This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
The recent surge in commercially available wearable technology has allowed real-time self-monitoring of behavior (eg, physical activity) and physiology (eg, glucose levels). However, there is limited neuroimaging work (ie, functional magnetic resonance imaging [fMRI]) to identify how people’s brains respond to receiving this personalized health feedback and how this impacts subsequent behavior.
Identify regions of the brain activated and examine associations between activation and behavior.
This was a pilot study to assess physical activity, sedentary time, and glucose levels over 14 days in 33 adults (aged 30 to 60 years). Extracted accelerometry, inclinometry, and interstitial glucose data informed the construction of personalized feedback messages (eg, average number of steps per day). These messages were subsequently presented visually to participants during fMRI. Participant physical activity levels and sedentary time were assessed again for 8 days following exposure to this personalized feedback.
Independent tests identified significant activations within the prefrontal cortex in response to glucose feedback compared with behavioral feedback (
Presenting personalized glucose feedback resulted in significantly more brain activation when compared with behavior. Participants reduced time spent sedentary at follow-up. Research on deploying behavioral and physiological feedback warrants further investigation.
Physical inactivity, insufficient levels of physical activity, is attributable to 9% of premature mortality and 7% of type 2 diabetes cases [
Over the last decade, wearable activity monitors have grown in popularity in consumer markets to help users physically track their movement behaviors (eg, active minutes, step counts, distance traveled, time spent sitting) [
Neuroimaging techniques are useful to recognize and identify the intricate relationships between cognitions, brain functions, and behavior [
fMRI can improve our understanding of how cognitive processes vary between those who do change their behavior following exposure to a stimulus and those who do not subsequently change [
A total of 33 participants (57% female) were recruited from a university in the United Kingdom via advertisement posters and email. Participants were aged 30 to 60 years, had no mobility-related musculoskeletal problems, had no confirmed diagnosis of diabetes, were willing and able to comply with the study protocol, met standard fMRI safety criteria (no metal in body, not claustrophobic, not pregnant), and were right-handed. All participants completed a physical activity readiness questionnaire [
Each participant’s consent was obtained according to the Declaration of Helsinki, and all experimental procedures were approved by the Loughborough University Ethics Advisory Committee (R15-P142).
Data were collected between June and September 2016. The study design is presented in
Study design.
Weight and body fat percentage were measured using the MC 780 MA scale (Tanita) following the removal of shoes and socks. Body mass index was calculated as weight (kg) divided by height (m) squared (weight/height2). Glucose and hemoglobin A1c (HbA1c) were analyzed using a Cholestech LDX system and Afinion AS100 Analyzer (both Alere Inc), respectively. Participants arrived fasted for ≥8 hours prior to the collection of a capillary blood sample.
An ActiGraph wGT3x-BT accelerometer (ActiGraph LLC) was worn on a waistband (on the right anterior axillary line) to objectively measure physical activity. Participants were asked to wear the validated device [
A Lumo (Lumo Bodytech Inc) posture sensor was worn on a waistband (in the lumbosacral region) in contact with the skin to measure sedentary behavior (time spent sitting, driving, lying, standing, stepping, and number of sit-to-stand transitions) during baseline and follow-up. Devices were calibrated to the wearer. Participants were asked to wear the device only during waking hours, remove it for any water-based activities (eg, showers or bathing), and place the device on charge overnight each day. The Lumo has been found to produce valid measurements of sedentary behavior compared with the ActivPAL (PAL Technologies Ltd), with a mean error of 9.5% [
The Freestyle Libre flash glucose monitor (Abbott Laboratories) measures interstitial glucose levels via a minimally invasive 5 mm flexible filament inserted into the posterior upper arm. The sensor works based on the glucose-oxidase process by measuring an electrical current proportional to the concentration of glucose. Tegaderm transparent film dressing (3M Health Care) was applied on top of the sensor to maintain its position. Participants were informed not to remove the sensor and to scan at least once every 7 hours (a conservative decision as the manufacturer states 8 hours to avoid data loss). As a result, participants were able to see their real-time glucose levels during baseline wear. An indication of how many times participants viewed this information (level of exposure) was identified by the number of time they scanned. Missing data were obtained because of a fault (sensor lasted <14 days) or the participant failed to scan at least once every 8 hours. The Freestyle Libre has been previously validated against venous sampling with an overall mean absolute relative difference of 11.4% with consistent accuracy throughout the 14 days [
Twenty personalized feedback messages were created for the purposes of this study and covered 4 topics: MVPA, light physical activity, sedentary behavior, and glucose levels (all presented in
Stimuli were presented on a monitor located 2.8 m behind the center of the scanner bore and viewed by a mirror mounted on the head coil. Adjustments to the positioning of the mirror were made for participants to ensure that the full monitor screen could be seen. We examined neural activity while participants were presented with feedback and were requested to maintain attention throughout. Prior to the start of the fMRI task, there was an initial period of 40 seconds of dummy scans which were immediately discarded. The fMRI task is outlined in
Personalized feedback stimuli.
Trial setup including 8 of the 24 blocks presented.
Brain imaging data were acquired on a 3T Discovery MR750w scanner (General Electric) using a 32-channel head coil at the National Centre for Sport and Exercise Medicine, Loughborough University, United Kingdom. Structural images (T1-weighted) were acquired using a fast spoiled gradient echo (FSPGR) Bravo sequence (3D volume, FSPGR; TR=8.2 ms; TE=3.1 ms; matrix size 240×240; 160 sagittal slices; FOV=240 mm; 1 mm thick). One functional scan lasting 16 minutes (480 volumes) was acquired during the task (2D gradient echo EPI; TR=2000 ms; TE=30 ms; flip angle=75 degrees; matrix size 64×64; 35 axial slices; FOV=205 mm; 3 mm thick). Stimulus presentation and synchronization to scanner acquisition were performed using Presentation version 18.1 (Neurobehavioral Systems Inc).
Functional data were preprocessed and analyzed using statistical parametric mapping (SPM12, Wellcome Department of Cognitive Neurology). All data reported are from scans that exhibited ≤3 mm in translational movement. Data were processed using a standard statistical parametric mapping approach, which consisted of scan realignment, coregistration, segmentation, normalization, and smoothing. Data were spatially aligned to the first functional image using 4th degree B-spline interpolation. Scans were then coregistered (mean functional image aligned with T1 then parameters applied to all functional images). Functional images were normalized into the Montreal Neurological Institute (MNI) standard stereotactic space with parameters applied to all functional images. A final smoothing step with a Gaussian Kernel with full width half maximum of 8 mm was applied to improve signal-to-noise ratio. The onsets and durations of each of the conditions of interest were modeled according to the block design described in the protocol. For each participant, brain activation was estimated using a general linear model (GLM) and included movement parameters (3 translations, 3 rotations) and a session constant as regressors. All regressors were convolved with SPM12’s canonical difference of the hemodynamic response function. Data were high-pass filtered with a cut off of 128 seconds to remove low-frequency noise and slow drifts in the signal. Family-wise error (FWE) correction was used to correct for multiple comparisons at
Second level random effects models for each task were constructed that averaged across participants and were subjected to further region of interest (ROI) and between-group analysis (described below). Exploratory whole brain searches were conducted for each contrast with a threshold set at
To examine demographic and self-report data, we conducted descriptive analyses using SPSS version 22.0 (IBM Corp). Two group
Parameter estimates corresponding to each significantly activated region, identified via fMRI data analysis, were extracted for each participant. Linear regressions provided partial correlation coefficients between the parameter estimates from the significant regions of interest and subsequent behavior at follow-up (ie, time spent in MVPA, light physical activity, and sedentary), controlling for wear time. The relationships between behavior change and activity from the ROIs were examined in separate models for each ROI, and the analyses were repeated to assess behavior via both accelerometry and inclinometry data.
A flow chart of individuals through the study and the characteristics of the study sample are presented in
Flowchart of individuals at each stage of the study.
Sample characteristics.
Characteristics | Whole sample (n=28) | |
Age (years), mean (SD) | 44.2 (9.5) | |
Male, % | 42.9 | |
Weight (kg), mean (SD) | 75.2 (15.3) | |
Body mass index (kg/m2), mean (SD) | 25.2 (4.3) | |
Body fat (%), mean (SD) | 26.7 (9.3) | |
HbA1ca (%), mean (SD) | 5.4 (0.4) | |
Glucose (mmol/L), mean (SD) | 5.0 (0.6) |
aHbA1c: hemoglobin A1c.
The 28 participants (43% male) had a mean age of 44.2 (SD 9.5) years (range 30 to 59 years). Three (11%) participants completed secondary school, 5 (18%) completed some additional training, and 20 (71%) received a bachelor’s degree or higher. Twenty-five (89%) were white, 2 (7%) were Chinese, and 1 (4%) was Asian or Asian British. Males were significantly taller (178.7 versus 167.5 cm), had a lower body fat percentage (18.8% versus 32.6%), and scanned the Freestyle Libre more frequently (9.5 versus 5.7 scans per day).
First, we contrasted each of the 4 topics with a fixation cross. The brain regions significantly activated in response to the initial contrasts of interest are presented in
We then proceeded to the main analysis that contrasted the topics between themselves. The brain regions identified as significantly activated are presented in
Average contrasting differences (thresholded at
Region | MNIa coordinates | ||||||||||||||||
Hemb | x | y | z | Voxels | Z | t | |||||||||||
Middle occipital gyrus | L | –38 | –74 | –14 | 178 | 6.29 | 9.99 | <.001 | |||||||||
Lingual gyrus | L | –14 | –94 | –10 | — | 6.25 | 9.89 | <.001 | |||||||||
Inferior occipital gyrus | L | –22 | –90 | –14 | — | 6.21 | 9.76 | <.001 | |||||||||
Subgyral | R | 36 | –62 | –16 | 11 | 6.06 | 9.29 | <.001 | |||||||||
Fusiform gyrus | L | –36 | –54 | –16 | 9 | 5.97 | 9.03 | <.001 | |||||||||
Subgyral | R | 34 | –84 | –6 | 93 | 5.95 | 8.97 | <.001 | |||||||||
Lingual gyrus | R | 24 | –92 | –10 | — | 5.86 | 8.74 | <.001 | |||||||||
Lingual gyrus | R | 16 | –90 | –8 | — | 5.63 | 8.11 | .001 | |||||||||
Inferior occipital gyrus | R | 44 | –76 | –12 | 2 | 5.62 | 8.09 | .001 | |||||||||
Middle occipital gyrus | R | 30 | –88 | 4 | 1 | 5.57 | 7.97 | .001 | |||||||||
Cuneus | L | –16 | –96 | –2 | 101 | 6.47 | 10.61 | <.001 | |||||||||
Middle occipital gyrus | L | –32 | –84 | –14 | 119 | 6.23 | 9.80 | <.001 | |||||||||
Middle occipital gyrus | L | –38 | –72 | –14 | — | 6.05 | 9.28 | <.001 | |||||||||
Subgyral | R | 34 | –84 | –6 | 83 | 6.05 | 9.26 | <.001 | |||||||||
Middle occipital gyrus | R | 30 | –84 | –14 | — | 5.68 | 8.24 | .001 | |||||||||
Middle occipital gyrus | R | 46 | –76 | –10 | 23 | 6.01 | 9.14 | <.001 | |||||||||
Subgyral | R | 36 | –62 | –16 | 23 | 5.90 | 8.83 | <.001 | |||||||||
Middle occipital gyrus | R | 28 | –98 | 6 | 19 | 5.77 | 8.48 | <.001 | |||||||||
Fusiform gyrus | L | –36 | –54 | –16 | 3 | 5.77 | 8.47 | <.001 | |||||||||
Fusiform gyrus | L | –34 | –50 | –18 | 2 | 5.70 | 8.30 | <.001 | |||||||||
Inferior frontal gyrus | L | –54 | 18 | 20 | 4 | 5.69 | 8.27 | <.001 | |||||||||
Lingual gyrus | R | 16 | –90 | –10 | 10 | 5.65 | 8.18 | .001 | |||||||||
Middle occipital gyrus | L | –36 | –72 | –16 | 19 | 5.99 | 9.11 | <.001 | |||||||||
Inferior occipital gyrus | L | –38 | –82 | –10 | 46 | 5.95 | 8.98 | <.001 | |||||||||
Subgyral | L | –20 | –94 | –6 | 36 | 5.87 | 8.77 | <.001 | |||||||||
Middle occipital gyrus | R | 36 | –84 | –4 | 4 | 5.78 | 8.50 | <.001 | |||||||||
Inferior frontal gyrus | L | –48 | 14 | 22 | 3 | 5.65 | 8.16 | .001 | |||||||||
Middle occipital gyrus | R | 48 | –76 | –10 | 3 | 5.59 | 8.02 | .001 | |||||||||
Subgyral | R | 28 | –88 | –6 | 1 | 5.57 | 7.97 | .001 | |||||||||
Cuneus | L | –16 | –96 | –6 | 218 | 6.69 | 11.38 | <.001 | |||||||||
Middle occipital gyrus | L | –36 | –74 | –16 | — | 6.13 | 9.50 | <.001 | |||||||||
Middle occipital gyrus | L | –20 | –90 | –14 | — | 5.90 | 8.83 | <.001 | |||||||||
Subgyral | R | 36 | –62 | –16 | 13 | 5.99 | 9.10 | <.001 | |||||||||
Lingual gyral | R | 14 | –90 | –8 | 28 | 5.97 | 9.05 | <.001 | |||||||||
Subgyral | R | 28 | –84 | –6 | 56 | 5.88 | 8.78 | <.001 | |||||||||
Middle frontal gyrus | L | –40 | 10 | 30 | 6 | 5.69 | 8.27 | <.001 | |||||||||
Middle occipital gyrus | R | 44 | –76 | –14 | 1 | 5.60 | 8.05 | .001 | |||||||||
Middle occipital gyrus | R | 30 | –84 | –14 | 2 | 5.58 | 8.00 | .001 | |||||||||
Middle occipital gyrus | L | –38 | –72 | –16 | 272 | 6.44 | 10.49 | <.001 | |||||||||
Cuneus | L | –16 | –96 | –6 | — | 6.33 | 10.12 | <.001 | |||||||||
Middle occipital gyrus | L | –32 | –84 | –14 | — | 6.07 | 9.33 | <.001 | |||||||||
Subgyral | R | 36 | –62 | –16 | 27 | 6.16 | 9.61 | <.001 | |||||||||
Subgyral | R | 34 | –84 | –6 | 135 | 6.14 | 9.53 | <.001 | |||||||||
Lingual gyral | R | 22 | –92 | –10 | — | 5.85 | 8.69 | <.001 | |||||||||
Middle occipital gyrus | R | 30 | –84 | –14 | — | 5.75 | 8.42 | <.001 | |||||||||
Superior parietal lobule | L | –32 | –62 | 58 | 5 | 6.06 | 9.28 | <.001 | |||||||||
Middle occipital gyrus | R | 46 | –76 | –12 | 24 | 5.96 | 9.00 | <.001 | |||||||||
Middle occipital gyrus | R | 48 | –66 | –14 | — | 5.88 | 8.79 | <.001 | |||||||||
Fusiform gyrus | L | –36 | –54 | –16 | 8 | 5.95 | 8.98 | <.001 | |||||||||
Middle frontal gyrus | L | –52 | 26 | 26 | 9 | 5.73 | 8.38 | <.001 | |||||||||
Thalamus | R | 22 | –28 | –2 | 2 | 5.69 | 8.27 | .001 |
aMNI: Montreal Neurological Institute.
bhem: hemisphere.
cFWE: family-wise error.
dMVPA: moderate-to-vigorous physical activity.
Average contrasting differences controlling for age, gender, and average daily number of glucose scans (thresholded at
Region | MNIa coordinates | ||||||||
Hemb | x | y | z | Voxels | Z | t | |||
Middle frontalgyrus | L | –32 | 36 | –12 | 25 | 5.60 | 8.17 | <.001 | |
Subgyral | L | –26 | 48 | 4 | 16 | 5.33 | 7.48 | <.001 | |
Cuneus | L | –2 | –80 | 4 | 34 | 5.05 | 6.85 | <.001 | |
Middle frontal gyrus | L | –32 | 36 | –12 | 8 | 4.95 | 6.63 | <.001 | |
Middle frontal gyrus | L | –20 | 34 | 42 | 11 | 4.94 | 6.61 | <.001 | |
Superior frontal gyrus | L | –26 | 50 | 4 | 3 | 4.79 | 6.29 | <.001 | |
Subgyral | R | 28 | –52 | 24 | 1 | 4.66 | 6.04 | <.001 |
aMNI: Montreal Neurological Institute.
bHem: hemisphere.
cFWE: family-wise error.
Group level significant activation pattern for the contrast glucose>behavior at the MNI coordinates (a) –32, 36, –12 and (b) –26, 48, 4.
Behavioral characteristics derived from accelerometry and inclinometry.
Accelerometrya, mean (SD) | Inclinometryb, mean (SD) | |||||
Baseline | Follow-up | Baseline | Follow-up | |||
Number of valid days | 7.0 (0.0) | 7.0 (1.0) | — | 4.2 (2.1) | 5.5 (1.7) | — |
Wear time | 903.5 (67.7) | 868.2 (70.4) | 924.3 (61.9) | 884.0 (61.6) | .001 | |
Step count | 9065.2 (3456.2) | 9634.0 (3699.3) | — | 8660.9 (2995.7) | 9580.3 (4326.0) | — |
Counts per minute | 194.0 (82.0 | 410.0 | <.001 | — | — | — |
Sedentary (min) | 589.0 (84.7) | 560.0 (75.6) | .014 | 602.2 (91.1) | 554.5 (89.4) | .001 |
Light PAc (min) | 265.0 (69.0) | 254.2 (71.1) | — | — | — | — |
Moderate (min) | 45.8 (31.0) | 50.7 (33.2) | — | — | — | — |
Vigorous (min) | 3.6 (6.6) | 3.2 (6.2) | — | — | — | — |
MVPAd (min) | 49.4 (34.2) | 53.9 (35.5) | — | — | — | — |
LVPAe (min) | 314.4 (66.4) | 308.1 (72.1) | — | — | — | — |
Stepping (min) | — | — | — | 93.5 (26.7) | 103.2 (44.1) | — |
Standing (min) | — | — | — | 228.5 (98.5) | 226.5 (67.8) | — |
a≥4 valid days, n=28 (100% compliance to ≥600 mins of accelerometer wear).
b≥1 valid day, n=23, (100% compliance to ≥600 mins of inclinometry wear).
cPA: physical activity.
dMVPA: moderate-to-vigorous physical activity.
eLVPA: light-to-vigorous physical activity.
The behavioral characteristics obtained via accelerometry and inclinometry are presented in
To investigate the relationship between brain activation and subsequent behavior, parameter estimates were calculated for the patterns of neural activation. Of these, only glucose feedback was positively associated with subsequent minutes of MVPA (
As recent neuroimaging work has highlighted value in analyzing individual responses to feedback relating to lifestyle behaviors [
Our findings identified activations within regions of the prefrontal cortex, in particular the middle frontal gyrus, subgyral, cuneus, and superior frontal gyrus upon comparison of personalized glucose feedback with behavioral feedback. Previous studies have also identified regions within the prefrontal cortex following exposure to antismoking images [
Investigating how individuals responded to personalized health-related feedback was an important component of this study as it has been well documented that receiving tailored feedback can result in greater resonance and consequently result in desirable health behaviors [
Our study identified a significant reduction of 29 minutes (or 47 minutes using inclinometry) in time spent sedentary from baseline to follow-up. Previous findings support this finding, having observed a 39.6 minute per day reduction in time spent sedentary [
In regard to the relationship of activation and subsequent behavior during the follow-up period, findings identified a positive partial correlation with minutes of MVPA. Previous studies have investigated behavior change subsequent to fMRI and have demonstrated positive associations between neural response (eg, to aversive smoking-related images) and smoking cessation [
Positioned at the intersection of a number of evolving interest areas, this interdisciplinary study offers a number of strengths. One strength was presenting the personalized feedback pertaining to both movement behaviors and physiology to participants. These components were objectively measured during baseline and follow-up using novel self-monitoring technologies, obtaining data to directly inform the feedback. In addition, the information that was presented in the fMRI tasks were designed based on feedback commonly presented via wearable devices or smartphone apps, reflecting what could be received in real-time in a real-world setting. Objective quantification of behavior at follow-up permitted the assessment of behavior following exposure and associations between neural activation and behavior.
Limitations of our study include the situation that participants viewed their glucose levels during baseline wear, an unavoidable situation given intentions to minimize data loss. This protocol confirms that participants had prior exposure to the glucose-related feedback subsequently presented during fMRI. However, to help try and account for this, analysis included the number of scans as a covariate because we thought the number of scans suggested the frequency with which participants viewed their glucose levels (eg, more scans equaled more exposure and so a greater awareness of their glucose levels). In addition, a lack of behavior change could be attributable to the sample that we recruited (ie, well educated and relatively healthy) and as such they could be profiled as a highly motivated audience who may not have viewed their behavior as in need of improvement. Furthermore, our unpowered sample size was another limitation, as we are unable to offer definitive interpretation of the findings. In addition, because of the number of people as active and inactive, we were unable to make any comparisons between groups of participants (eg, patterns of brain activation between those most active and least active). Finally, the pattern of neural activity observed and related psychological processes should be interpreted with caution due to the nature of reverse inference [
This multidisciplinary study highlighted that fMRI can be used to assess the neural response to personalized health feedback. In particular, greater activation in the prefrontal cortex during exposure to glucose compared with behavioral feedback was observed. A reduction in time spent sedentary and a negative association between the parameter estimates and subsequent minutes of MVPA were observed. Future research deploying behavioral feedback in parallel with physiological feedback to encourage positive behavior change is warranted.
analysis of covariance
counts per minute
functional magnetic resonance imaging
fast spoiled gradient echo
family-wise error
general linear model
hemoglobin A1c
hemisphere
light-to-vigorous physical activity
metabolic equivalent of task
Montreal Neurological Institute
medial prefrontal cortex
moderate-to-vigorous physical activity
National Centre for Sports and Exercise Medicine
National Institute for Health Research
region of interest
statistical parametric mapping
Wake Forest University
The authors thank all the participants for taking part in this study. The authors would also like to thank Julie Thompson (senior radiographer, computed tomography/MRI, University Hospitals of Leicester) for her technical support and Dr Mark Orme (Centre for Exercise and Rehabilitation Science, Respiratory Biomedical Research Unit, Glenfield Hospital) and Dr Ruth Trethewey (School of Sport, Exercise and Health Sciences, Loughborough University) for their assistance with data collection. This work was supported by the School of Sport, Exercise and Health Sciences, the National Centre for Sport and Exercise Medicine (NCSEM) at Loughborough University, and the National Institute for Health Research (NIHR) Leicester Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the National Health Service, NIHR, NCSEM England, the Department of Health, or the partners involved.
None declared.