Implementing neuroimaging and eye tracking methods to assess neurocognitive development of young infants in low- and middle-income countries

Abstract

Infants and children in low- and middle-income countries (LMICs) are frequently exposed to a range of environmental risk factors which may negatively affect their neurocognitive development. The mechanisms by which factors such as undernutrition and poverty impact development and cognitive outcomes in early childhood are poorly understood. This lack of knowledge is due in part to a paucity of objective assessment tools which can be implemented across different cultural settings and in very young infants. Over the last decade, technological advances, particularly in neuroimaging, have opened new avenues for research into the developing human brain, allowing us to investigate novel biological associations. This paper presents functional near-infrared spectroscopy (fNIRS), electroencephalography (EEG) and eye tracking (ET) as objective, cross-cultural methods for studying infant neurocognitive development in LMICs, and specifically their implementation in rural Gambia, West Africa. These measures are currently included, as part of a broader battery of assessments, in the Brain Imaging for Global Health (BRIGHT) project, which is developing brain function for age curves in Gambian and UK infants from birth to 24 months of age. The BRIGHT project combines fNIRS, EEG and ET with behavioural, growth, health and sociodemographic measures. The implementation of these measures in rural Gambia are discussed, including methodological and technical challenges that needed to be addressed to ensure successful data acquisition. The aim is to provide guidance to other groups seeking to implement similar methods in their research in other LMICs to better understand associations between environmental risk and early neurocognitive development.

Keywords: EEG; eye tracking; fNIRS; global health; infancy; low-and middle-income countries (LMIC); neurocognitive development.

Figure 1.. Set up and output of one BRIGHT fNIRS study.

( a) A sleeping infant wearing the fNIRS headband holding sources (red) and detectors (blue) in a similar set up to the data shown in the lower panel. ( b) Set up of fNIRS study in The Gambia for awake infants. Infant is seated on parent’s lap wearing the fNIRS headgear while attending to stimuli on the screen. ( c) Typical haemodynamic responses to an auditory stimulus across nine channels distributed across infants’ right temporal lobe. Responses are averaged across a group of 129 one-month-old infants (from the first wave of data collection in the BRIGHT project) showing localised increases in HbO ₂ (red) and a corresponding decrease in HHb (blue) in response to auditory stimulation; red frames illustrate a significant increase of HbO ₂, blue shading indicates a significant decrease in HHb. N, number of infants with valid data for each channel; ch, channel number. Image copyright: Gates Foundation ( a), Ian Farrell ( b).

Figure 2.. BRIGHT Project Habituation and Novelty Detection (HaND) study.

Left: haemodynamic HbO ₂ eight-month-old group response to repeated stimuli. Right: individual amplitudes. Participants: UK (orange), N=42; Gambia (green), N=60. For familiarisation trial (Fam.) 3, lower = better. High data quality allows for a clear differentiation between cohorts as indicated by non-overlapping error bars. (Adapted from figures in Lloyd-Fox et al., 2019).

Figure 3.. Illustration of fNIRS headgear used at BRIGHT study sites.

Upper panel: fNIRS bespoke headgear worn by infants in the BRIGHT project in the UK and The Gambia. Lower panel: optodes and fibres (far left), sensor arrays with clip-on optode holders (left), source (red) and detector (blue) fibres clipped into array (right) and headband and optode array combined (far right). Image copyright: Ian Farrell (top panel), Sarah Lloyd-Fox (bottom panel).

Figure 4.. Effects of hair braids on fNIRS signal quality in posterior temporal and frontal regions.

Plots show changes in the haemodynamic response of HbO ₂ (red) and HHb (blue) at each channel position on the child’s head. A close correspondence can be seen between the location of hair braids and noise levels in the signal. It can also be seen that it is possible to obtain good signal to noise in channels neighbouring those affected by the braiding. Image copyright: Sarah Lloyd-Fox.

Figure 5.. Stimuli used in fNIRS social vs. non-social paradigm used in The Gambia (left) and the UK (right).

Top row shows examples of social stimuli used at either site. Bottom rows show examples of non-social still images specific to cultural contexts.

Figure 6.. Set up and output of BRIGHT EEG study.

( a) Set up of EEG assessment at 5 months at MRCG Keneba. Infant is seated on parent’s lap and silently interacts with one of the researchers. The wireless EEG system and headphones allow for some flexibility regarding the infant’s position in the room. ( b) Grand Average of ERP response of n = 170 one-month-old infants tested at MRCG Keneba for frequent tones (blue), infrequent sounds (red) and trial unique sounds (yellow) showing differential neural responses across these sound categories. Image copyright: Laura Katus.

Figure 7.. Set up and data quality metrics of BRIGHT ET study.

Upper panels: BRIGHT project eye tracker setup, with infant seated on parent’s lap while attending to the screen. Eye tracking is integrated in a panel just below the screen (upper left panel). Lower panels: six measures of data quality. Clockwise from top left, ( a) proportion of samples in which the eyes were detected; ( b) spatial precision; ( c) flicker, the ratio of validity for sample n to sample n+1; ( d) the duration of each eye tracking session; ( e) distance of the infant from the centre of the "track box", the volume in which the eye tracker can detect eyes (the centre of which yields optimal data quality); ( e) the distance of the infant from the screen. Red lines represent the sample mean, red boxes represent 95% confidence intervals, blue boxes represent 1 SD.

Figure 8.. Headcap used to shield fNIRS headband from near infrared light.

The cap is attached to the fNIRS headgear at the infant’s forehead using Velcro. The lower edge of the headband has elastic spanning the lower rim of the fNIRS cap. Velcro strips in the back of the headband allow quick adjustment for different head sizes to prevent delays in capping.