Applying independent component analysis to clinical FMRI at 7 t

7 T and beyond: toward a synergy between fMRI-based presurgical mapping at ultrahigh magnetic fields, AI, and robotic neurosurgery

Presurgical evaluation with functional magnetic resonance imaging (fMRI) can reduce postsurgical morbidity. Here, we discuss presurgical fMRI mapping at ultra-high magnetic fields (UHF), i.e., ≥ 7 T, in the light of the current growing interest in artificial intelligence (AI) and robot-assisted neurosurgery. The potential of submillimetre fMRI mapping can help better appreciate uncertainty on resection margins, though geometric distortions at UHF might lessen the accuracy of fMRI maps. A useful trade-off for UHF fMRI is to collect data with 1-mm isotropic resolution to ensure high sensitivity and subsequently a low risk of false negatives. Scanning at UHF might yield a revival interest in slow event-related fMRI, thereby offering a richer depiction of the dynamics of fMRI responses. The potential applications of AI concern denoising and artefact removal, generation of super-resolution fMRI maps, and accurate fusion or coregistration between anatomical and fMRI maps. The latter can benefit from the use of T1-weighted echo-planar imaging for better visualization of brain activations. Such AI-augmented fMRI maps would provide high-quality input data to robotic surgery systems, thereby improving the accuracy and reliability of robot-assisted neurosurgery. Ultimately, the advancement in fMRI at UHF would promote clinically useful synergies between fMRI, AI, and robotic neurosurgery.Relevance statement This review highlights the potential synergies between fMRI at UHF, AI, and robotic neurosurgery in improving the accuracy and reliability of fMRI-based presurgical mapping.Key points• Presurgical fMRI mapping at UHF improves spatial resolution and sensitivity.• Slow event-related designs offer a richer depiction of fMRI responses dynamics.• AI can support denoising, artefact removal, and generation of super-resolution fMRI maps.• AI-augmented fMRI maps can provide high-quality input data to robotic surgery systems.

Gradual Restraint Habituation for Awake Functional Magnetic Resonance Imaging Combined With a Sparse Imaging Paradigm Reduces Motion Artifacts and Stress Levels in Rodents

Functional magnetic resonance imaging, as a non-invasive technique, offers unique opportunities to assess brain function and connectivity under a broad range of applications, ranging from passive sensory stimulation to high-level cognitive abilities, in awake animals. This approach is confounded, however, by the fact that physical restraint and loud unpredictable acoustic noise must inevitably accompany fMRI recordings. These factors induce marked stress in rodents, and stress-related elevations of corticosterone levels are known to alter information processing and cognition in the rodent. Here, we propose a habituation strategy that spans specific stages of adaptation to restraint, MRI noise, and confinement stress in awake rats and circumvents the need for surgical head restraint. This habituation protocol results in stress levels during awake fMRI that do not differ from pre-handling levels and enables stable image acquisition with very low motion artifacts. For this, rats were gradually trained over a period of three weeks and eighteen training sessions. Stress levels were assessed by analysis of fecal corticosterone metabolite levels and breathing rates. We observed significant drops in stress levels to below pre-handling levels at the end of the habituation procedure. During fMRI in awake rats, after the conclusion of habituation and using a non-invasive head-fixation device, breathing was stable and head motion artifacts were minimal. A task-based fMRI experiment, using acoustic stimulation, conducted 2 days after the end of habituation, resulted in precise whole brain mapping of BOLD signals in the brain, with clear delineation of the expected auditory-related structures. The active discrimination by the animals of the acoustic stimuli from the backdrop of scanner noise was corroborated by significant increases in BOLD signals in the thalamus and reticular formation. Taken together, these data show that effective habituation to awake fMRI can be achieved by gradual and incremental acclimatization to the experimental conditions. Subsequent BOLD recordings, even during superimposed acoustic stimulation, reflect low stress-levels, low motion and a corresponding high-quality image acquisition. Furthermore, BOLD signals obtained during fMRI indicate that effective habituation facilitates selective attention to sensory stimuli that can in turn support the discrimination of cognitive processes in the absence of stress confounds.

Both activation and deactivation of functional networks support increased sentence processing costs

The research on the neural correlates underlying the language system has gradually moved away from the traditional Broca-Wernicke framework to a network perspective in the past 15 years. Language processing is found to be supported by the co-activation of both core and peripheral brain regions. However, the dynamic co-activation patterns of these brain regions serving different language functions remain to be fully revealed. The present functional magnetic resonance imaging (fMRI) study focused on sentence processing at different syntactic complexity levels to examine how the co-activation of different brain networks will be modulated by increased processing costs. Chinese relative clauses were used to probe the two dimensions of syntactic complexity: embeddedness (left-branching vs. center-embedded) and gap-filler dependency (subject-gap vs. object-gap) using the general linear model (GLM) approach, independent component analysis (ICA) and graph theoretical analysis. In contrast to localized activation revealed by the GLM approach, ICA identified more extensive networks both positively and negatively correlated with the task. We found that the posterior default mode network was anti-correlated to the gap-filler integration costs with increased deactivation for the left-branching object relative clauses compared to subject relative clauses, suggesting the involvement of this network in leveraging the cognitive resources based on the complexity level of the language task. Concurrent activation and deactivation of networks were found to be associated with the higher costs induced by center-embedding and its interaction with gap-filler integration. The graph theoretical analysis further unveiled that center-embeddedness imposed more attentional demand on the subject relative clause, as characterized by its higher degree and strength in the ventral attention network, and higher processing costs of syntactic reanalysis on the object relative clause, as characterized by increased intermodular connections of the language network with other networks. The results suggest that network activation and deactivation profiles are modulated by different dimensions of syntactic complexity to serve the higher demand of creating a coherent semantic representation.

Overarching Principles and Dimensions of the Functional Organization in the Inferior Parietal Cortex

The parietal cortex (PC) is implicated in a confusing myriad of different cognitive processes/tasks. Consequently, understanding the nature and organization of the core underlying neurocomputations is challenging. According to the Parietal Unified Connectivity-biased Computation model, two properties underpin PC function and organization. Firstly, PC is a multidomain, context-dependent buffer of time- and space-varying input, the function of which, over time, becomes sensitive to the statistical temporal/spatial structure of events. Secondly, over and above this core buffering computation, differences in long-range connectivity will generate graded variations in task engagement across subregions. The current study tested these hypotheses using a group independent component analysis technique with two independent functional magnetic resonance imaging datasets (task and resting state data). Three functional organizational principles were revealed: Factor 1, inferior PC was sensitive to the statistical structure of sequences for all stimulus types (pictures, sentences, numbers); Factor 2, a dorsal–ventral variation in generally task-positive versus task-negative (variable) engagement; and Factor 3, an anterior–posterior dimension in inferior PC reflecting different engagement in verbal versus visual tasks, respectively. Together, the data suggest that the core neurocomputation implemented by PC is common across domains, with graded task engagement across regions reflecting variations in the connectivity of task-specific networks that interact with PC.

Salience-Default Mode Functional Network Connectivity Linked to Positive and Negative Symptoms of Schizophrenia

Schizophrenia is a complex, debilitating mental disorder characterized by wide-ranging symptoms including delusions, hallucinations (so-called positive symptoms), and impaired motor and speech/language production (so-called negative symptoms). Salience-monitoring theorists propose that abnormal functional communication between the salience network (SN) and default mode network (DMN) begets positive and negative symptoms of schizophrenia, yet prior studies have predominately reported links between disrupted SN/DMN functional communication and positive symptoms. It remains unclear whether disrupted SN/DMN functional communication explains (1) solely positive symptoms or (2) both positive and negative symptoms of schizophrenia. To address this question, we incorporate time-lag-shifted functional network connectivity (FNC) analyses that explored coherence of the resting-state functional magnetic resonance imaging signal of 3 networks (anterior DMN, posterior DMN, and SN) with fixed time lags introduced between network time series (1 TR = 2 s; 2 TR = 4 s). Multivariate linear regression analysis revealed that severity of disordered thought and attentional deficits were negatively associated with 2 TR-shifted FNC between anterior DMN and posterior DMN. Meanwhile, severity of flat affect and bizarre behavior were positively associated with 1 TR-shifted FNC between anterior DMN and SN. These results provide support favoring the hypothesis that lagged SN/DMN functional communication is associated with both positive and negative symptoms of schizophrenia.

Key clinical benefits of neuroimaging at 7T

The growing interest in ultra-high field MRI, with more than 35.000 MR examinations already performed at 7T, is related to improved clinical results with regard to morphological as well as functional and metabolic capabilities. Since the signal-to-noise ratio increases with the field strength of the MR scanner, the most evident application at 7T is to gain higher spatial resolution in the brain compared to 3T. Of specific clinical interest for neuro applications is the cerebral cortex at 7T, for the detection of changes in cortical structure, like the visualization of cortical microinfarcts and cortical plaques in Multiple Sclerosis. In imaging of the hippocampus, even subfields of the internal hippocampal anatomy and pathology may be visualized with excellent spatial resolution. Using Susceptibility Weighted Imaging, the plaque-vessel relationship and iron accumulations in Multiple Sclerosis can be visualized, which may provide a prognostic factor of disease. Vascular imaging is a highly promising field for 7T which is dealt with in a separate dedicated article in this special issue. The static and dynamic blood oxygenation level-dependent contrast also increases with the field strength, which significantly improves the accuracy of pre-surgical evaluation of vital brain areas before tumor removal. Improvement in acquisition and hardware technology have also resulted in an increasing number of MR spectroscopic imaging studies in patients at 7T. More recent parallel imaging and short-TR acquisition approaches have overcome the limitations of scan time and spatial resolution, thereby allowing imaging matrix sizes of up to 128×128. The benefits of these acquisition approaches for investigation of brain tumors and Multiple Sclerosis have been shown recently. Together, these possibilities demonstrate the feasibility and advantages of conducting routine diagnostic imaging and clinical research at 7T.

Functional MRI for Surgery of Gliomas

PURPOSE OF REVIEW

Advanced neuroimaging techniques such as functional MRI (fMRI) and diffusion MR tractography have been increasingly used at every stage of the surgical management of brain gliomas, as a means to improve tumor resection while preserving brain functions. This review provides an overview of the last advancements in the field of functional MRI techniques, with a particular focus on their current clinical use and reliability in the preoperative and intraoperative setting, as well as their future perspectives for personalized multimodal management of patients with gliomas.

RECENT FINDINGS

fMRI and diffusion MR tractography give relevant insights on the anatomo-functional organization of eloquent cortical areas and subcortical connections near or inside a tumor. Task-based fMRI and diffusion tensor imaging (DTI) tractography have proven to be valid and highly sensitive tools for localizing the distinct eloquent cortical and subcortical areas before surgery in glioma patients; they also show good accuracy when compared with intraoperative stimulation mapping data. Resting-state fMRI functional connectivity as well as new advanced HARDI (high angular resolution diffusion imaging) tractography methods are improving and reshaping the role of functional MRI for surgery of gliomas, with potential benefit for personalized treatment strategies. Noninvasive functional MRI techniques may offer the opportunity to perform a multimodal assessment in brain tumors, to be integrated with intraoperative mapping and clinical data for improving surgical management and oncological and functional outcome in patients affected by gliomas.

Robust presurgical functional MRI at 7 T using response consistency

Functional MRI is valuable in presurgical planning due to its non-invasive nature, repeatability, and broad availability. Using ultra-high field MRI increases the specificity and sensitivity, increasing the localization reliability and reducing scan time. Ideally, fMRI analysis for this application should identify unreliable runs and work even if the patient deviates from the prescribed task timing or if there are changes to the hemodynamic response due to pathology. In this study, a model-free analysis method-UNBIASED-based on the consistency of fMRI responses over runs was applied, to ultra-high field fMRI localizations of the hand area. Ten patients with brain tumors and epilepsy underwent 7 Tesla fMRI with multiple runs of a hand motor task in a block design. FMRI data were analyzed with the proposed approach (UNBIASED) and the conventional General Linear Model (GLM) approach. UNBIASED correctly identified and excluded fMRI runs that contained little or no activation. Generally, less motion artifact contamination was present in UNBIASED than in GLM results. Some cortical regions were identified as activated in UNBIASED but not GLM results. These were confirmed to show reproducible delayed or transient activation, which was time-locked to the task. UNBIASED is a robust approach to generating activation maps without the need for assumptions about response timing or shape. In presurgical planning, UNBIASED can complement model-based methods to aid surgeons in making prudent choices about optimal surgical access and resection margins for each patient, even if the hemodynamic response is modified by pathology. Hum Brain Mapp 38:3163-3174, 2017. © 2017 The Authors Human Brain Mapping Published by Wiley Periodicals, Inc.

The clinical relevance of distortion correction in presurgical fMRI at 7T

Presurgical planning with fMRI benefits from increased reliability and the possibility to reduce measurement time introduced by using ultra-high field. Echo-planar imaging suffers, however, from geometric distortions which scale with field strength and potentially give rise to clinically significant displacement of functional activation. We evaluate the effectiveness of a dynamic distortion correction (DDC) method based on unmodified single-echo EPI in the context of simulated presurgical planning fMRI at 7T and compare it with static distortion correction (SDC). The extent of distortion in EPI and activation shifts are investigated in a group of eleven patients with a range of neuropathologies who performed a motor task. The consequences of neglecting to correct images for susceptibility-induced distortions are assessed in a clinical context. It was possible to generate time series of EPI-based field maps which were free of artifacts in the eloquent brain areas relevant to presurgical fMRI, despite the presence of signal dropouts caused by pathologies and post-operative sites. Distortions of up to 5.1mm were observed in the primary motor cortex in raw EPI. These were accurately corrected with DDC and slightly less accurately with SDC. The dynamic nature of distortions in UHF clinical fMRI was demonstrated via investigation of temporal variation in voxel shift maps, confirming the potential inadequacy of SDC based on a single reference field map, particularly in the vicinity of pathologies or in the presence of motion. In two patients, the distortion correction was potentially clinically significant in that it might have affected the localization or interpretation of activation and could thereby have influenced the treatment plan. Distortion correction is shown to be effective and clinically relevant in presurgical planning at 7T.

A Hitchhiker's Guide to Functional Magnetic Resonance Imaging

Functional Magnetic Resonance Imaging (fMRI) studies have become increasingly popular both with clinicians and researchers as they are capable of providing unique insights into brain functions. However, multiple technical considerations (ranging from specifics of paradigm design to imaging artifacts, complex protocol definition, and multitude of processing and methods of analysis, as well as intrinsic methodological limitations) must be considered and addressed in order to optimize fMRI analysis and to arrive at the most accurate and grounded interpretation of the data. In practice, the researcher/clinician must choose, from many available options, the most suitable software tool for each stage of the fMRI analysis pipeline. Herein we provide a straightforward guide designed to address, for each of the major stages, the techniques, and tools involved in the process. We have developed this guide both to help those new to the technique to overcome the most critical difficulties in its use, as well as to serve as a resource for the neuroimaging community.

A method for the dynamic correction of B0-related distortions in single-echo EPI at 7T

We propose a method to calculate field maps from the phase of each EPI in an fMRI time series. These field maps can be used to correct the corresponding magnitude images for distortion caused by inhomogeneity in the static magnetic field. In contrast to conventional static distortion correction, in which one ‘snapshot’ field map is applied to all subsequent fMRI time points, our method also captures dynamic changes to B₀ which arise due to motion and respiration. The approach is based on the assumption that the non-B₀-related contribution to the phase measured by each radio-frequency coil, which is dominated by the coil sensitivity, is stable over time and can therefore be removed to yield a field map from EPI.

Our solution addresses imaging with multi-channel coils at ultra-high field (7 T), where phase offsets vary rapidly in space, phase processing is non-trivial and distortions are comparatively large. We propose using dual-echo gradient echo reference scan for the phase offset calculation, which yields estimates with high signal-to-noise ratio. An extrapolation method is proposed which yields reliable estimates for phase offsets even where motion is large and a tailored phase unwrapping procedure for EPI is suggested which gives robust results in regions with disconnected tissue or strong signal decay.

Phase offsets are shown to be stable during long measurements (40 min) and for large head motions. The dynamic distortion correction proposed here is found to work accurately in the presence of large motion (up to 8.1°), whereas a conventional method based on single field map fails to correct or even introduces distortions (up to 11.2 mm). Finally, we show that dynamic unwarping increases the temporal stability of EPI in the presence of motion.

Our approach can be applied to any EPI measurements without the need for sequence modification.

Improving the clinical potential of ultra-high field fMRI using a model-free analysis method based on response consistency

Objective

To develop an analysis method that is sensitive to non-model-conform responses often encountered in ultra-high field presurgical planning fMRI. Using the consistency of time courses over a number of experiment repetitions, it should exclude low quality runs and generate activation maps that reflect the reliability of responses.

Materials and methods

7 T fMRI data were acquired from six healthy volunteers: three performing purely motor tasks and three a visuomotor task. These were analysed with the proposed approach (UNBIASED) and the GLM.

Results

UNBIASED results were generally less affected by false positive results than the GLM. Runs that were identified as being of low quality were confirmed to contain little or no activation. In two cases, regions were identified as activated in UNBIASED but not GLM results. Signal changes in these areas were time-locked to the task, but were delayed or transient.

Conclusion

UNBIASED is shown to be a reliable means of identifying consistent task-related signal changes regardless of response timing. In presurgical planning, UNBIASED could be used to rapidly generate reliable maps of the consistency with which eloquent brain regions are activated without recourse to task timing and despite modified hemodynamics.

Electronic supplementary material

The online version of this article (doi:10.1007/s10334-016-0533-8) contains supplementary material, which is available to authorized users.

Attention shifts the language network reflecting paradigm presentation

OBJECTIVES

Functional magnetic resonance imaging (fMRI) is a reliable and non-invasive method with which to localize language function in pre-surgical planning. In clinical practice, visual stimulus presentation is often difficult or impossible, due to the patient's restricted language or attention abilities. Therefore, our aim was to investigate modality-specific differences in visual and auditory stimulus presentation.

METHODS

Ten healthy subjects participated in an fMRI study comprising two experiments with visual and auditory stimulus presentation. In both experiments, two language paradigms (one for language comprehension and one for language production) used in clinical practice were investigated. In addition to standard data analysis by the means of the general linear model (GLM), independent component analysis (ICA) was performed to achieve more detailed information on language processing networks.

RESULTS

GLM analysis revealed modality-specific brain activation for both language paradigms for the contrast visual > auditory in the area of the intraparietal sulcus and the hippocampus, two areas related to attention and working memory. Using group ICA, a language network was detected for both paradigms independent of stimulus presentation modality. The investigation of language lateralization revealed no significant variations. Visually presented stimuli further activated an attention-shift network, which could not be identified for the auditory presented language.

CONCLUSION

The results of this study indicate that the visually presented language stimuli additionally activate an attention-shift network. These findings will provide important information for pre-surgical planning in order to preserve reading abilities after brain surgery, significantly improving surgical outcomes. Our findings suggest that the presentation modality for language paradigms should be adapted on behalf of individual indication.