cerebral vasomotor reactivity, and the association is independent from daily mean systolic and diastolic blood pressure levels [94]. Glycemic variability was also found to be associated with several stroke risk factors and with poorer short-term prognosis [95]. Furthermore, glycemic variability was found to be associated with cerebral metabolic distress (defined as lactate/pyruvate ratio >40), as well as hospital mortality, after subarachnoid hemorrhage [96]. On the other hand, a recent study emphasized that it was still partially unclear whether the association between glycemic variability and brain dysfunction following subarachnoid hemorrhage was due to direct deleterious effects induced by glucose dysregulation or if hyperglycemia was an epiphenomenon related to initial bleeding severity [97]. The study concluded that further research was needed to understand the most appropriate and informative timing for glucose monitoring and how to combine different glucose metrics with markers of brain injury [97]. The importance of selecting the most appropriate metrics for CGM data analysis for the study of brain function was also indicated in another recent study, which showed that, in T2DM patients, the glucose coefficient of variation (CV) and time below range (TBR) were not associated with any of the investigated aspects of cognitive function [98]. In contrast, some hyperglycemia metrics and the time in range (TIR) were associated with cognitive function, especially with executive function and working memory [98]. The reported findings indicate that there is certainly an association between glycemic variability and several aspects of brain function; however, there are still issues to be elucidated, including the identification of the most appropriate CGM metrics and brain function markers to shed further light onto such an association. In fact, further studies, even of a longitudinal type, are ongoing at present [99].
It also has to be noted that some studies also investigated the possible molecular mechanisms determining the association between glycemic variability and altered brain function. In fact, one study investigated the possible mechanistic basis for diabetes-induced cerebrovascular damage, due to both hyperglycemia and high glycemic fluctuations [100]. It was shown that carbonyl stress, indicated by the formation of occluding-methylglyoxal carbonyls, was a possible mechanism of endothelial barrier dysfunction [100]. The study also demonstrated that hyperglycemia compromised the elimination of methylglyoxal by the glyoxalase pathway and, hence, increased the glycating potential of methylglyoxal, thus exacerbating the dysfunction of the cerebral microvasculature [100]. Another study evaluated the responses of neuronal cells to different glycemic exposures and investigated the role of mitochondrial uncoupling proteins (UCPs) in regulating such responses, with a focus on one such protein (UPC2) [101]. Indeed, UCPs are mitochondrial anion carriers with a crucial contribution to regulating mitochondrial homeostasis, and have a demonstrated role in physiological and pathological adaptations of the brain [101]. The study findings suggested that UCP2 was at the core of neuronal cell protection or adaptation against the effects mediated by glycemic variability, and that other isoforms of neuronal UCPs can be upregulated to compensate for the possible inhibition of UCP2 activity [101].
The studies summarized above provide evidence of the clear association between glycemic variability and brain function impairment, while, at the same time, indicating that further research is still necessary to clarify some issues, as previously reported. In our opinion, one of the possible future directions for such new studies is the integration of CGM data with modern neuroimaging techniques, such as fNIRS. Indeed, among the previous studies conducted on glycemic variability and brain function, those that specifically exploit neuroimaging data are rare. One study investigating the association between glycemic variability at multiple time scales and brain volumes exploited MRI [71]. The anatomical MRI data were segmented to calculate regional gray and white matter and cerebrospinal fluid volumes in the main anatomical lobes and their sub-regions [71]. MRI was also used to assess the association between glycemic variability and brain damage in hypoglycemia neonates [102]. Another study, assessing the association between glycemic variability and the presence of cerebral lacunes (sign of lacunar infarction potentially leading to dementia or disability), retrospectively analyzed MRI data to identify such lacunes [103]. FMRI was used in a study investigating the associations between glycemic variability, hypothalamic function, and diet during the first 18 months of life [104]. With regard to studies integrating glycemic variability data with NIRS data, we did not identify any study, thus indicating that the combination of CGM and NIRS would definitely be a new strategy.
In addition to the CGM data for the assessment of glycemic variability, in our opinion, NIRS imaging may also be conveniently integrated with glucometabolic information derived by a glucose (or meal) tolerance test, with special interest in the traditional oral glucose tolerance test (OGTT). In fact, some previous studies, though still few, performed OGTT while investigating some aspects of brain function. Specifically, most of these studies focused on the investigation of the brain-derived neurotrophic factor (BDNF), which is a member of the neurotrophin family and plays an important role in neural protection, synaptic activity, and endothelial survival. One study investigated whether circulating BDNF levels change during an OGTT and found that plasma levels of BDNF were sensitive to acute changes in both OGTT glucose and insulin levels [105]. In another study, it was shown that, after an oral glucose challenge, a lower serum BDNF response was associated with higher central pulse pressure [106]. Another study by the same research group of study [106] showed that lower serum BDNF levels during the OGTT were associated with high cardiovascular risk [107]. Another study in the field addressed a different topic, since it focused on multiple sclerosis (MS), which is a chronic neurologic condition predominantly affecting young people and, in fact, one of the major causes of disability in this population [108]. That study investigated the association of glucose metabolism, as assessed by an OGTT, with the level of MS progression and related degree of disability, and a strong association was found between impaired glucose metabolism and disability [108].
In summary, based on the reported previous studies, we claimed that the integration of fNIRS, CGM, and OGTT data would likely provide invaluable information and increase the scientific scenarios concerning the study of brain function. In fact, such a "triple integration" would be an extremely innovative approach.

Conclusions
In conclusion, fNIRS is a relatively cheap neuroimaging tool that allows for the study of cortical networks in conditions resembling free-living conditions. Since it does not employ the use of radiation, experiments can be performed in different occasions. Several alterations have already been identified in both patients with T1DM, T2DM, and patients with obesity. Further applications of fNIRS, for instance, in concomitance of CGM, or before and after interventions (weight loss, drug treatment, bariatric surgery, exercise), may improve our understanding of the interplay between systemic metabolism and cortical networks.