Wearable Functional Near-Infrared (Fnir) Technology and Its Applications in Naturalistic Conditions

Wearable functional near-infrared technology (fNIR) is a noninvasive neuroimaging technology that is affordable, low cost, and user-friendly, making it a promising technology to study how our brain works in naturalistic conditions. A practical user-friendly fNIR device that allows naturalistic use (e.g., home-use) can open unprecedented opportunities with health, psychologic, political/economic, and physiologic impacts. Wearable fNIR devices impose additional requirements compared to laboratory-based fNIR systems, including size, weight, and motion/ambient artifacts, without sacrificing key performance. Furthermore, to ensure user-friendliness, special attentions to cost, power, connectivity, flexibility to adapt to various applications, and comfort have to be properly considered. This manuscript reviews wearable fNIR in naturalistic conditions, covering reported applications and technologies developed to enable fNIR wearability. We also highlight some of the remaining challenges of current wearable fNIR devices including the limited number of optodes. uncorrected artifacts and noise, and the need for user alignment and manual event recording.

Traditionally, many of the mental tasks designed for fNIR cognitive tests were performed in clinical settings, which pose environmental and confounding factors that are different from naturalistic conditions and do not reflect our everyday events [8].
To enable cognitive tests in naturalistic conditions, several groups have developed wearable fNIR systems [9]. For example, Almajidy et al. developed a multichannel NIRS system [10]. Piper et al. later developed a wearable multi-channel fNIRS system for freely moving subjects using a head cap [11]. Strangman et al. developed a wearable device with a multimodal physiological monitor with leads for biopotentials; 64 NIRS optodes (a typical optode consists of a light source and detector pair); force, acceleration, gyroscope, and temperature sensors; and the ability to connect to a commercial respirometer [12]. A wearable fNIRS-EEG system with 128 channels for fNIRS, 32 channels for EEG, two channels for auxiliary electrocardiogram (ECG) and an accelerometer for correction in the form of a cap was developed by Kassab et al [13]. However, despite the technological advances of wearable fNIR technology, the use of a cap to fix the optodes on the subject makes it uncomfortable for prolonged use [14] (e.g., everyday home use). As an alternative design, application of fNIR in a naturalistic environment with a fiber-less, multichannel wearable device in the form of a headset connected to a portable processing unit was demonstrated by Pinti et al [15]. Wyser et al. later developed a modular wearable device with a flexible form factor allowing multiple distances between optode source and detectors; it also featured multiple wavelengths [16]. Wyser's optode modules are about the size of a quarter and allows the use of headgear other than a cap; however, there is still a somewhat bulky control unit.
In addition, wearable fNIR has also demonstrated capability in identifying patients with Pakinsonian syndromes having higher prefrontal cortex activity to maintain posture stability [32] and subjects with traumatic brain injury [33]. Brain-computer interface [34] and correlation with fNIR hybrid system [35] have also been demonstrated. In summary, wearable fNIR allows unprecedented studies and diagnosis. However, many of these studies have been confined to laboratories or clinics (except campus navigation using wearable fNIR 29), rather than in naturalistic conditions (i.e., homes).

Methods
Given the promises, challenges, and advances of wearable fNIR technologies, a number of excellent reviews on wearable fNIR technologies deployable in naturalistic environments have been published [5,21,36]. We have performed a systematic search on recent advances and summarize below the various applications that have been published using wearable fNIR technologies in naturalistic environment as well as the features that have enabled fNIR technologies to move from the laboratory to naturalistic environments. Our search was based on entering the keywords "wearable fNIR," "functional near-infrared natural environment," and "fNIR natural" in Web of Science to search for relevant publications in this topic. A total of 30 matches was found with 20 unique and relevant publications from 1900-2019. (Table 1)  Under each broader category, there can also be sub-categories. For instance, under health, there are sub-categories pertaining to monitoring/screening and prevention applications. Example applications under these subcategories include application of fNIR to the monitoring of preterm infant brain health (hypoxia) [37], neural activities during cardiac and vascular surgery, cerebrovascular disease, epilepsy, and headache [38]. In addition, fNIR can also be used to predict likelihood of obesity and potentially used to prevent obesity [39]. Furthermore, it can be used to study decision-making, which can find applications in marketing, advertising, learning and training optimization by understanding demands placed on individuals as they engage with various tasks [5]. By capturing cognitive capacity/limit (i.e., maximum brain activity to maintain performance), operator efficiency can be maximized while human error can be minimized [5,40]. To study social interactions, a technique called hyperscanning has been employed. Hyperscanning refers to simultaneous scanning of the brains of interacting subjects. It is a way to study social interaction and neural functions and has been demonstrated in prefrontal

Am J Biomed Sci & Res
Copy@ Audrey Bowden cortex activation during acting and other activities [4]. Combining virtual reality and fNIR allows control of subject's experience in dynamic environment [5]. fNIR can also serve as a research tool for studying brain activation in various activities, decision-making and cognitive states where selective attention and physical work compete with executive processing (e.g., balancing a ping-pong ball on a small card while walking) [40,41]. It has also been applied to distinguish which part of a story the subject was listening and to understand the neural mechanism underlying imagined rather than intelligible speech for communication [42,43] (Table 1).

Wearable fNIR requirements
A wearable fNIR device has additional requirements compared to a laboratory-based fNIR system. In a laboratory, the subjects are more stationary and the environment is well controlled. If the subjects are engaged in activities, more motion can be expected. Due to the sensitivity of the optical detection module of the fNIR device, unaccounted motions will manifest as noise. In general. the heavier the unit, the more it tends to move due to inertia, resulting in more motion artifacts (i.e., noise) and adding stress on the connecting wires/cables. To reduce motion artifacts, attempts have been made to apply correction algorithms [44], to incorporate motion sensors such as an accelerometers and gyroscopes and to optimize the size and mass of the optodes [2]. Several groups have studied the effects of the distance between the source and detector of an optode [10], and some have employed a short channel (5mm to 1cm) [16,45] to correct for artifacts due to superficial tissue and motion. Wireless designs have also been implemented to eliminate connecting wires/ cables [21,33,41] to enhance comfort and to remove restriction to subjects' activities.
Furthermore, many wearable fNIR devices adopt the use of a cap to fix the optodes in place, although prolonged use of the cap can result in discomfort. A smaller and more flexible form factor (e.g., a headband) can provide comfort to the user and stability to the optodes [5,15,26]. Finally, a modular design can streamline the number of components to be mounted on the subject, thus reducing the overall weight, size, and cost of the wearable fNIR device. Furthermore, the use of a modular design can allow parts to be easily switched out, thus affording the use of disposable components for hygiene or performance reasons [1,5,16,46,47,51].
We have summarized key features from the searched literatures implemented by wearable fNIR devices ( Table 2).
Although there has been tremendous technical progress in wearable fNIR devices, there are still remaining challenges that needs to be addressed for a truly user-friendly, home-use, wearable system. While there are existing algorithms and sensors that can help correct for motion artifacts, the correction still has room to improve. Furthermore, in order to capture as much neural activity as possible, an fNIR device should have wide coverage over different brain regions. As a result, multiple optodes will have to spread over the entire surface of the head where the brain resides.
Increasing the number of optodes is often difficult for a low cost, low power fNIR device, as each additional optode adds cost and power consumption to the device. To reduce the required number of optodes, optode alignment will be critical. A way to help align different brain regions for different targeted functions, including alignment with dorsolateral prefrontal cortex for working memory and attention, pars opercularis for language production and motor cortices for articulation, would be helpful in accurately assessing cognitive and behavior functions [31,48]. In addition, hair can often get in the way of obtaining high quality fNIR signal. In order to probe areas of the scalp where hair might be present, a brush-like structure that allows hair to be pushed aside has been demonstrated by Khan et al. [3]. Some groups have also taken the approach of focusing only on the prefrontal cortex (PFC), which has been shown to be critical to working memory [49], avoiding the issue of hair altogether. Nonetheless, measuring only the prefrontal cortex will miss out on other important regions such as the motor cortex 3. have focused on assisting researchers in finding the right brain region to study [48] and [50][51][52][53][54][55]. A topological mapping ability of the fNIR device can help address this challenge independent of the availability of MRI images.

Conclusion
Currently, wearable fNIR devices have a limited number of optodes. The number is typically fixed at the design stage, as each optode is hardwired to the ADC converter. To make the design more flexible, optodes can be wirelessly connected to the ADC converter.
To reduce power and cost for the wireless implementation, RFID or acoustic approaches can be employed. Furthermore, data from various channels should be multiplexed to further save on power and cost. Frequency-division multiplexing, time-division multiplexing, or code-division multiplexing can be used. We anticipate a user-friendly, low-cost, wearable fNIR device would allow many important social/behavioral/psychological questions to be answered. Further, we expect the results will lead to important contributions to behavior science, health care and engineering: for example, in understanding the recognitive role in obesity and in child development, enhancing medical diagnosis, creating novel brain-machine interface, and developing new marketing and advertising tools.