To pool or not to pool: Can we ignore cross-trial variability in FMRI?

Test-retest reliability of the attention network test from the perspective of intrinsic network organization

The attention network test (ANT), developed based on the triple-network taxonomy by Posner and colleagues, has been widely used to examine the efficacy of alerting, orienting and executive control in clinical and developmental neuroscience studies. Recent research suggests the imperfect reliability of the behavioural ANT and its variants. However, the classical ANT fMRI task's test-retest reliability has received little attention. Moreover, it remains ambiguous whether the attention-related intrinsic network components, especially the dorsal attention, ventral attention and frontoparietal network, manifest acceptable reliability. The present study approaches these issues by utilizing an openly available ANT fMRI dataset for participants with Parkinson's disease and healthy elderly. The reproducibility of group-level activations across sessions and participant groups and the test-retest reliability at the individual level were examined at the voxel, region and network levels. The intrinsic network was defined using the Yeo-Schaefer atlas. Our results reveal three critical facets: (1) the overlapping of the group-level contrast map between sessions and between participant groups was unsatisfactory; (2) the reliability of alerting, orienting and executive, defined as a contrast between conditions, was worse than estimates of specific conditions. (3) Dorsal attention, ventral attention, visual and somatomotor networks showed acceptable reliability for the congruent and incongruent conditions. Our results suggest that specific condition estimates might be used instead of the contrast map for individual or group-difference studies.

Clinical Response to Neurofeedback in Major Depression Relates to Subtypes of Whole-Brain Activation Patterns During Training

Major Depressive Disorder (MDD) poses a significant public health challenge due to its high prevalence and the substantial burden it places on individuals and healthcare systems. Real-time functional magnetic resonance imaging neurofeedback (rtfMRI-NF) shows promise as a treatment for this disorder, although its mechanisms of action remain unclear. This study investigated whole-brain response patterns during rtfMRI-NF training to explain interindividual variability in clinical efficacy in MDD. We analyzed data from 95 participants (67 active, 28 control) with MDD from previous rtfMRI-NF studies designed to increase left amygdala activation through positive autobiographical memory recall. Significant symptom reduction was observed in the active group (t=-4.404, d=-0.704, p<0.001) but not in the control group (t=-1.609, d=-0.430, p=0.111). However, left amygdala activation did not account for the variability in clinical efficacy. To elucidate the brain training process underlying the clinical effect, we examined whole-brain activation patterns during two critical phases of the neurofeedback procedure: activation during the self-regulation period, and transient responses to feedback signal presentations. Using a systematic process involving feature selection, manifold extraction, and clustering with cross-validation, we identified two subtypes of regulation activation and three subtypes of brain responses to feedback signals. These subtypes were significantly associated with the clinical effect (regulation subtype: F=8.735, p=0.005; feedback response subtype: F=5.326, p=0.008; subtypes' interaction: F=3.471, p=0.039). Subtypes associated with significant symptom reduction were characterized by selective increases in control regions, including lateral prefrontal areas, and decreases in regions associated with self-referential thinking, such as default mode areas. These findings suggest that large-scale brain activity during training is more critical for clinical efficacy than the level of activation in the neurofeedback target region itself. Tailoring neurofeedback training to incorporate these patterns could significantly enhance its therapeutic efficacy.

Estimation of fMRI responses related to epileptic discharges using Bayesian hierarchical modeling

Simultaneous electroencephalography-functional MRI (EEG-fMRI) is a unique and noninvasive method for epilepsy presurgical evaluation. When selecting voxels by null-hypothesis tests, the conventional analysis may overestimate fMRI response amplitudes related to interictal epileptic discharges (IEDs), especially when IEDs are rare. We aimed to estimate fMRI response amplitudes represented by blood oxygen level dependent (BOLD) percentage changes related to IEDs using a hierarchical model. It involves the local and distributed hemodynamic response homogeneity to regularize estimations. Bayesian inference was applied to fit the model. Eighty-two epilepsy patients who underwent EEG-fMRI and subsequent surgery were included in this study. A conventional voxel-wise general linear model was compared to the hierarchical model on estimated fMRI response amplitudes and on the concordance between the highest response cluster and the surgical cavity. The voxel-wise model overestimated fMRI responses compared to the hierarchical model, evidenced by a practically and statistically significant difference between the estimated BOLD percentage changes. Only the hierarchical model differentiated brief and long-lasting IEDs with significantly different BOLD percentage changes. Overall, the hierarchical model outperformed the voxel-wise model on presurgical evaluation, measured by higher prediction performance. When compared with a previous study, the hierarchical model showed higher performance metric values, but the same or lower sensitivity. Our results demonstrated the capability of the hierarchical model of providing more physiologically reasonable and more accurate estimations of fMRI response amplitudes induced by IEDs. To enhance the sensitivity of EEG-fMRI for presurgical evaluation, it may be necessary to incorporate more appropriate spatial priors and bespoke decision strategies.

Effects of schema on the relationship between post-encoding brain connectivity and subsequent durable memory

Schemas can facilitate memory consolidation. Studies have suggested that interactions between the hippocampus and the ventromedial prefrontal cortex (vmPFC) are important for schema-related memory consolidation. However, in humans, how schema accelerates the consolidation of new information and relates to durable memory remains unclear. To address these knowledge gaps, we used a human analogue of the rodent spatial schema task and resting-state fMRI to investigate how post-encoding brain networks can predict long-term memory performance in different schema conditions. After participants were trained to obtain schema-consistent or schema-inconsistent object-location associations, they learned new object-location associations. The new associations were tested after the post-encoding rest in the scanner and 24 h later outside the scanner. The Bayesian multilevel modelling was applied to analyse the post-encoding brain networks. The results showed that during the post-encoding, stronger vmPFC- anterior hippocampal connectivity was associated with durable memory in the schema-consistent condition, whereas stronger object-selective lateral occipital cortex (LOC)-ventromedial prefrontal connectivity and weaker connectivity inside the default mode network were associated with durable memory in the schema inconsistent condition. In addition, stronger LOC-anterior hippocampal connectivity was associated with memory in both schema conditions. These results shed light on how schemas reconfigure early brain networks, especially the prefrontal-hippocampal and stimuli-relevant cortical networks and influence long-term memory performance.

Evaluating increases in sensitivity from NORDIC for diverse fMRI acquisition strategies

As the neuroimaging field moves towards detecting smaller effects at higher spatial resolutions, and faster sampling rates, there is increased attention given to the deleterious contribution of unstructured, thermal noise. Here, we critically evaluate the performance of a recently developed reconstruction method, termed NORDIC, for suppressing thermal noise using datasets acquired with various field strengths, voxel sizes, sampling rates, and task designs. Following minimal preprocessing, statistical activation (t-values) of NORDIC processed data was compared to the results obtained with alternative denoising methods. Additionally, we examined the consistency of the estimates of task responses at the single-voxel, single run level, using a finite impulse response (FIR) model. To examine the potential impact on effective image resolution, the overall smoothness of the data processed with different methods was estimated. Finally, to determine if NORDIC alters or removes temporal information important for modeling responses, we employed an exhaustive leave-p-out cross validation approach, using FIR task responses to predict held out timeseries, quantified using R². After NORDIC, the t-values are increased, an improvement comparable to what could be achieved by 1.5 voxels smoothing, and task events are clearly visible and have less cross-run error. These advantages are achieved with smoothness estimates increasing by less than 4%, while 1.5 voxel smoothing is associated with increases of over 140%. Cross-validated R²s based on the FIR models show that NORDIC is not measurably distorting the temporal structure of the data under this approach and is the best predictor of non-denoised time courses. The results demonstrate that analyzing 1 run of data after NORDIC produces results equivalent to using 2 to 3 original runs and that NORDIC performs equally well across a diverse array of functional imaging protocols. Significance Statement: For functional neuroimaging, the increasing availability of higher field strengths and ever higher spatiotemporal resolutions has led to concomitant increase in concerns about the deleterious effects of thermal noise. Historically this noise source was suppressed using methods that reduce spatial precision such as image blurring or averaging over a large number of trials or sessions, which necessitates large data collection efforts. Here, we critically evaluate the performance of a recently developed reconstruction method, termed NORDIC, which suppresses thermal noise. Across datasets varying in field strength, voxel sizes, sampling rates, and task designs, NORDIC produces substantial gains in data quality. Both conventional t-statistics derived from general linear models and coefficients of determination for predicting unseen data are improved. These gains match or even exceed those associated with 1 voxel Full Width Half Max image smoothing, however, even such small amounts of smoothing are associated with a 52% reduction in estimates of spatial precision, whereas the measurable difference in spatial precision is less than 4% following NORDIC.

Brain and behavioral contributions to individual choices in response to affective-cognitive persuasion

Affective and cognitive information conveyed by persuasive stimuli is evaluated and integrated by individuals according to their behavioral predispositions. However, the neurocognitive structure that supports persuasion based on either affective or cognitive content is poorly understood. Here, we examine the neural and behavioral processes supporting choices based on affective and cognitive persuasion by integrating 4 information processing features: intrinsic brain connectivity, stimulus-evoked brain activity, intrinsic affective-cognitive orientation, and explicit target evaluations. We found that the intrinsic cross-network connections of a multimodal fronto-parietal network are associated with individual affective-cognitive orientation. Moreover, using a cross-validated classifier, we found that individuals' intrinsic brain-behavioral dimensions, such as affective-cognitive orientation and intrinsic brain connectivity, can predict individual choices between affective and cognitive targets. Our findings show that affective- and cognitive-based choices rely on multiple sources, including behavioral orientation, stimulus evaluation, and intrinsic functional brain architecture.

Hierarchical Bayesian modeling of the relationship between task-related hemodynamic responses and cortical excitability

Investigating the relationship between task-related hemodynamic responses and cortical excitability is challenging because it requires simultaneous measurement of hemodynamic responses while applying noninvasive brain stimulation. Moreover, cortical excitability and task-related hemodynamic responses are both associated with inter-/intra-subject variability. To reliably assess such a relationship, we applied hierarchical Bayesian modeling. This study involved 16 healthy subjects who underwent simultaneous Paired Associative Stimulation (PAS10, PAS25, Sham) while monitoring brain activity using functional Near-Infrared Spectroscopy (fNIRS), targeting the primary motor cortex (M1). Cortical excitability was measured by Motor Evoked Potentials (MEPs), and the motor task-related hemodynamic responses were measured using fNIRS 3D reconstructions. We constructed three models to investigate: (1) PAS effects on the M1 excitability, (2) PAS effects on fNIRS hemodynamic responses to a finger tapping task, and (3) the correlation between PAS effects on M1 excitability and PAS effects on task-related hemodynamic responses. Significant increase in cortical excitability was found following PAS25, whereas a small reduction of the cortical excitability was shown after PAS10 and a subtle increase occurred after sham. Both HbO and HbR absolute amplitudes increased after PAS25 and decreased after PAS10. The probability of the positive correlation between modulation of cortical excitability and hemodynamic activity was 0.77 for HbO and 0.79 for HbR. We demonstrated that PAS stimulation modulates task-related cortical hemodynamic responses in addition to M1 excitability. Moreover, the positive correlation between PAS modulations of excitability and hemodynamics brought insight into understanding the fundamental properties of cortical function and cortical excitability.

Reliability of task-evoked neural activation during face-emotion paradigms: Effects of scanner and psychological processes

Assessing and improving test–retest reliability is critical to efforts to address concerns about replicability of task‐based functional magnetic resonance imaging. The current study uses two statistical approaches to examine how scanner and task‐related factors influence reliability of neural response to face‐emotion viewing. Forty healthy adult participants completed two face‐emotion paradigms at up to three scanning sessions across two scanners of the same build over approximately 2 months. We examined reliability across the main task contrasts using Bayesian linear mixed‐effects models performed voxel‐wise across the brain. We also used a novel Bayesian hierarchical model across a predefined whole‐brain parcellation scheme and subcortical anatomical regions. Scanner differences accounted for minimal variance in temporal signal‐to‐noise ratio and task contrast maps. Regions activated during task at the group level showed higher reliability relative to regions not activated significantly at the group level. Greater reliability was found for contrasts involving conditions with clearly distinct visual stimuli and associated cognitive demands (e.g., face vs. nonface discrimination) compared to conditions with more similar demands (e.g., angry vs. happy face discrimination). Voxel‐wise reliability estimates tended to be higher than those based on predefined anatomical regions. This work informs attempts to improve reliability in the context of task activation patterns and specific task contrasts. Our study provides a new method to estimate reliability across a large number of regions of interest and can inform researchers' selection of task conditions and analytic contrasts.

Effect sizes and test-retest reliability of the fMRI-based neurologic pain signature

Identifying biomarkers that predict mental states with large effect sizes and high test-retest reliability is a growing priority for fMRI research. We examined a well-established multivariate brain measure that tracks pain induced by nociceptive input, the Neurologic Pain Signature (NPS). In N = 295 participants across eight studies, NPS responses showed a very large effect size in predicting within-person single-trial pain reports (d = 1.45) and medium effect size in predicting individual differences in pain reports (d = 0.49). The NPS showed excellent short-term (within-day) test-retest reliability (ICC = 0.84, with average 69.5 trials/person). Reliability scaled with the number of trials within-person, with ≥60 trials required for excellent test-retest reliability. Reliability was tested in two additional studies across 5-day (N = 29, ICC = 0.74, 30 trials/person) and 1-month (N = 40, ICC = 0.46, 5 trials/person) test-retest intervals. The combination of strong within-person correlations and only modest between-person correlations between the NPS and pain reports indicate that the two measures have different sources of between-person variance. The NPS is not a surrogate for individual differences in pain reports but can serve as a reliable measure of pain-related physiology and mechanistic target for interventions.

Striving toward translation: strategies for reliable fMRI measurement

fMRI has considerable potential as a translational tool for understanding risk, prioritizing interventions, and improving the treatment of brain disorders. However, recent studies have found that many of the most widely used fMRI measures have low reliability, undermining this potential. Here, we argue that many fMRI measures are unreliable because they were designed to identify group effects, not to precisely quantify individual differences. We then highlight four emerging strategies [extended aggregation, reliability modeling, multi-echo fMRI (ME-fMRI), and stimulus design] that build on established psychometric properties to generate more precise and reliable fMRI measures. By adopting such strategies to improve reliability, we are optimistic that fMRI can fulfill its potential as a clinical tool.

Migraine in Multiple Sclerosis Patients Affects Functional Connectivity of the Brain Circuitry Involved in Pain Processing

Migraine is particularly common in patients with multiple sclerosis (MS) and has been linked to the dysfunction of the brain circuitry modulating the peripheral nociceptive stimuli. Using MRI, we explored whether changes in the resting state-functional connectivity (RS-FC) may characterize the occurrence of migraine in patients with MS. The RS-FC characteristics in concerned brain regions were explored in 20 MS patients with migraine (MS+M) during the interictal phase, and compared with 19 MS patients without migraine (MS-M), which served as a control group. Functional differences were correlated to the frequency and severity of previous migraine attacks, and with the resulting impact on daily activities. In MS+M, the loss of periaqueductal gray matter (PAG) positive connectivity with the default mode network and the left posterior cranial pons was associated with an increase of migraine attacks frequency. In contrast, the loss of PAG negative connectivity with sensorimotor and visual network was linked to migraine symptom severity and related daily activities impact. Finally, a PAG negative connection was established with the prefrontal executive control network. Migraine in MS+M patients and its impact on daily activities, underlies RS-FC rearrangements between brain regions involved in pain perception and modulation.

Neural Coding of Cognitive Control: The Representational Similarity Analysis Approach

Cognitive control relies on distributed and potentially high-dimensional frontoparietal task representations. Yet, the classical cognitive neuroscience approach in this domain has focused on aggregating and contrasting neural measures - either via univariate or multivariate methods - along highly abstracted, 1D factors (e.g., Stroop congruency). Here, we present representational similarity analysis (RSA) as a complementary approach that can powerfully inform representational components of cognitive control theories. We review several exemplary uses of RSA in this regard. We further show that most classical paradigms, given their factorial structure, can be optimized for RSA with minimal modification. Our aim is to illustrate how RSA can be incorporated into cognitive control investigations to shed new light on old questions.

Beyond linearity in neuroimaging: Capturing nonlinear relationships with application to longitudinal studies

The ubiquitous adoption of linearity for quantitative predictors in statistical modeling is likely attributable to its advantages of straightforward interpretation and computational feasibility. The linearity assumption may be a reasonable approximation especially when the variable is confined within a narrow range, but it can be problematic when the variable's effect is non-monotonic or complex. Furthermore, visualization and model assessment of a linear fit are usually omitted because of challenges at the whole brain level in neuroimaging. By adopting a principle of learning from the data in the presence of uncertainty to resolve the problematic aspects of conventional polynomial fitting, we introduce a flexible and adaptive approach of multilevel smoothing splines (MSS) to capture any nonlinearity of a quantitative predictor for population-level neuroimaging data analysis. With no prior knowledge regarding the underlying relationship other than a parsimonious assumption about the extent of smoothness (e.g., no sharp corners), we express the unknown relationship with a sufficient number of smoothing splines and use the data to adaptively determine the specifics of the nonlinearity. In addition to introducing the theoretical framework of MSS as an efficient approach with a counterbalance between flexibility and stability, we strive to (a) lay out the specific schemes for population-level nonlinear analyses that may involve task (e.g., contrasting conditions) and subject-grouping (e.g., patients vs controls) factors; (b) provide modeling accommodations to adaptively reveal, estimate and compare any nonlinear effects of a predictor across the brain, or to more accurately account for the effects (including nonlinear effects) of a quantitative confound; (c) offer the associated program 3dMSS to the neuroimaging community for whole-brain voxel-wise analysis as part of the AFNI suite; and (d) demonstrate the modeling approach and visualization processes with a longitudinal dataset of structural MRI scans.

The longitudinal stability of fMRI activation during reward processing in adolescents and young adults

BACKGROUND

The use of functional neuroimaging has been an extremely fruitful avenue for investigating the neural basis of human reward function. This approach has included identification of potential neurobiological mechanisms of psychiatric disease and examination of environmental, experiential, and biological factors that may contribute to disease risk via effects on the reward system. However, a central and largely unexamined assumption of much of this research is that neural reward function is an individual difference characteristic that is relatively stable and trait-like over time.

METHODS

In two independent samples of adolescents and young adults studied longitudinally (Ns = 145 & 139, 100% female and 100% male, ages 15-21 and 20-22, 2-4 scans and 2 scans respectively), we tested within-person stability of reward-task BOLD activation, with a median of 1 and 2 years between scans. We examined multiple commonly used contrasts of active states and baseline in both the anticipation and feedback phases of a card-guessing reward task. We examined the effects of cortical parcellation resolution, contrast, network (reward regions and resting-state networks), region-size, and activation strength and variability on the stability of reward-related activation.

RESULTS

In both samples, contrasts of an active state relative to a baseline were more stable (ICC: intra-class correlation; e.g., Win>Baseline; mean ICC = 0.13 - 0.33) than contrasts of two active states (e.g., Win>Loss; mean ICC = 0.048 - 0.05). Additionally, activation in reward regions was less stable than in many non-task networks (e.g., dorsal attention), and activation in regions with greater between-subject variability showed higher stability in both samples.

CONCLUSIONS

These results show that some contrasts from functional neuroimaging activation during a card guessing reward task have partially trait-like properties in adolescent and young adult samples over 1-2 years. Notably, results suggest that contrasts intended to map cognitive function and show robust group-level effects (i.e. Win > Loss) may be less effective in studies of individual differences and disease risk. The robustness of group-level activation should be weighed against other factors when selecting regions of interest in individual difference fMRI studies.