Deep Learning Provides Exceptional Accuracy to ECoG-Based Functional Language Mapping for Epilepsy Surgery.

The success of surgical resection in epilepsy patients depends on preserving functionally critical brain regions, while removing pathological tissues. Being the gold standard, electro-cortical stimulation mapping (ESM) helps surgeons in localizing the function of eloquent cortex through electrical stimulation of electrodes placed directly on the cortical brain surface. Due to the potential hazards of ESM, including increased risk of provoked seizures, electrocorticography based functional mapping (ECoG-FM) was introduced as a safer alternative approach. However, ECoG-FM has a low success rate when compared to the ESM. In this study, we address this critical limitation by developing a new algorithm based on deep learning for ECoG-FM and thereby we achieve an accuracy comparable to ESM in identifying eloquent language cortex. In our experiments, with 11 epilepsy patients who underwent presurgical evaluation (through deep learning-based signal analysis on 637 electrodes), our proposed algorithm obtained an accuracy of 83.05% in identifying language regions, an exceptional 23% improvement with respect to the conventional ECoG-FM analysis (∼60%). Our findings have demonstrated, for the first time, that deep learning powered ECoG-FM can serve as a stand-alone modality and avoid likely hazards of the ESM in epilepsy surgery. Hence, reducing the potential for developing post-surgical morbidity in the language function.

How meaning unfolds in neural time: Embodied reactivations can precede multimodal semantic effects during language processing.

Research on how the brain construes meaning during language use has prompted two conflicting accounts. According to the 'grounded view', word understanding involves quick reactivations of sensorimotor (embodied) experiences evoked by the stimuli, with simultaneous or later engagement of multimodal (conceptual) systems integrating information from various sensory streams. Contrariwise, for the 'symbolic view', this capacity depends crucially on multimodal operations, with embodied systems playing epiphenomenal roles after comprehension. To test these contradictory hypotheses, the present magnetoencephalography study assessed implicit semantic access to grammatically constrained action and non-action verbs (n = 100 per category) while measuring spatiotemporally precise signals from the primary motor cortex (M1, a core region subserving bodily movements) and the anterior temporal lobe (ATL, a putative multimodal semantic hub). Convergent evidence from sensor- and source-level analyses revealed that increased modulations for action verbs occurred earlier in M1 (∼130-190 ms) than in specific ATL hubs (∼250-410 ms). Moreover, machine-learning decoding showed that trial-by-trial classification peaks emerged faster in M1 (∼100-175 ms) than in the ATL (∼345-500 ms), with over 71% accuracy in both cases. Considering their latencies, these results challenge the 'symbolic view' and its implication that sensorimotor mechanisms play only secondary roles in semantic processing. Instead, our findings support the 'grounded view', showing that early semantic effects are critically driven by embodied reactivations and that these cannot be reduced to post-comprehension epiphenomena, even when words are individually classified. Briefly, our study offers non-trivial insights to constrain fine-grained models of language and understand how meaning unfolds in neural time.

Identification of Focal Epileptogenic Networks in Generalized Epilepsy Using Brain Functional Connectivity Analysis of Bilateral Intracranial EEG Signals.

Simultaneous bilateral onset and bi-synchrony epileptiform discharges in electroencephalogram (EEG) remain hallmarks for generalized seizures. However, the possibility of an epileptogenic focus triggering rapidly generalized epileptiform discharges has been documented in several studies. Previously, a new multi-stage surgical procedure using bilateral intracranial EEG (iEEG) prior to and post complete corpus callosotomy (CC) was developed to uncover seizure focus in non-lateralizing focal epilepsy. Five patients with drug-resistant generalized epilepsy who underwent this procedure were included in the study. Their bilateral iEEG findings prior to complete CC showed generalized epileptiform discharges with no clear lateralization. Nonetheless, the bilateral ictal iEEG findings post complete CC indicated lateralized or localized seizure onset. This study hypothesized that brain functional connectivity analysis, applied to the pre CC bilateral iEEG recordings, could help identify focal epileptogenic networks in generalized epilepsy. The results indicated that despite diffuse epileptiform discharges, focal features can still be observed in apparent generalized seizures through brain connectivity analysis. The seizure onset localization/lateralization from connectivity analysis demonstrated a good agreement with the bilateral iEEG findings post complete CC and final surgical outcomes. Our study supports the role of focal epileptic networks in generalized seizures.

Is it time to replace the Wada test and put awake craniotomy to sleep?

The question we address here is whether the invasive presurgical brain mapping approaches of direct cortical stimulation and of the Wada procedure can be replaced by noninvasive functional neuroimaging methods (functional magnetic resonance imaging [fMRI], magnetoencephalography [MEG], transcranial magnetic stimulation and [TMS]). First, we outline the reasons for contemplating such a replacement. Second, we present evidence to the effect that the efficacy of the invasive and noninvasive methods, while suboptimal, is comparable. Third, we discuss additional advantages of noninvasive presurgical brain mapping and conclude that there are no longer compelling reasons for opting for invasive mapping in many if not most cases provided that the non-invasive methods are available.

Interhemispheric effective and functional cortical connectivity signatures of spina bifida are consistent with callosal anomaly.

The impact of the posterior callosal anomalies associated with spina bifida on interhemispheric cortical connectivity is studied using a method for estimating cortical multivariable autoregressive models from scalp magnetoencephalography data. Interhemispheric effective and functional connectivity, measured using conditional Granger causality and coherence, respectively, is determined for the anterior and posterior cortical regions in a population of five spina bifida and five control subjects during a resting eyes-closed state. The estimated connectivity is shown to be consistent over the randomly selected subsets of the data for each subject. The posterior interhemispheric effective and functional connectivity and cortical power are significantly lower in the spina bifida group, a result that is consistent with posterior callosal anomalies. The anterior interhemispheric effective and functional connectivity are elevated in the spina bifida group, a result that may reflect compensatory mechanisms. In contrast, the intrahemispheric effective connectivity is comparable in the two groups. The differences between the spina bifida and control groups are most significant in the θ and α bands.

Extracting biomarkers of autism from MEG resting-state functional connectivity networks.

The present study is a preliminary attempt to use graph theory for deriving distinct features of resting-state functional networks in young adults with autism spectrum disorder (ASD). Networks modeled neuromagnetic signal interactions between sensors using three alternative interdependence measures: (a) a non-linear measure of generalized synchronization (robust interdependence measure [RIM]), (b) mutual information (MI), and (c) partial directed coherence (PDC). To summarize the information contained in each network model we employed well-established global graph measures (average strength, assortativity, clustering, and efficiency) as well as graph measures (average strength of edges) tailored to specific hypotheses concerning the spatial distribution of abnormalities in connectivity among individuals with ASD. Graph measures then served as features in leave-one-out classification analyses contrasting control and ASD participants. We found that combinations of regionally constrained graph measures, derived from RIM, performed best, discriminating between the two groups with 93.75% accuracy. Network visualization revealed that ASD participants displayed significantly reduced interdependence strength, both within bilateral frontal and temporal sensors, as well as between temporal sensors and the remaining recording sites, in agreement with previous studies of functional connectivity in this disorder.

Time course of electromagnetic activity associated with detection of rare events.

The neural origins of the cortical response to rare sensory events remain poorly understood. Using simultaneous event-related potentials and magnetic resonance imaging, we investigated the anatomical profile of regional activity at various processing stages during performance of auditory and visual variants of an oddball paradigm. The earliest rarity-detection response was found in sensory-specific cortices, rapidly spreading to tertiary association areas, mesial temporal and frontal cortices by 150-200 ms. P3m-related activity was not found in sensory-specific cortices. On the basis of the anatomic distribution of P3m-related activity, this component is likely to reflect more generalized cognitive abilities hosted by association cortical regions.

Increased response variability in autistic brains?

One of the key ideas regarding atypical connectivity in autistic brains is the hypothesis of noisier networks. The systems level version of this hypothesis predicts reduced reliability or increased variability in the evoked responses of individuals with autism. Using magnetoencephalography, we examined the response of individuals with autism spectrum disorder versus matched typically developing persons to passive tactile stimulation of the thumb and index finger of the dominant (right) hand. A number of different analyses failed to show higher variability in the evoked response to the thumb or to the index finger in the autism group as compared with typicals. Our results argue against the hypothesis that the brain networks in autism are noisier than normal.

Transcallosal connectivity and cortical rhythms: findings in children with spina bifida.

We studied the relation between cortical oscillatory rhythms and the structural integrity of the corpus callosum in 21 children with spina bifida and hydrocephalus. Participants underwent resting state neuromagnetic recordings and diffusion tensor imaging. Areas of three segments of the corpus callosum (genu, body, splenium) were derived through diffusion tensor imaging-based morphometrics. Children with spina bifida showed reduced values of spectral power in the theta, alpha and beta bands when compared with age-matched controls, but only in the posterior and temporal regions. Reduced spectral power in posterior regions correlated with decreased area of the posterior segments of the corpus callosum. Atypical cortical oscillatory activity is associated with reduced transcallosal connectivity in children with spina bifida.

How somatic cortical maps differ in autistic and typical brains.

The comorbidity of 'core characteristics' and sensorimotor abnormalities in autism implies abnormalities in brain development of a general and pervasive nature and atypical organization of sensory cortex. By using magnetoencephalography, we examined the cortical response to passive tactile stimulation of the thumb and index finger of the dominant hand and lip of the individuals with autism spectrum disorder and typically developing persons. The distance between the cortical representations of thumb and the lip was significantly larger in the autism group than in typicals. Moreover, in cortex, the thumb is typically closer to the lip than the index finger. This was not observed in persons with autism. Our findings are arguably the first demonstration of abnormality in sensory organization in the brains of persons with autism.

Aberrant cortical functionality and somatosensory deficits after stroke.

Damage and/or disconnection of the primary somatosensory cortex (SI) after stroke leads to deficits in touch perception. We used magnetoencephalography to test whether specific patterns of functionality of the somatosensory cortex are associated with different degrees of postacute somatosensory deficit. Nineteen postacute unilateral stroke patients suffering different degrees of somatosensory deficit (six nonexistent, six moderate, and seven severe) and eight aged-matched controls underwent high-resolution MRI and whole-head magnetoencephalography recordings of somatosensory-evoked fields and of spontaneous slow oscillatory activity. Amplitude of SI activation after tactile stimulation in the affected and nonaffected hemispheres and delta dipole density (DDD) in the postcentral areas were estimated and compared across the four groups. Severe postacute somatosensory deficit was accompanied, in all cases, with absence of SI responses to stimulation in the affected hand and a significant asymmetry in postcentral DDD toward the affected hemisphere. Patients with moderate sensory loss showed asymmetry in their postcentral DDD (four cases toward the affected hemisphere and two toward the unaffected) but no atypical amplitudes in SI activation. Recordings in stroke patients without somatosensory deficit did not differ from those obtained in controls for SI amplitude or postcentral DDD. In stroke patients, amplitude of SI responses and postcentral DDD show a negative correlation. Lack of activation of SI cortex after stimulation of the affected hand and spontaneous slow oscillatory activity in postcentral areas are neurophysiological correlates of somatosensory deficit in the postacute phase of stroke.

When epilepsy interferes with word comprehension: findings in Landau-Kleffner syndrome.

Landau-Kleffner syndrome is characterized by a regression in receptive language. The factors that affect the clinical expression of this syndrome remain unclear. This study presents neuroimaging findings in 2 patients showing different clinical evolutions. Linguistic regression persisted in 1 patient and evolved positively in the other. In patient A (with severe linguistic regression) there was an overlap between areas engaged during word recognition and those involved in generating the epileptiform activity; in patient B (with better linguistic evolution), receptive language was predominantly represented in the right hemisphere (unaffected). Patient A underwent multiple subpial transections. The 2-year follow-up indicated linguistic improvement, absence of epileptiform activity, and activation of the left temporal cortex during word comprehension. These results suggest that the resolution of the linguistic deficit in Landau-Kleffner syndrome may be modulated by the language-specific cortex freed from interfering epileptiform activity or by reorganization of the receptive language cortex triggered by the epileptic activity.

A comparison of functional MRI and magnetoencephalography for receptive language mapping.

We compared functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) for the mapping of receptive language function. Participants performed the same language task in the two different imaging environments. MEG activation profiles showed prominent bilateral activity in superior temporal gyrus and left-lateralized activity in middle temporal gyrus. fMRI activation profiles revealed bilateral activity in prefrontal, superior temporal, middle temporal, and visual areas. Laterality quotients derived from the two modalities showed poor agreement between the two methods for commonly active regions of interest. Locations of peak activity also varied considerably within participants between the two methods.

Magnetoencephalographic studies of language reorganization after cerebral insult.

We review our experience with the application of magnetoencephalography (MEG) to the study of reorganization of the mechanisms supporting auditory language comprehension. In 3 studies, patient populations with cerebral insult of differing etiology, including epilepsy, surgical resection, and stroke, performed a running recognition task for spoken words while MEG data were collected using a whole-head magnetometer. Increased activation in the right hemisphere after left temporal lobectomy was associated with greater relative activation in that hemisphere preoperatively. Patients with chronic seizure disorder secondary to mesial temporal sclerosis exhibited a tendency toward an interhemispheric shift of language function, and those with epilepsy secondary to neoplasm showed a tendency toward an intrahemispheric shift. Patients with aphasia secondary to unilateral left-hemisphere stroke exhibited a more bilateral and diffuse overall profile of activation within the left hemisphere than control subjects of similar age. Taken together, results provide evidence that reorganization of cortex subserving auditory comprehension can occur well into the fifth and sixth decades and that the nature of the plastic response is dependent on variables such as premorbid language laterality, etiology, and, in specific groups, age at insult.

Time-locked brain activity associated with emotion: a pilot MEG study.

OBJECTIVE: To examine the time course of brain activation in response to emotionally evocative pictures. METHODS AND PROCEDURES: Regions of the brain involved in the processing of affective stimuli in response to picture sets rated unpleasant, pleasant and affectively neutral, as well as the order of activation of each region, were investigated using magnetoencephalography in 10 normal adult volunteers. RESULTS: Spatiotemporal maps were found consisting of two basic components. The first involving activation in the occipital and basal aspects of the temporal cortex- lasted, on average, 270 ms post-stimulus. The second component involving activation in the mesial temporal lobes (MTL) extended from 270 to 850 ms post-stimulus. After (serial) activating the mesial temporal lobe structures or simultaneous (parallel) to it, activation is also observed in the frontal structures. CONCLUSIONS: The temporal organization in the brain of an emotional stimulus requires the serial and alternating engagement of frontal and posterior cortices. It is suggested that lesions to the brain may disrupt this temporal course, altering the emotional response commonly observed in patients with brain injury.

Early development of neurophysiological processes involved in normal reading and reading disability: a magnetic source imaging study.

This longitudinal study examined the development of the brain mechanism involved in phonological decoding in beginning readers using magnetic source imaging. Kindergarten students were assigned to 2 groups: those who showed mastery of skills that are important predictors of proficient reading (low-risk group) and those who initially did not show mastery but later benefited from systematic reading instruction and developed average-range reading skills at the end of Grade 1 (high-risk responders). Spatiotemporal profiles of brain activity were obtained during performance of letter-sound and pseudoword naming tasks before and after Grade 1 instruction. With few exceptions, low-risk children showed early development of brain activation profiles that are typical of older skilled readers. Provided that temporoparietal and visual association areas were recruited into the brain mechanism that supported reading, the majority of high-risk responder children benefited from systematic reading instruction and developed adequate reading abilities.

Toward the substitution of invasive electroencephalography in epilepsy surgery.

The authors compared the localization accuracy of interictal magnetoencephalography (MEG) with ictal and interictal invasive video electroencephalography (VEEG) in identifying the epileptogenic zone in epilepsy surgery candidates. Forty-one patients, 29 with temporal lobe epilepsy (TLE) and 12 with extratemporal lobe epilepsy (ETLE), participated. Only patients with interictal changes during the MEG recordings were included. A comparison of the accuracy of invasive VEEG and MEG seizure zone identification was based on the degree of overlap between the location of the actual surgical resection and the zone identified by each method, and the success of surgery in reducing seizure activity. No statistical differences were observed between the accuracy of invasive VEEG and MEG in determining the location of the seizure zone across TLE and ETLE cases. Invasive VEEG and MEG localization judgments were correct in 54% and 56% of the cases, respectively. Separate group analyses suggested that MEG may be less beneficial relative to invasive VEEG in ETLE than TLE cases. MEG is of statistically equivalent accuracy to invasive VEEG, despite the fact that its use has not reached optimal conditions. The authors predict the replacement of the more invasive procedure with MEG in the near future for TLE cases, subsequent to the optimization of the conditions under which preoperative MEG is performed.

Reproducibility of measures of neurophysiological activity in Wernicke's area: a magnetic source imaging study.

OBJECTIVE: The purpose of this study was to evaluate the reproducibility of estimates of neurophysiological activity obtained with Magnetic Source Imaging. METHODS: Split-half data sets were obtained from 14 healthy volunteers during performance of a continuous recognition task for spoken words. The concurrent validity of spatiotemporal activation maps obtained with this task has been previously verified through comparisons with the Wada test and electrocortical stimulation mapping. Consecutive late activity sources (> 200 ms after stimulus onset) were modeled independently as equivalent current dipoles (ECDs) and used to identify the location of language-specific cortex in the left hemisphere (Wernicke's area). RESULTS: Linear displacement of the geometric center of the cluster of ECDs in this region ranged between 2 and 8 mm across subjects. Intraparticipant variability (range) in the onset latency of activity was +/-50 ms, while the range of change in global field power for the entire set of ECDs in Wernicke's area was less than 17% in all cases. CONCLUSIONS: The results indicate that despite its many conceptual limitations the ECD model can provide reliable estimates of regional cortical activity associated with the engagement of linguistic processes. SIGNIFICANCE: The results highlight the need for reproducibility studies when research questions pose particular requirements for precision of estimates of regional neurophysiological activity.

Cortical representation of dermatomes: MEG-derived maps after tactile stimulation.

Mechanical stimulation of skin receptors is known to evoke cortical responses arising from the somatosensory cortex. Here we present a magnetoencephalographic (MEG) study where dermatomal somatosensory-evoked fields (DSSEFs) were recorded after mechanical stimulation of sacral (S1), lumbar (L3), thoracic (Th7), and cervical (C4) dermatomes in three healthy volunteers. All MEG measurements were repeated in order to test the replicability of the results. DSSEFs were successfully measured and modeled in all three participants. The topography and temporal dynamics of cortical responses derived after stimulation of each dermatome are described. We found that cortical-evoked responses can be reliably recorded using MEG after mechanical stimulation of dermatomes when a sufficiently large skin region within the dermatome is stimulated. Primary sensory cortex response (SI) to each of the four dermatomes was replicable and showed stability over time. The MEG-derived individual maps of activation confirm the somatotopic representation of dermatomes in primary sensory cortex and the utility of MEG recordings in disentangling the interactions between primary and secondary sensory cortex during somatic perception.