Reproducibility of fMRI activations associated with auditory sentence comprehension

Reliability of BOLD signals in chronic stroke-induced aphasia

Investigating the neurobiology of language impairment and treatment in chronic stroke aphasia using fMRI requires an understanding of measurement variability within and between participants. In this multicenter study, we evaluated the scan-rescan reliability of an auditory and visual (written) story comprehension paradigm in stroke participants with aphasia (N = 65) and healthy controls (N = 22). The multi-modal task was conducted twice (~1 week apart) on separate visits upon study enrolment and twice again at completion three months later. A non-language visuomotor task was studied in the aphasia group only, which was conducted once per time point (3 months apart). While participants were asked to make responses during the comprehension task, these in-scanner responses were not recorded. Reliability was assessed using intraclass correlation coefficient (ICC) at both group and individual participant levels. The visual story comprehension condition had higher reliability than the auditory condition in both groups, with participants with aphasia exhibiting lower reliability than controls in both conditions (stroke ICC = .43, healthy ICC = .81). Differences in reliability within the group of participants with aphasia were found to be partially explained by overall language impairment as well as greater head motion. In the participants with aphasia, the visuomotor paradigm was found to have greater reliability than the story comprehension task at equivalent interscan intervals (visuomotor = 0.50, comprehension = 0.34), and its reliability was not associated with language impairment. This work highlights the importance of considering the reliability of fMRI tasks in aphasia research, provides strategies to improve reliability and has potential implications for the field of clinical neuroimaging in general.

Quality assurance of human functional magnetic resonance imaging: a literature review

Functional magnetic resonance imaging (fMRI) has been a popular approach in brain research over the past 20 years. It offers a noninvasive method to probe the brain and uses blood oxygenation level dependent (BOLD) signal changes to access brain function. However, the BOLD signal only represents a small fraction of the total MR signal. System instability and various noise have a strong impact on the BOLD signal. Additionally, fMRI applies fast imaging technique to record brain cognitive process over time, requiring high temporal stability of MR scanners. Furthermore, data acquisition, image quality, processing, and statistical analysis methods also have a great effect on the results of fMRI studies. Quality assurance (QA) programs for fMRI can test the stability of MR scanners, evaluate the quality of fMRI and help to find errors during fMRI scanning, thereby greatly enhancing the success rate of fMRI. In this review, we focus on previous studies which developed QA programs and methods in SCI/SCIE citation peer-reviewed publications over the last 20 years, including topics on existing fMRI QA programs, QA phantoms, image QA metrics, quality evaluation of existing preprocessing pipelines and fMRI statistical analysis methods. The summarized studies were classified into four categories: QA of fMRI systems, QA of fMRI data, quality evaluation of data processing pipelines and statistical methods and QA of task-related fMRI. Summary tables and figures of QA programs and metrics have been developed based on the comprehensive review of the literature.

Social cognition in schizophrenia: Validation of an ecological fMRI task

Neuroimaging studies have revealed brain regions involved in social cognition, which reportedly show functional alterations in schizophrenia. However, the social neural network has not been investigated with regards to language perception and social interactions in daily life. Here we developed and validated an integrative fMRI task to explore the neural basis of social cognition with regards to language perception in schizophrenia. The task comprised listening to film extracts and inferring mental states to characters. We first identified the functional network activated during the task in 28 healthy controls (HC). Next, we evaluated the reproducibility of Blood-Oxygen-Level Dependent (BOLD) variations in 14 HC participants. Finally, we investigated network impairment in 20 patients with schizophrenia (SZ) compared to HC. The HC group exhibited bilateral activation in the superior and middle temporal gyri (including the poles and the temporo-parietal junction). Overall, our novel integrative task induced activation of a functional network with good reproducibility and involved in language conveying social information. Compared to the HC group, the SZ group showed decreased recruitment of the right temporo-parietal junction. These findings may be useful for testing the impact of remediation on the brain, particularly on the network of language conveying social information.

Test-retest reliability in an fMRI study of naming in dementia

fMRI has been used as an outcome measure in dementia treatment studies, with many previous studies comparing only single pre- and post-treatment fMRI scans to determine treatment-induced neural changes, while utilizing single subject experimental designs. The purpose of the current study was to evaluate fMRI test-retest reliability in dementia patients and typical older adults using noun and verb confrontation naming to evaluate the validity of using a single pre/post-treatment scan comparison. Seven individuals with dementia and 9 control participants were tested three times over two months using the same fMRI procedures. Differences in individual and group level activation patterns were observed that varied across time. Additionally, the extent of variability fluctuated across individuals, groups, and the grammatical category of target words. Our findings suggested that one time fMRI scanning may inadequately represent an individual's typical brain activation pattern, particularly an individual with dementia. Thus, multiple imaging baselines are recommended.

Small sample sizes reduce the replicability of task-based fMRI studies

Despite a growing body of research suggesting that task-based functional magnetic resonance imaging (fMRI) studies often suffer from a lack of statistical power due to too-small samples, the proliferation of such underpowered studies continues unabated. Using large independent samples across eleven tasks, we demonstrate the impact of sample size on replicability, assessed at different levels of analysis relevant to fMRI researchers. We find that the degree of replicability for typical sample sizes is modest and that sample sizes much larger than typical (e.g., N = 100) produce results that fall well short of perfectly replicable. Thus, our results join the existing line of work advocating for larger sample sizes. Moreover, because we test sample sizes over a fairly large range and use intuitive metrics of replicability, our hope is that our results are more understandable and convincing to researchers who may have found previous results advocating for larger samples inaccessible.

Short- and long-term reliability of language fMRI

When using functional magnetic resonance imaging (fMRI) for mapping important language functions, a high test-retest reliability is mandatory, both in basic scientific research and for clinical applications. We, therefore, systematically tested the short- and long-term reliability of fMRI in a group of healthy subjects using a picture naming task and a sparse-sampling fMRI protocol. We hypothesized that test-retest reliability might be higher for (i) speech-related motor areas than for other language areas and for (ii) the short as compared to the long intersession interval. 16 right-handed subjects (mean age: 29 years) participated in three sessions separated by 2-6 (session 1 and 2, short-term) and 21-34 days (session 1 and 3, long-term). Subjects were asked to perform the same overt picture naming task in each fMRI session (50 black-white images per session). Reliability was tested using the following measures: (i) Euclidean distances (ED) between local activation maxima and Centers of Gravity (CoGs), (ii) overlap volumes and (iii) voxel-wise intraclass correlation coefficients (ICCs). Analyses were performed for three regions of interest which were chosen based on whole-brain group data: primary motor cortex (M1), superior temporal gyrus (STG) and inferior frontal gyrus (IFG). Our results revealed that the activation centers were highly reliable, independent of the time interval, ROI or hemisphere with significantly smaller ED for the local activation maxima (6.45 ± 1.36 mm) as compared to the CoGs (8.03 ± 2.01 mm). In contrast, the extent of activation revealed rather low reliability values with overlaps ranging from 24% (IFG) to 56% (STG). Here, the left hemisphere showed significantly higher overlap volumes than the right hemisphere. Although mean ICCs ranged between poor (ICC<0.5) and moderate (ICC 0.5-0.74) reliability, highly reliable voxels (ICC>0.75) were found for all ROIs. Voxel-wise reliability of the different ROIs was influenced by the intersession interval. Taken together, we could show that, despite of considerable ROI-dependent variations of the extent of activation over time, highly reliable centers of activation can be identified using an overt picture naming paradigm.

Variance decomposition for single-subject task-based fMRI activity estimates across many sessions

Here we report an exploratory within-subject variance decomposition analysis conducted on a task-based fMRI dataset with an unusually large number of repeated measures (i.e., 500 trials in each of three different subjects) distributed across 100 functional scans and 9 to 10 different sessions. Within-subject variance was segregated into four primary components: variance across-sessions, variance across-runs within a session, variance across-blocks within a run, and residual measurement/modeling error. Our results reveal inhomogeneous and distinct spatial distributions of these variance components across significantly active voxels in grey matter. Measurement error is dominant across the whole brain. Detailed evaluation of the remaining three components shows that across-session variance is the second largest contributor to total variance in occipital cortex, while across-runs variance is the second dominant source for the rest of the brain. Network-specific analysis revealed that across-block variance contributes more to total variance in higher-order cognitive networks than in somatosensory cortex. Moreover, in some higher-order cognitive networks across-block variance can exceed across-session variance. These results help us better understand the temporal (i.e., across blocks, runs and sessions) and spatial distributions (i.e., across different networks) of within-subject natural variability in estimates of task responses in fMRI. They also suggest that different brain regions will show different natural levels of test-retest reliability even in the absence of residual artifacts and sufficiently high contrast-to-noise measurements. Further confirmation with a larger sample of subjects and other tasks is necessary to ensure generality of these results.

The modulatory effect of semantic familiarity on the audiovisual integration of face-name pairs

To recognize individuals, the brain often integrates audiovisual information from familiar or unfamiliar faces, voices, and auditory names. To date, the effects of the semantic familiarity of stimuli on audiovisual integration remain unknown. In this functional magnetic resonance imaging (fMRI) study, we used familiar/unfamiliar facial images, auditory names, and audiovisual face-name pairs as stimuli to determine the influence of semantic familiarity on audiovisual integration. First, we performed a general linear model analysis using fMRI data and found that audiovisual integration occurred for familiar congruent and unfamiliar face-name pairs but not for familiar incongruent pairs. Second, we decoded the familiarity categories of the stimuli (familiar vs. unfamiliar) from the fMRI data and calculated the reproducibility indices of the brain patterns that corresponded to familiar and unfamiliar stimuli. The decoding accuracy rate was significantly higher for familiar congruent versus unfamiliar face-name pairs (83.2%) than for familiar versus unfamiliar faces (63.9%) and for familiar versus unfamiliar names (60.4%). This increase in decoding accuracy was not observed for familiar incongruent versus unfamiliar pairs. Furthermore, compared with the brain patterns associated with facial images or auditory names, the reproducibility index was significantly improved for the brain patterns of familiar congruent face-name pairs but not those of familiar incongruent or unfamiliar pairs. Our results indicate the modulatory effect that semantic familiarity has on audiovisual integration. Specifically, neural representations were enhanced for familiar congruent face-name pairs compared with visual-only faces and auditory-only names, whereas this enhancement effect was not observed for familiar incongruent or unfamiliar pairs. Hum Brain Mapp 37:4333-4348, 2016. © 2016 Wiley Periodicals, Inc.

Validity and reliability of four language mapping paradigms

Language areas of the brain can be mapped in individual participants with functional MRI. We investigated the validity and reliability of four language mapping paradigms that may be appropriate for individuals with acquired aphasia: sentence completion, picture naming, naturalistic comprehension, and narrative comprehension. Five neurologically normal older adults were scanned on each of the four paradigms on four separate occasions. Validity was assessed in terms of whether activation patterns reflected the known typical organization of language regions, that is, lateralization to the left hemisphere, and involvement of the left inferior frontal gyrus and the left middle and/or superior temporal gyri. Reliability (test-retest reproducibility) was quantified in terms of the Dice coefficient of similarity, which measures overlap of activations across time points. We explored the impact of different absolute and relative voxelwise thresholds, a range of cluster size cutoffs, and limitation of analyses to a priori potential language regions. We found that the narrative comprehension and sentence completion paradigms offered the best balance of validity and reliability. However, even with optimal combinations of analysis parameters, there were many scans on which known features of typical language organization were not demonstrated, and test-retest reproducibility was only moderate for realistic parameter choices. These limitations in terms of validity and reliability may constitute significant limitations for many clinical or research applications that depend on identifying language regions in individual participants.

•

Validity and reliability were investigated for four language mapping paradigms.

•

Narrative comprehension and sentence completion paradigms performed best.

•

Lateralization to the left hemisphere was not always apparent.

•

Test-retest reproducibility was only moderate.

Functional imaging of the thalamus in language

Herein, the literature regarding functional imaging of the thalamus during language tasks is reviewed. Fifty studies met criteria for analysis. Two of the most common task paradigms associated with thalamic activation were generative tasks (e.g. word or sentence generation) and naming, though activation was also seen in tasks that involve lexical decision, reading and working memory. Typically, thalamic activation was seen bilaterally, left greater than right, along with activation in frontal and temporal cortical regions. Thalamic activation was seen with perceptually challenging tasks, though few studies rigorously correlated thalamic activation with measures of attention or task difficulty. The peaks of activation loci were seen in virtually all thalamic regions, with a bias towards left-sided and midline activation. These analyses suggest that the thalamus may be involved in processes that involve manipulations of lexical information, but point to the need for more systematic study of the thalamus using language tasks.

Neural correlates of adaptation in freely-moving normal hearing subjects under cochlear implant acoustic simulations

Neurobiological correlates of adaptation to spectrally degraded speech were investigated with fMRI before and after exposure to a portable real-time speech processor that implements an acoustic simulation model of a cochlear implant (CI). The speech processor, in conjunction with isolating insert earphones and a microphone to capture environment sounds, was worn by participants over a two week chronic exposure period. fMRI and behavioral speech comprehension testing were conducted before and after this two week period. After using the simulator each day for 2h, participants significantly improved in word and sentence recognition scores. fMRI shows that these improvements came accompanied by changes in patterns of neuronal activation. In particular, we found additional recruitment of visual, motor, and working memory areas after the perceptual training period. These findings suggest that the human brain is able to adapt in a short period of time to a degraded auditory signal under a natural learning environment, and gives insight on how a CI might interact with the central nervous system. This paradigm can be furthered to investigate neural correlates of new rehabilitation, training, and signal processing strategies non-invasively in normal hearing listeners to improve CI patient outcomes.

Modulation of auditory cortex response to pitch variation following training with microtonal melodies

We tested changes in cortical functional response to auditory patterns in a configural learning paradigm. We trained 10 human listeners to discriminate micromelodies (consisting of smaller pitch intervals than normally used in Western music) and measured covariation in blood oxygenation signal to increasing pitch interval size in order to dissociate global changes in activity from those specifically associated with the stimulus feature that was trained. A psychophysical staircase procedure with feedback was used for training over a 2-week period. Behavioral tests of discrimination ability performed before and after training showed significant learning on the trained stimuli, and generalization to other frequencies and tasks; no learning occurred in an untrained control group. Before training the functional MRI data showed the expected systematic increase in activity in auditory cortices as a function of increasing micromelody pitch interval size. This function became shallower after training, with the maximal change observed in the right posterior auditory cortex. Global decreases in activity in auditory regions, along with global increases in frontal cortices also occurred after training. Individual variation in learning rate was related to the hemodynamic slope to pitch interval size, such that those who had a higher sensitivity to pitch interval variation prior to learning achieved the fastest learning. We conclude that configural auditory learning entails modulation in the response of auditory cortex to the trained stimulus feature. Reduction in blood oxygenation response to increasing pitch interval size suggests that fewer computational resources, and hence lower neural recruitment, is associated with learning, in accord with models of auditory cortex function, and with data from other modalities.

Design choices in imaging speech comprehension: an Activation Likelihood Estimation (ALE) meta-analysis

The localisation of spoken language comprehension is debated extensively: is processing located anterior or posterior on the left temporal lobe, and is it left- or bilaterally organised? An Activation Likelihood Estimation (ALE) analysis was conducted on functional MRI and PET studies investigating speech comprehension to identify the neural network involved in comprehension processing. Furthermore, the analysis aimed to establish the effect of four design choices (scanning paradigm, non-speech baseline, the presence of a task, and the type of stimulus material) on this comprehension network. The analysis included 57 experiments contrasting intelligible with less intelligible or unintelligible stimuli. A large comprehension network was found across bilateral Superior Temporal Sulcus (STS), Middle Temporal Gyrus (MTG) and Superior Temporal (STS) bilaterally, in left Inferior Frontal Gyrus (IFG), left Precentral Gyrus, and Supplementary Motor Area (SMA) and pre-SMA. The core network for post-lexical processing was restricted to the temporal lobes bilaterally with the highest ALE values located anterior to Heschl's Gyrus. Activations in the ALE comprehension network outside the temporal lobes (left IFG, SMA/pre-SMA, and Precentral Gyrus) were driven by the use of sentences instead of words, the scanning paradigm, or the type of non-speech baseline.

The neural bases of difficult speech comprehension and speech production: Two Activation Likelihood Estimation (ALE) meta-analyses

The role of speech production mechanisms in difficult speech comprehension is the subject of on-going debate in speech science. Two Activation Likelihood Estimation (ALE) analyses were conducted on neuroimaging studies investigating difficult speech comprehension or speech production. Meta-analysis 1 included 10 studies contrasting comprehension of less intelligible/distorted speech with more intelligible speech. Meta-analysis 2 (21 studies) identified areas associated with speech production. The results indicate that difficult comprehension involves increased reliance of cortical regions in which comprehension and production overlapped (bilateral anterior Superior Temporal Sulcus (STS) and anterior Supplementary Motor Area (pre-SMA)) and in an area associated with intelligibility processing (left posterior MTG), and second involves increased reliance on cortical areas associated with general executive processes (bilateral anterior insulae). Comprehension of distorted speech may be supported by a hybrid neural mechanism combining increased involvement of areas associated with general executive processing and areas shared between comprehension and production.

Neural adaptation and perceptual learning using a portable real-time cochlear implant simulator in natural environments

A portable real-time speech processor that implements an acoustic simulation model of a cochlear implant (CI) has been developed on the Apple iPhone / iPod Touch to permit testing and experimentation under extended exposure in real-world environments. This simulator allows for both a variable number of noise band channels and electrode insertion depth. Utilizing this portable CI simulator, we tested perceptual learning in normal hearing listeners by measuring word and sentence comprehension behaviorally before and after 2 weeks of exposure. To evaluate changes in neural activation related to adaptation to transformed speech, fMRI was also conducted. Differences in brain activation after training occurred in the inferior frontal gyrus and areas related to language processing. A 15-20% improvement in word and sentence comprehension of cochlear implant simulated speech was also observed. These results demonstrate the effectiveness of a portable CI simulator as a research tool and provide new information about the physiological changes that accompany perceptual learning of degraded auditory input.

Advances in longitudinal MRI diagnostic tests

INTRODUCTION

The past decade has seen an explosion of functional magnetic resonance imaging (MRI) studies in neuroscience. As the technology progresses, it is now possible to carry out longitudinal studies using functional MRI. Such studies can be used to understand the progression of mental and neurological disorders and the effectiveness of different treatments by obtaining direct measures of brain activity as well as markers of tissue health and connectivity.

AREAS COVERED

We review six popular neuroimaging tools that can be used for longitudinal studies: blood oxygen level-dependent (BOLD)-weighted imaging, BOLD-based functional connectivity, arterial spin labeling, dynamic R2* imaging, voxel-based morphometry, and diffusion tensor imaging.

EXPERT OPINION

Each of these techniques is targeted to probe a specific feature of brain function or brain structure and can reveal important information about the progression of a pathological condition. We anticipate that in the near future, the MRI techniques discussed here may become standard tools in clinical use and will not be used for research purposes only.

Whole-brain, time-locked activation with simple tasks revealed using massive averaging and model-free analysis

The brain is the body's largest energy consumer, even in the absence of demanding tasks. Electrophysiologists report on-going neuronal firing during stimulation or task in regions beyond those of primary relationship to the perturbation. Although the biological origin of consciousness remains elusive, it is argued that it emerges from complex, continuous whole-brain neuronal collaboration. Despite converging evidence suggesting the whole brain is continuously working and adapting to anticipate and actuate in response to the environment, over the last 20 y, task-based functional MRI (fMRI) have emphasized a localizationist view of brain function, with fMRI showing only a handful of activated regions in response to task/stimulation. Here, we challenge that view with evidence that under optimal noise conditions, fMRI activations extend well beyond areas of primary relationship to the task; and blood-oxygen level-dependent signal changes correlated with task-timing appear in over 95% of the brain for a simple visual stimulation plus attention control task. Moreover, we show that response shape varies substantially across regions, and that whole-brain parcellations based on those differences produce distributed clusters that are anatomically and functionally meaningful, symmetrical across hemispheres, and reproducible across subjects. These findings highlight the exquisite detail lying in fMRI signals beyond what is normally examined, and emphasize both the pervasiveness of false negatives, and how the sparseness of fMRI maps is not a result of localized brain function, but a consequence of high noise and overly strict predictive response models.

Biomechanical correlates of symptomatic and asymptomatic neurophysiological impairment in high school football

Concussion is a growing public health issue in the United States, and chronic traumatic encephalopathy (CTE) is the chief long-term concern linked to repeated concussions. Recently, attention has shifted toward subconcussive blows and the role they may play in the development of CTE. We recruited a cohort of high school football players for two seasons of observation. Acceleration sensors were placed in the helmets, and all contact activity was monitored. Pre-season computer-based neuropsychological tests and functional magnetic resonance imaging (fMRI) tests were also obtained in order to assess cognitive and neurophysiological health. In-season follow-up scans were then obtained both from individuals who had sustained a clinically-diagnosed concussion and those who had not. These changes were then related through stepwise regression to history of blows recorded throughout the football season up to the date of the scan. In addition to those subjects who had sustained a concussion, a substantial portion of our cohort who did not sustain concussions showed significant neurophysiological changes. Stepwise regression indicated significant relationships between the number of blows sustained by a subject and the ensuing neurophysiological change. Our findings reinforce the hypothesis that the effects of repetitive blows to the head are cumulative and that repeated exposure to subconcussive blows is connected to pathologically altered neurophysiology.

Effects of combining field strengths on auditory functional MRI group analysis: 1.5T and 3T

PURPOSE

To evaluate effects of combining functional magnetic resonance imaging (fMRI) data acquired from different field strengths on group analysis as a function of the number of subjects at each field strength.

MATERIALS AND METHODS

In all, 28 subjects (18 at 3T) participated in an auditory task of passively listening to a 0.75s segment of jazz music in an event-related design. Results of single-subject analysis were combined to create all possible subject combinations for a group size of eight subjects from each of the 3T and 1.5T pools, comprising subject mixtures of (3T/1.5T) 0/8, 2/6, 4/4, 6/2, and 8/0. Group analysis performance of each subject permutation was measured by receiver operating characteristic (ROC) curves and activation overlap maps.

RESULTS

While area under ROC curves, extent of activation in the gold standard region, and reliability of activation increased with the number of 3T subjects, marginal gain decreased. ROC performance overlap across mixtures was observed, indicating that some combinations of subjects markedly outperformed others. For detection of activation, 4/4 was arguably the minimum mixture level that was comparable to 3T-only group results.

CONCLUSION

Inclusion of 1.5T data does not necessarily reduce the validity of group analysis. Lower field strength data was found only to limit detection power, but did not affect specificity. Within the limits of realignment error, these results should also extend to group longitudinal analyses of subject mixtures from different field strengths.