Monitoring Brain Activities Using fNIRS to Avoid Stroke

*Yun-Hsuan Chen and Mohamad Sawan*

## **Abstract**

Functional near-infrared spectroscopy (fNIRS) is an emerging wearable neuroimaging technique based on monitoring the hemodynamics of brain activity. First, the operation principle of fNIRS is described. This includes introducing the absorption spectra of the targeted molecule: the oxygenated and deoxygenated hemoglobin. Then, the optical path formed by emitters and detectors and the concentration of the molecules is determined using Beer-Lambert law. In the second part, the advantages of applying fNIRS are compared with other neuroimaging techniques, such as computed tomography and magnetic resonance imaging. The compared parameters include time and spatial resolution, immobility, etc. Next, the evolution of the fNIRS devices is shown. It includes the commercially available systems and the others under construction in academia. In the last section, the applications of fNIRS to avoid stroke are presented. The challenges of achieving good signal quality and high user comfort monitoring on stroke patients are discussed. Due to the wearable, user-friendly, and accessibility characteristics of fNIRS, it has the potential to be a complementary technique for real-time bedside monitoring of stroke patients. A stroke risk prediction system can be implemented to avoid stroke by combining the recorded fNIRS signals, routinely monitored physiological parameters, electronic health records, and machine learning models.

**Keywords:** neuroimaging, functional near-infrared spectroscopy (fNIRS), hemodynamics, monitoring, wearable devices, stroke

## **1. Introduction**

Studying the brain signals helps to gain knowledge of its function. Clinically, cortical signal recordings provide information for researchers to investigate the mechanisms of the cerebral diseases and help the clinicians identify the type of diseases, locate the lesions, and further prescribe the treatments and medications [1, 2]. Therefore, various neuroimaging techniques flourished this century [3]. These techniques are implemented in various applications these decades.

Neuroimaging techniques can be separated into invasive and non-invasive types. The former includes electrocorticography (ECoG), intracortical implants, stereoelectroencephalography (SEEG), spikes, and local field potential (LPF). The latter includes scalp encephalography (EEG), computed tomography (CT), functional magnetic resonance imaging (fMRI), magnetoencephalography (MEG), photoacoustic

imaging (PAI), functional near-infrared spectroscopy (fNIRS), and Positron emission tomography (PET). Invasive approaches encounter more clinical issues, such as infection, biocompatibility, etc. Non-invasive techniques, such as CT, fMRI, MEG, and PET, can achieve high temporal and spatial resolution. However, these tools are bulky and not easy to access, which is not appropriate for real-time monitoring [4]. Thus, lightweight, compact systems like EEG and fNIRS attract attention [5]. EEG records the electrical signals of many neurons, resulting in a low-spatial resolution. In comparison, fNIRS is an approach to overcome the limitations mentioned above [6].

fNIRS is often used to study brain activity during cognitive and motion tasks [7]. Clinically, fNIRS is applied to diagnose, monitor the progression of the diseases, predict the outcomes of the disorders, and track the effect of rehabilitation [8]. The brain function changes and the evolution of the brain diseases or the recovery process suggest the effectiveness of treatments or rehabilitation. Furthermore, the fNIRS biomarkers can be identified to distinguish the abnormalities and further stratify the severity of the diseases. With the development of various neuromodulation approaches, such as transcranial direct current stimulation (tDCS), transcranial magnetic stimulation (TMS), and ultrasound, combining fNIRS and these techniques to alter the brain activity, a closed-loop brain-machine/−computer interface can be achieved [9, 10]. These closed-loop systems can be used for abstinence of addiction, to reduce the symptoms of epileptic seizures, to release the migraine, and even to enhance attention [11, 12].

While fNIRS have been applied to various neural diseases, this chapter brings an overview of the application to stroke, with special attention to stroke prevention and prediction. Stroke is the second leading cause of death worldwide. Over 50% of stroke patients suffer from various levels of disability after the onset resulting in heavy family and social burdens [13]. However, it is reported in the countries' guidelines that 85% of stroke is preventable. A stroke happens when the blood flow in the brain is interrupted or significantly reduced. This is highly related to the changes in the amount of hemoglobin, which can be detected using fNIRS. This neuroimaging technique has been applied in various stroke cases, such as stroke prediction in people with specific diseases, monitoring during the treatment, stroke management in the perioperative period, and rehabilitation evaluation [14]. Since stroke is known as a chronic disease, investigations are carried out to predict the sign of the stroke onset by tracking the risk factors changes. With user-friendly, compact, and light-weighted characteristics, fNIRS is a promising tool to monitor the hemodynamic parameters continuously or in a high-frequency mode. Combining the real-time monitored physiological parameters and the electronic health records (EHRs), a prediction model can be established by introducing machine learning models [15].

In this chapter, a detailed introduction to the mechanism of neurovascular coupling (Section 2.1), the principles of fNIRS (Section 2.2), the advantages of fNIRS (Section 2.3), the evolutions of fNIRS devices, and the commercially available or self-developed devices (Section 2.4) are presented. Then, the potential of using fNIRS to avoid stroke is elaborated in Section 3. Section 4 is the conclusion.

## **2. Functional near-infrared spectroscopy (fNIRS)**

In this section, an overview of fNIRS, an emerging neuroimaging technique, is presented. The principle, the advantages, and the evolution of the fNIRS systems are discussed in these sections.

## **2.1 Neurovascular coupling**

Neurovascular coupling is a mechanism that the hemoglobin concentration variations result from brain activities. This is the phenomenon that an fNIRS can monitor. When the neurons are in the rest state, the concentrations of the surrounding oxygenated hemoglobin (OxyHb) and the deoxygenated hemoglobin (DeoxyHb) are in a particular state to support the normal metabolism (**Figure 1a**). When the neurons are activated, they need oxygen to support their activation (**Figure 1b**). Therefore, the blood flow of the vessels around the neurons increases, increasing OxyHb to provide additional oxygen to the neurons. Though the neurons consume the oxygen, converting the OxyHb to DeoxyHb during activation, the increased OxyHb is more than the OxyHb which is transformed. Therefore, an increase of OxyHb and a reduction of DeoxyHb are observed during this neuronal activation phase, called neurovascular coupling (**Figure 1c**).

## **2.2 Principles of fNIRS**

The incident lights at the wavelengths that can react the most with the target molecules are applied to monitor the concentration variation optically. The target molecules mean OxyHb and DeoxyHb, whose concentrations are altered due to brain activities. However, OxyHb and DeoxyHb are not the only molecules in the brain. Water is another molecule that accounts for a large percentage of the brain.

## **Figure 1.**

*Schematics of the neurovascular coupling: (a) At rest state, the concentration of oxygenated and deoxygenated hemoglobins (OxyHb and DeoxyHb) is at a certain metabolic state, (b) during neuronal activation, the blood flow, meaning the OxyHb increases to provide the needed oxygen to the activated neurons. While more OxyHb is supplied, more DeoxyHb is generated due to the consumption of the oxygen from the OxyHb molecules, and (c) the overview of the variation of OxyHb and DeoxyHb resulted from the neurovascular coupling when neurons are activated. The red cell is the oxygenated hemoglobin. White circles on the red cells are oxygen. When the oxygen is provided to the neurons, the red cell converts to the blue cell, called deoxygenated hemoglobin.*

The absorption spectra of these three molecules are shown in **Figure 2**. An optical window with the wavelength from 700 to 900 nm is an accepted range for the public to detect the concentration of OxyHb and DeoxyHb with limited interferences from the water molecules. To maximize the detection accuracy, fNIRS devices are often equipped with light emitters at 760 nm and 850 nm, which is the wavelength of maximum absorption factors of DeoxyHb and OxyHb, respectively.

After penetrating the cerebral tissues, the incident light is collected by detectors. The detected light is attenuated due to the scattering, absorption, and transmission effects of the interaction of the light and the medium (**Figure 3**). Beer-Lambert law defines the relationship between attenuation and the attenuating species' concentration and the optical path length [6]. For the near-infrared spectroscopy (NIRS) applied on the scalp, since the optical path between the emitter and the detector is not straight, a modified Beer-Lambert law with differential path length factor (DPF) as an additional parameter is applied. Besides, another parameter, G, representing photon loss due to light scattering, is added in the modified formula. Assuming DPF and d are known and remain constant, G remains constant, and the attenuation difference can determine the concentration difference of the attenuating species.

**Figure 2.**

*Absorption spectra of the molecules in the brain. fNIRS is an optical approach to monitor the concentration variation of HbOxy and HbDeoxy.*

#### **Figure 3.**

*Modified beer-Lambert law is applied to the recorded fNIRS signals to determine the concentration changes of HbOxy and HbDeoxy: I0: The intensity of the inserted light; I: The intensity of the detected light; a: Attenuation; ε: Attenuation factor; C: Concentration of attenuating species; d: Optical path length; G: Photon loss due to light scattering, which depends mainly on geometrical factors introduced; DPF: Differential path length factor.*

Since it is challenging to define the value G, most of the available NIRS devices can monitor the relative concentration, meaning the variation of concentration, not the absolute value of the concentration of the hemoglobins. This is the imaging principle of a continuous-wave NIRS (cw-NIRS).

To obtain the absolute values of hemoglobins, time-domain and frequency-domain modalities are designed, known as TD-NIRS and FD-NIRS. Diffused optical tomography (DOT) is an upgrade version of TD-NIRS to achieve a high spatial resolution and distinguish the signals from different depth layers under the scalp [16]. This NIRS approach being applied to study the functional changes of the brain cortex during various activities is later known as functional near-infrared spectroscopy, shortened as fNIRS. This shares the same idea as functional MRI (fMRI), to investigate the brain's activated region during certain tasks.

**Figure 4** presents an example of the location arrangement of fNIRS's optical electrodes (optodes), meaning the emitters and the detectors. A banana-shaped optical path is formed between each emitter and detector pair. The concentration changes of the hemoglobin passing through the path region can be determined using the modified Beer-Lambert law. The commonly used distance between an emitter and a detector is 3 cm, attaining the deepest sensing depth at 1-1.5 cm beneath the scalp. This shallow layer is mostly located with capillary vessels. Depending on the region of interest, the emitters and detectors can be placed alternately or in other patterns on the scalp. Illustrations in **Figure 4** show the optical path of a cw-fNIRS. If applying a TD-fNIRS, the photons that arrive at different times can be detected, giving the hemodynamic information of various depths of the tissue.

## **2.3 Comparison of fNIRS and other neuroimaging techniques**

Magnetic resonance imaging (MRI), computed tomography (CT), and magnetoencephalography (MEG) are the most common techniques to characterize the abnormalities in the brain (**Figure 5**) [17]. Though MRI and CT offer brain images with a high spatial resolution within the millimeter range, the temporal resolution is low, impeding their application in real-time monitoring. In addition, MRI and CT tests are rather expensive and time-consuming. The MRI, CT, and MEG tests equipment is bulky and should occupy a custom-designed room. Therefore, the patients need to be transported for every measurement. Moreover, the MRI measurements

#### **Figure 4.**

*Placement of fNIRS optodes on the scalp. The hemoglobin changes in the blood vessels located in the optical path formed by a pair of emitter and detector can be monitored.*

#### **Figure 5.**

*Comparison of various types of neuroimaging techniques: CT (computed tomography); EEG (electroencephalogram); fMRI (functional magnetic resonance imaging); fNIRS (functional near-infrared spectroscopy); and MEG (magnetoencephalography).*

cause loud noise and are sensitive to motions, reducing its usability in a certain group of patients, such as infants. Besides, there are other restrictions for receiving MRI measurements, such as for patients with a pacemaker or metallic implants due to the magnetic fields. CT has the additional disadvantage that patients are exposed to ionizing radiations. Due to the discussed disadvantages, there is no appropriate neuroimaging technique to real-time monitor the hemodynamic response in routine clinical practice.

fNIRS uses pairs of optodes, including a near-infrared light source and a detector, to monitor the concentration change of OxyHb, DeoxyHb, and cerebral blood flow resulting from the brain's hemodynamic activity [18]. Both EEG and fNIRS recording systems are non-invasive, flexible, low-cost, compact, light-weighted, small-sized, portable, undemanding set-up and can be used ambulatory. In addition, they both have less physical restriction and less motion artifact interference compared with functional MRI (fMRI) and CT. Moreover, they have no high intensity (>1 T) magnetic field or ionizing radiations compared with fMRI and CT. Regarding EEG, the spatial resolution is not increased since the recorded signal is generated by the bunch of neurons close to the electrode. The studies using fNIRS to monitor the variations of pathophysiological parameters are introduced in the following paragraphs.

## **2.4 Evolution of fNIRS devices**

The components forming an fNIRS system are the emitter, the detector, and the recording module. The emitters can be a laser, laser diode, or light-emitting diode (LED). The detectors can be photomultiplier tubes (PMTs), silicon photodiodes (SPDs), avalanche photodiodes (APDs), or a novel type called silicon photomultiplier (SiPM). SiPM overcomes the shortcomings of the bulky size of PMTs and the lower sensitivity to light resulting in a limited detection depth of the brain [19].

The evolution of fNIRS technologies in terms of the parameters related to signal quality and user-friendly is shown in **Table 1**. The early fNIRS was often equipped with fiber laser as emitter, therefore using mainline power resulting in a rather bulky system and low mobility. The subjects wearing the cap with optic fiber bundles need to stay around the device and experience movements restriction. A compact fNIRS system with LED powered by batteries is designed to become a wearable device to extend the application scenarios. The fiber laser fNIRS devices can be weighted up to 100 kg, while the LED devices' weight is around 500 g for the helmet, including the optodes.

The oximeter is the original device using the NIRS principle to detect the concentration variation of the hemoglobin [20]. The device is a clip to be placed on the finger or two patches attached to the forehead to detect the oxygen saturation of the blood (SpO2). These oximeters are single sites with emitters of often two wavelengths each, which are used to detect the OxyHb and DeoxyHb. The oximeters are considered simple prototypes of various emerging increasing complexity fNIRS devices. Increasing the number of optode channels is mandatory to simultaneously monitor the hemodynamic states of multiple areas of the cerebral cortex. The challenges of upgrading a single site to multiple sites fNIRS are applying the frequency encoding method to minimize the interferences from other optical paths. After overcoming this issue, the channel number of fNIRS boomed. It is noted that one single light emitter can create multiple channels if numerous detectors surround it. Using laser fibers, around one hundred channels can be reached. However, achieving these numerous channels requires heavy cables and emitters mounted on the cap. The weighty system causes discomfort. In addition, the relatively rigid cables limit the number of optodes can be placed on the scalp. In contrast, the channel number of fNIRS easily achieves one to two thousand channels if using LED as emitters. Thus, the hemodynamic conditions of the brain cortex's large region of interest can be imaged.

Due to the limitation of the light penetration, cw-fNIRS is only sensitive to the hemodynamic variation of the vessels at 1-2 cm beneath the scalp, around half of the commonly used distance between the light source and the detector, 3 cm.


#### **Table 1.**

*Evolution of the key parameters of the fNIRS devices.*

With TD-fNIRS, the OxyHb and DeoxyHb concentration changes at various depths of the cerebral tissue can be determined [21].

Only the hemodynamic states of light scattered spots can be determined with limited fNIRS channels. High density diffused optical tomography (HD-DOS) is designed to construct the 3D images of hemodynamics based on NIRS principles [16]. This HD-DOS contained various distances between the emitter and the detector, from 1 to 6 cm, to increase the spatial resolutions of a conventional cw-fNIRS.

As introduced in Section 2.3, every neuroimaging technique has its advantages and shortcomings. Therefore, applying multiple them simultaneously compensates for the shortcomings of each other and achieves a more comprehensive view of the brain activity. The fNIRS-EEG system can simultaneously record the neuron electrical signals and hemodynamic signals, which helps investigate the neurovascular coupling effect. However, fNIRS-fMRI obtains cross-validation for brain activity. The high-temporal-resolution fNIRS and the high-spatial-resolution fMRI are complementary.

Concerning the commercially available fNIRS devices, around 20 companies worldwide, such as NIRx, Artinis, ISS, SHIMADZU, g.tec, etc. Most of their products are cw-fNIRS; only one company, ISS Inc., introduced the multi-channel FD-fNIRS. Also, no company offers a TD-fNIRS yet. Both laser fiber and LED are used in these commercially available systems for the light emitters.

Regarding the emerging platforms in academia, research groups are focusing on gaining more flexibility when applying fNIRS in specific applications. Most of them focus on miniaturization and improving the temporal and spatial resolutions. A review paper summarized fNIRS hardware implanted in academia from 2012 to 2017 [22]. In recent 5 years, a 16-source and 16-detector fNIRS intended for epilepsy study was recently published [23]. An open-source, light-weighted (142 g) fNIRS is developed [24]. WearLight, a wearable fNIRS, is designed [25]. Growing research has been conducted to implement TD-fNIRS to compensate for the missing product in the market [26].

Furthermore, many multimodal fNIRS-EEG systems are designed to better explore both the electrical and hemodynamic signals of the neurovascular coupling [27]. The co-author of this chapter is one of the first few group to present the multimodal fNIRS-EEG system [28]. This prototype was further optimized and proved to be valid when monitoring signals from patients with epilepsy and stroke [29]. In addition, Mobile Modular Multimodal Biosignal Acquisition (M3BA architecture) is an architecture that can wirelessly transmit EEG, fNIRS, and accelerometer signals [30].

## **3. Applications of fNIRS on stroke**

The stroke-related applications using fNIRS include predictions, diagnosis, monitoring during treatments, outcome prediction, stroke management, and rehabilitations [14]. fNIRS monitoring during the treatments is the most challenging part since the protocol needs to be fine-tuned to reduce the impede on the planned surgeries and treatments. In Section 3.1, the studies that reported the fNIRS measurements during the treatments of stroke patients are summarized. In Section 3.2, the role of fNIRS on stroke risk prediction and prevention is discussed. The challenges of applying fNIRS to avoid stroke are discussed in Section 3.3.

### **3.1 fNIRS monitoring during treatments**

A stroke happens when the blood flow in the brain is interrupted or significantly reduced. Lack of oxygen and nutrients supply to the brain cells results in cell death. Ischemic stroke contributes to 85% of strokes occurs when the blood flow in cerebral vessels is either blocked by a blood clot or reduced by atherosclerosis. Removing the obstruction and restoring the blood flow of the affected area as early as possible reduces the impact of ischemia and improves the outcome of stroke.

Nowadays, intravenous thrombolysis therapy using tissue plasminogen activator (tPA) is the only medicine treatment approved by U.S. Food and Drug Administration for acute ischemic stroke [31]. tPA is an enzyme involved in clot breakdown. However, less than 50% of the patients receiving tPA therapy are successfully recanalized. At least 30% of them encounter the main complication, hemorrhage of tPA. Half of them show no response to the tPA therapy due to the complex conditions of the combination of characteristics of the blood clot and their health conditions [32].

Since another treatment named mechanical thrombectomy (MT) has a longer time window, a higher recanalization rate, fewer contraindications, and fewer complications than intravenous thrombolysis, it has become a more popular approach for treatment. MT improves functional outcome and prevents severe disability and mortality for those eligible for this treatment nowadays [33]. MT is a surgery to remove a blood clot from a blood vessel of patients with large vessel occlusion (LVO). Though the surgical approaches and devices of MT have been rapidly improved these years, 4–29% of the patients encounter complications after receiving MT [34]. Around 6% of patients have hemorrhage and 2-20% have reocclusion. Without immediate detection and proper treatments for these complications, the recanalization of MT treatment is in vain.

Real-time monitoring the hemodynamic status prior to, during, and after the treatments can help detect the abnormalities in time and alert the clinicians to take action. Continuous fNIRS monitoring of the cerebral hemodynamic variations of patients who underwent tPA therapy, known as thrombolysis, was reported in a few publications [35–37]. One research group suggested that the OxyHb and the total hemoglobin concentration (THC) of both ipsilesional and contralesional hemispheres increased during the first 2.5 h after tPA treatment [35]. While another group studied the relative change in regional O2 saturation (rSO2) and the interhemispheric rSO2 (IHΔrSO2) difference of 1 hour, 6 hours, 12 hours, 18 hours, and 24 hours after the tPA treatment are analyzed and compared with the National Institutes of Health Stroke Scale (NIHSS scores) or either MRI and CT images at the same time frame [36]. A protocol is proposed in a clinical trial to monitor the fNIRS data along with the blood pressure, heart rate, respiratory rate, rSO2, Glasgow coma scale and the NIHSS at every 15 minutes from the beginning of the thrombolysis till 120 minutes after the starting point [37].

On the other hand, recent and important studies on continuous fNIRS monitoring of the cerebral hemodynamic variations of patients receiving MT treatment are summarized in **Table 2**. Most studies start fNIRS monitoring prior to MT treatment. However, since the preparation of fNIRS monitoring can impede an urgent surgery, the number of measurements starting prior to the MT is limited.

Regarding the locations of optodes, the optodes are placed on the forehead of patients in most studies. While the optodes are placed on the region of interest (ROI), the area is close to the location of ischemic stroke and the opposite hemisphere in only one study. The measured parameters: OxyHb, DeoxyHb, HbT


*rSO2: regional cerebral tissue oxygenation; StO2: tissue oxygen saturation; ROI: region of interest; CAS: Carotid artery stenting; IHΔrSO2: interhemispheric rSO2 difference; DCS: Diffuse correlation spectroscopy; DOS: diffuse optical spectroscopy.*

*† : only 25% of the analyzed data includes recording before the MT.*

#### **Table 2.**

*Studies regarding fNIRS monitoring of acute stroke patients with mechanical thrombectomy recanalization or stenting.*

*Monitoring Brain Activities Using fNIRS to Avoid Stroke DOI: http://dx.doi.org/10.5772/intechopen.105461*

(the combination of the previous two parameters), and rSO2 (regional cerebral tissue oxygenation) are used to evaluate the efficacy of MT and monitor the hemodynamic states after the surgery. In some studies, the difference between the parameters of two hemispheres over time is characterized. It is found that the variations of the recorded parameters suggest the success of recanalization and are associated with the clinical outcome.

From the above examples, it is concluded that NIRS technique can provide information of real-time changes in hemodynamic response revealing the effects of thrombolysis and thrombectomy. Further applying this information to avoid stroke will be presented in Section 3.2.

### **3.2 Using fNIRS-EEG to predict and further avoid stroke**

Since multimodal fNIRS-EEG is becoming wearable devices with high accessibility for brain monitoring, this section will focus on combining these two techniques. The concept of applying multimodal fNIRS-EEG neuroimaging for stroke risk prediction is shown in **Figure 6**. The first stage is to identify the fNIRS-EEG biomarkers associated with the hemodynamic conditions of stroke onset zones. Studies are launched to monitor the physiological signals of a stroke patient before the treatments are carried out. The measurements continue along with the treatments to the recovering phase in the perioperative period. The variation of the biomarkers when the treatments succeed, the MT or the tPA, or when any complications occur is valuable to train a model for stroke risk prediction. Combining fNIRS and EEG data recordings with other physiological parameters and electronic health records (EHRs) to implement into a machine learning model, a prediction system can be obtained to predict either the onset of complications in stroke patients or on people with a high risk of stroke.

The first milestone to achieve the aim of stroke prediction is real-time tracking of the occurrence of complications after MT or tPA according to the parameters recorded from the multimodal fNIRS-EEG system and the routine clinical recordings.

#### **Figure 6.**

*Combining the fNIRS-EEG, other physiological parameters, EHRs, and ML model, a system can be established to predict the onset of the complications after stroke treatments or the onset of stroke cases in people with a high risk of stroke. DAR: The ratio of the power of the delta band to the power of alpha bands in EEG signals. ML: Machine learning. MT: Mechanical thrombectomy. rSO2: Regional cerebral tissue oxygenation.*

Investigating the mechanisms of the variation of the recorded parameters while combining other available physiological parameters over time distinguish the different impacts on brain activity caused by different complications. When understanding the variation trend of the recorded physiological parameters prior to and during the occurrence of complications after treatments (the yellow time range in **Figure 6**), these parameters are potential biomarkers as an alert for close attention and even further personalized medical or surgical treatments.

The conventional neuroimaging technologies and routine monitored physiological parameters have limitations for stroke management. CT or MRI examinations are impracticable for continuous bedside monitoring of post-treatment patients. Routine measured physiological parameters (i.e., blood pressure (BP), peripheral pulse oximetry (SpO2), electrocardiography (ECG), heart rate (HR), etc.) indirectly suggest the reduction of blood flow in a certain area of the brain. Promisingly, the fNIRS-EEG recordings provide additional parameters directly reflect the variation of hemodynamic activity and neuron electrical in real-time. The first step to creating the prediction system is to demonstrate this multimodal fNIRS-EEG system is a promising tool to fill in the vacancy of current technologies for post-treatment management.

The second step is to develop a machine learning (ML) model for stroke risk prediction based on the database collected during the monitoring. With the emerging of applying ML models on electronic health records (EHRs) for risk prediction of diseases, the recorded data will be cleaned and organized as a database for future research use. Besides the parameters derived from fNIRS-EEG recordings, the routine measurements of physiological parameters (i.e., BP, SpO2, ECG, HR, etc.) for post-treatment are included in the database. In addition, the medical history of the recruited patients is included in the database since these factors impact the outcome. Furthermore, the clinical outcomes (the disability and mortality of certain periods after MT) are included. To deal with such a high dimension EHRs and the recorded physiological data, machine learning is an emerging technique to explore the relationship between input features and outcome. Therefore, it is becoming a popular technique for disease prediction. Examples are implementing ML algorithms for predicting seizures using intracortical EEG (iEEG) recordings and predicting bladder volume using a smart neuroprosthesis [45]. The association of the recorded fNIRS-EEG parameters and other data in EHRs with the reduction of blood flow when complications occur is analyzed using machine learning techniques. The importance of features (recorded parameters and patients' health information) is ranked according to the occurrence and progression of reocclusion or hemorrhage during modeling.

With the non-invasive and wearable fNIRS-EEG system and the development prediction model, the variation of selected features recorded from post-treatment can identify or predict the occurrence of complications. In addition, the model can be used to guide personalized medical management based on each patient's condition. Moreover, the fNIRS-EEG system and prediction model applied to people with high stroke risk can suggest the progress of atherosclerosis (accounts for 20% of ischemic stroke) or predict the occurrence of ischemic stroke.

#### **3.3 Challenges of using fNIRS to avoid stroke**

The challenges of acquiring high-quality fNIRS signals of brain activity and the corresponding possible solutions have been well explored [14]. The difficulties of applying fNIRS to stroke patients to generate a stroke risk prediction system are introduced here.

## *3.3.1. Appropriate patients with electrodes and optodes located in the region of interest benefit the most from the fNIRS-EEG measurement*

To record the effective signals revealing the occurrence of complications after the treatments, placing the optodes and electrodes on the area that may present the most variation of neuron electrical signals and hemodynamics is a key scientific issue. This issue can be resolved by selecting appropriate patients and relevant locations of the electrodes and optodes.

fNIRS has limited spatial resolution due to the properties of light penetration, thus, the recorded data is sensitive to the hemodynamic changes of the blood vessels located no deeper than 1-2 cm beneath the scalp [46, 47]. Therefore, patients with revascularized vessels located at a shallow layer of the cerebral cortex are appropriate for the proposed EEG-fNIRS measurement. Accordingly, potential complications that occur close to the revascularized vessels can be monitored by the fNIRS-EEG system. Based on this criterion, the patients recruited in the study discussed in **Figure 6** were selected by physicians according to the CT, MRI, or other imaging techniques carried out immediately after admission and after the treatment. CT can distinguish ischemic stroke and hemorrhage stroke. Moreover, the location of infarction and the successfulness of recanalization before and after treatment can be determined. On the other hand, MRI detects the location of ischemia and the affected area with higher precision than CT images.

When potential patients for the measurement are selected, the locations of the fNIRS optodes and EEG electrodes are determined and arranged alternately to cover the area of interest. Such area of interest includes the brain area around the infarcted site and the surgical intervention as well as the symmetric area of the contralesional hemisphere.

### *3.3.2. Duration of monitoring*

Various types of complications occur at different times after treatment. For example, around 6% of patients receiving MT have hemorrhage and 2-20% of them have reocclusion. Hemorrhage often happens within 24 hours of the start of MT. Various types of hemorrhage are related to surgical procedures or the health condition of patients (blood pressure, location of ischemia region, severity of the stroke, etc.). Reocclusion happens when more platelets accumulate at a certain location of a vessel where the surgery takes place due to the damage of the wall of the vessel. In addition, patients with more embolic fragments and stenosis at the thrombectomy site have a high risk of reocclusion. The reocclusion often occurs 24 to 48 hours after MT and is associated with poor outcomes [48].

To monitor the progress of various types of complications, the ideal duration of monitoring should last 24 hours continuously after MT and then at least every 12 hours from 24 hours after MT till the patient is discharged from the hospital.

#### *3.3.3. Data analysis to extract the effective parameters recorded using fNIRS-EEG*

Investigating various approaches to handle the recorded data during data analysis to find the optimal parameters suggesting the occurrence and progress of complications is the 3rd key issue. The analyzed data include directly measured signals, derived values, or the combination of the data from fNIRS and EEG recordings.

For fNIRS, the investigated parameters are the change of total hemoglobin concentration (THC) and regional cerebral oxygen saturation (rSO2) can be computed based on ΔOxyHb and ΔDeoxyHb. The hemodynamics across the hemispheres can be compared by the interhemispheric rSO2 (IHΔrSO2) difference. The association of the recorded parameters with physiological conditions is analyzed to better understand the causes of the variation of parameters.

Regarding EEG, since it is not easy for people unfamiliar with EEG signals to understand the related neurological activity with the corresponding recording, one common approach to analyzing these signals is to convert them to a frequency domain. The converted data inspecting the signal strength as a function of frequency is defined as power spectrum density (PSD). This is useful since it is known that signal patterns of certain frequencies will occur during certain brain activities. It has been studied that the ratio of the power of the delta band to the power of alpha bands (DAR) correlated the most to the infarct area of ischemic stroke [49]. Also, the interhemispheric asymmetry can be characterized using the Brain Symmetry Index (BSI) derived based on the power of signals from the bilateral hemispheres.

The parameters for analysis are not restricted to those discussed above; other possible combinations of the recorded data are also explored. The factors integrated with the direct and indirect parameters from fNIRS and EEG recording can be potential features reflecting the occurrence of complications after stroke treatments.

Besides data collected from the recording facilities, other data recorded through conventional clinical procedures after MT are the key elements to be compared with. Ideally, continuous CT or MRI scans show the most detailed information on the occurrence and progress of complications. However, it is not feasible to attain the scans frequently. Therefore, the vital signs and the rating scales evaluating the impairment caused by a stroke are the parameters to suggest the occurrence of complications or the outcome of recanalization. Routine measurements recommended by worldwide guidelines for post-thrombectomy patients are BP, SpO2, ECG, heart rate, and blood glucose.

The National Institute of Health Stroke Scale (NIHSS) is the most used score scale for stroke management to quantify neurological dysfunction or deficit after stroke. According to the guidelines regarding the management after thrombectomy, the vital signs and NIHSS are monitored every 30 minutes within the first 6 hours, then every 1 hour from 6 to 24 hours, and then once per 6 hours from 24 to 72 hours. The correlation between the data from fNIRS-EEG recordings and conventional features for detecting complications can suggest the occurrence of complications or stroke onset.

## **4. Conclusions**

We introduced in this chapter the emerging optical technique, fNIRS, to monitor the hemodynamic condition reflecting the activities of the brain. fNIRS consists of emitting light with wavelengths near-infrared range to detect the concentration variation of OxyHb and DeoxyHb. Consequently, the detected light is converted to the absorber's concentration using the modified Beer-Lambert law. fNIRS is an outstanding technique to perform real-time monitoring compared with other common neuroimaging techniques in size, weight, and accessibility. fNIRS has been applied on stroke patients to monitor their hemodynamic conditions in prior-, peri-, and postoperative periods. The variations of images provided by fNIRS, before the onset of complications on stroke patients who received either thrombolysis or thrombectomy treatments are crucial to stroke prediction. Integrating the recorded fNIRS-EEG features, other physiological parameters, and EHRs, the ML model can provide a risk

of the stroke prediction system. With the feasibility of continuous or high-frequency fNIRS recordings, the prediction precision of the prediction system to avoid stroke can be increased.

## **Acknowledgements**

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Westlake University, grant number [041030080118], and Zhejiang Key R&D Program [2021C03002].

## **Conflict of interest**

The authors declare no conflict of interest.

## **Notes**

The copyrights of the figures in this book chapter belong to CenBRAIN Neurotech, School of Engineering, Westlake University. Please properly cite the origin of the figures if wanting to reuse them in other places.

## **Author details**

Yun-Hsuan Chen1,2\* and Mohamad Sawan1,2\*

1 CenBRAIN Neurotech, School of Engineering, Westlake University, Hangzhou, China

2 Institute of Advanced Technology, Westlake Institute for Advanced Study, Hangzhou, China

\*Address all correspondence to: chenyunxuan@westlake.edu.cn and sawan@westlake.edu.cn

© 2022 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

## **References**

[1] Zhou Y. Neuroimaging - neurobiology, multimodal and network applications [Internet]. London: IntechOpen; 2020. 140 p. [cited 2022 Jun 06]. Available from: https://www.intechopen.com/books/9347 DOI: 10.5772/intechopen.83577

[2] Zhang Y-D, Dong Z, Wang S-H, Yu X, Yao X, Zhou Q, et al. Advances in multimodal data fusion in neuroimaging: Overview, challenges, and novel orientation. Information Fusion. 2020;**64**:149-187

[3] Rybář M, Daly I. Neural decoding of semantic concepts: A systematic literature review. Journal of Neural Engineering. 2022;**19**(2):021002

[4] Kakkar P, Kakkar T, Patankar T, Saha S. Current approaches and advances in the imaging of stroke. Disease Models & Mechanisms. 2021;**14**(12):dmm048785

[5] Jiaming C, Theodore JH, Pulkit G, Jana MK. Enhanced spatiotemporal resolution imaging of neuronal activity using joint electroencephalography and diffuse optical tomography. Neurophotonics. 2021;**8**(1):1-21

[6] Leon-Carrion J, Leon-Dominguez U. Functional near-infrared spectroscopy (fNIRS): Principles and neuroscientific applications. In: Bright P, editor. Neuroimaging - Methods. London, UK: InTech; 2012

[7] Pinti P, Tachtsidis I, Hamilton A, Hirsch J, Aichelburg C, Gilbert S, et al. The present and future use of functional near-infrared spectroscopy (fNIRS) for cognitive neuroscience. Annals of the New York Academy of Sciences. 2020;**1464**(1):5-29

[8] Rahman MA, Siddik AB, Ghosh TK, Khanam F, Ahmad M. A narrative review on clinical applications of fNIRS. Journal of Digital Imaging. 2020;**33**(5):1167-1184

[9] Larrivee D, Rayegani SM. Neurostimulation and neuromodulation in contemporary therapeutic practice [Internet]. London: IntechOpen; 2020. p. 274 [cited 2022 Jun 06]. Available from: https://www.intechopen. com/books/8059 DOI: 10.5772/ intechopen.77890

[10] Paulmurugan K, Vijayaragavan V, Ghosh S, Padmanabhan P, Gulyás B. Brain–computer interfacing using functional near-infrared spectroscopy (fNIRS). Biosensors. 2021;**11**:389. DOI: 10.3390/bios11100389

[11] Curtin A, Tong S, Sun J, Wang J, Onaral B, Ayaz H. A systematic review of integrated functional near-infrared spectroscopy (fNIRS) and transcranial magnetic stimulation (TMS) studies. Frontiers in Neuroscience. 2019;**13**:84. DOI: 10.3389/fnins.2019.00084

[12] Figeys M, Zeeman M, Kim ES. Effects of transcranial direct current stimulation (tDCS) on cognitive performance and cerebral oxygen hemodynamics: A systematic review. Frontiers in Human Neuroscience. 2021;**15**:623315. DOI: 10.3389/fnhum.2021.623315

[13] Beltrán-Rodríguez I, Casado LG-T, Durán-Borrella Ó, González-Feito P, Sánchez-Serrano I, de la Fuente Blanco R. Post-stroke rehabilitation: A necessary step. In: Sanchetee P, editor. Post-Stroke Rehabilitation. London: IntechOpen; 2022

[14] Chen Y-H, Sawan M. Trends and challenges of wearable multimodal technologies for stroke risk prediction. Sensors. 2021;**21**(2):460. DOI: 10.3390/ s21020460

*Monitoring Brain Activities Using fNIRS to Avoid Stroke DOI: http://dx.doi.org/10.5772/intechopen.105461*

[15] Guo Y. A new paradigm of "real-time" stroke risk prediction and integrated Care Management in the Digital Health era: Innovations using machine learning and artificial intelligence approaches. Thrombosis and Haemostasis. 2022;**122**(1):5-7

[16] Wheelock MD, Culver JP, Eggebrecht AT. High-density diffuse optical tomography for imaging human brain function. Review of Scientific Instruments. 2019;**90**(5):051101

[17] Moradi S, Ferdinando H, Zienkiewicz A, Särestöniemi M, Myllylä T. Measurement of cerebral circulation in human. In: Scerrati A, Ricciardi L, Dones F, editors. Cerebral Circulation - Updates on Models, Diagnostics and Treatments of Related Diseases. London: IntechOpen; 2022

[18] Almajidy RK, Mankodiya K, Abtahi M, Hofmann UG. A newcomer's guide to functional near infrared spectroscopy experiments. IEEE Reviews in Biomedical Engineering. 2020;**13**:292-308. DOI: 10.1109/ RBME.2019.2944351

[19] Pagano R, Libertino S, Sanfilippo D, Fallica G, Lombardo S. Improvement of sensitivity in continuous wave near infra-red spectroscopy systems by using silicon photomultipliers. Biomedical Optics Express. 2016;**7**(4): 1183-1192

[20] Ferrari M, Quaresima V. A brief review on the history of human functional near-infrared spectroscopy (fNIRS) development and fields of application. NeuroImage. 2012;**63**(2):921-935

[21] Torricelli A, Contini D, Pifferi A, Caffini M, Re R, Zucchelli L, et al. Time domain functional NIRS imaging for human brain mapping. NeuroImage. 2014;**85**:28-50

[22] Hubin Z, Robert JC. Review of recent progress toward a fiberless, whole-scalp diffuse optical tomography system. Neurophotonics. 2017;**5**(1):1-11

[23] Liu G, Cui W, Hu X, Xiao R, Zhang S, Cai J, et al. Development of a miniaturized and modular probe for fNIRS instrument. Lasers in Medical Science. 2022;**37**(4):2269-2277

[24] Tsow F, Kumar A, Hosseini SMH, Bowden A. A low-cost, wearable, do-it-yourself functional near-infrared spectroscopy (DIY-fNIRS) headband. HardwareX. 2021;**10**:e00204

[25] Saikia MJ. A Wearable, Configurable Functional Near-Infrared Spectroscopy System for Wireless Neuroimaging. Kingston, USA: University of Rhode Island; 2019

[26] Lacerenza M, Buttafava M, Renna M, Torricelli A, Tosi A, Mora AD, et al., editors. Compact time-domain NIRS oximeter for non-invasive brain and muscle monitoring. In: Applied Industrial Optics 2021. Washington, DC: Optica Publishing Group; 2021

[27] Uchitel J, Vidal-Rosas EE, Cooper RJ, Zhao H. Wearable, integrated EEGfNIRS technologies: A review. Sensors (Basel). 12 Sep 2021;**21**(18):6106. DOI: 10.3390/s21186106. PMID: 34577313; PMCID: PMC8469799

[28] Sawan M, Salam MT, Lan JL, Kassab A, Gelinas S, Vannasing P, et al. Wireless recording systems: From noninvasive EEG-NIRS to invasive EEG devices. IEEE Transactions on Biomedical Circuits and Systems. 2013;**7**:186-195

[29] Kassab A, Le Lan J, Tremblay J, Vannasing P, Dehbozorgi M, Pouliot P, et al. Multichannel wearable fNIRS-EEG system for long-term clinical monitoring. Human Brain Mapping. 2018;**39**(1):7-23

[30] OpenNIRS.org – Modular open hardware for Near InfraRed Spectroscopy. 2022. Available from: http://opennirs. org/m3ba-mobile-modular-multimodalbiosignal-acquisition/

[31] Marshall RS. Progress in intravenous thrombolytic therapy for acute stroke. JAMA Neurology. 2015;**72**(8):928-934

[32] Çetiner M. Thrombolysis in acute stroke. In: Ambrosi PB, editor. Cerebrovascular Diseases - Elucidating Key Principles. London: IntechOpen; 2021

[33] Giles JA, Vellimana AK, Adeoye OM. Endovascular treatment of acute stroke. Current Neurology and Neuroscience Reports. 2022;**22**(1):83-91

[34] Pilgram-Pastor SM, Piechowiak EI, Dobrocky T, Kaesmacher J, Den Hollander J, Gralla J, et al. Stroke thrombectomy complication management. Journal of NeuroInterventional Surgery. 2021;**13**(10):912

[35] Delgado-Mederos R, Gregori-Pla C, Zirak P, Blanco I, Dinia L, Marín R, et al. Transcranial diffuse optical assessment of the microvascular reperfusion after thrombolysis for acute ischemic stroke. Biomedical Optics Express. 2018;**9**(3):1262-1271

[36] Annus A, Nagy A, Vecsei L, Klivenyi P. 24-hour near-infrared spectroscopy monitoring of acute Ischaemic stroke patients undergoing thrombolysis or Thrombectomy: A pilot study. Journal of Stroke and Cerebrovascular Diseases. 2019;**28**(8):2337-2342

[37] Sanci E. Evaluation of the Effectiveness of Thrombolytic Therapy in Acute Ischemic Stroke Patients Using NIRS (E-NESE). ClinicalTrials.gov; 2018 Available from: https://clinicaltrials.gov/

ct2/show/record/NCT03621917?term=N CT03621917&draw=2&rank=1

[38] Giacalone G, Zanoletti M, Re R, Germinario B, Contini D, Spinelli L, et al. Time-domain near-infrared spectroscopy in acute ischemic stroke patients. Neurophotonics. 2019;**6**(1):015003

[39] Hiramatsu R, Furuse M, Yagi R, Ohnishi H, Ikeda N, Nonoguchi N, et al. Limit of intraoperative near-infrared spectroscopy monitoring during endovascular thrombectomy in acute ischemic stroke. Interventional Neuroradiology. 2018;**24**(1):57-63

[40] Ritzenthaler T, Cho TH, Mechtouff L, Ong E, Turjman F, Robinson P, et al. Cerebral near-infrared spectroscopy: A potential approach for Thrombectomy monitoring. Stroke. 2017;**48**(12):3390-3392

[41] Hametner C, Stanarcevic P, Stampfl S, Rohde S, Veltkamp R, Bösel J. Noninvasive cerebral oximetry during endovascular therapy for acute ischemic stroke: An observational study. Journal of Cerebral Blood Flow & Metabolism. 2015;**35**(11):1722-1728

[42] Iversen HK. Monitoring Stroke Patients with Near-Infrared Spectroscopy Before, during and after Endovascular Treatment. ClinicalTrials.gov; 2018. Available from: https://clinicaltrials.gov/ ct2/show/NCT03738644

[43] Forti RM, Favilla CG, Cochran JM, Baker WB, Detre JA, Kasner SE, et al. Transcranial optical monitoring of cerebral hemodynamics in acute stroke patients during mechanical Thrombectomy. Journal of Stroke and Cerebrovascular Diseases. 2019;**28**(6):1483-1494

[44] Akiyama T, Ohira T, Hiraga K, Tanizaki Y, Akaji K, Kawase T.

*Monitoring Brain Activities Using fNIRS to Avoid Stroke DOI: http://dx.doi.org/10.5772/intechopen.105461*

Hemodynamic response following neuronal activity in cases with severe stenosis of the cervical internal carotid artery and its serial change through carotid artery stenting. Journal of Neuroendovascular Therapy. 2008;**2**(2):93-100

[45] Rong G, Mendez A, Bou Assi E, Zhao B, Sawan M. Artificial intelligence in healthcare: Review and prediction case studies. Engineering. 2020;**6**(3):291-301

[46] Quaresima V, Ferrari M. A minireview on functional near-infrared spectroscopy (fNIRS): Where do we stand, and where should we go? Photonics. 2019;**6**(3):87

[47] Almajidy RK, Mankodiya K, Abtahi M, Hofmann UG. A Newcomer's guide to functional near infrared spectroscopy experiments. IEEE Reviews in Biomedical Engineering. 2020;**13**:292-308

[48] Mosimann Pascal J, Kaesmacher J, Gautschi D, Bellwald S, Panos L, Piechowiak E, et al. Predictors of unexpected early Reocclusion after successful mechanical Thrombectomy in acute ischemic stroke patients. Stroke. 2018;**49**(11):2643-2651

[49] Shreve L, Kaur A, Vo C, Wu J, Cassidy JM, Nguyen A, et al. Electroencephalography measures are useful for identifying large acute ischemic stroke in the emergency department. Journal of Stroke and Cerebrovascular Diseases. 2019;**28**(8):2280-2286

## **Chapter 7**
