Ballistocardiogram Artifact Reduction in Simultaneous EEG-fMRI Using Deep Learning.
Ontology highlight
ABSTRACT: Objective
The concurrent recording of electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) is a technique that has received much attention due to its potential for combined high temporal and spatial resolution. However, the ballistocardiogram (BCG), a large-amplitude artifact caused by cardiac induced movement contaminates the EEG during EEG-fMRI recordings. Removal of BCG in software has generally made use of linear decompositions of the corrupted EEG. This is not ideal as the BCG signal propagates in a manner which is non-linearly dependent on the electrocardiogram (ECG). In this paper, we present a novel method for BCG artifact suppression using recurrent neural networks (RNNs).Methods
EEG signals were recovered by training RNNs on the nonlinear mappings between ECG and the BCG corrupted EEG. We evaluated our model's performance against the commonly used Optimal Basis Set (OBS) method at the level of individual subjects, and investigated generalization across subjects.Results
We show that our algorithm can generate larger average power reduction of the BCG at critical frequencies, while simultaneously improving task relevant EEG based classification.Conclusion
The presented deep learning architecture can be used to reduce BCG related artifacts in EEG-fMRI recordings.Significance
We present a deep learning approach that can be used to suppress the BCG artifact in EEG-fMRI without the use of additional hardware. This method may have scope to be combined with current hardware methods, operate in real-time and be used for direct modeling of the BCG.
SUBMITTER: McIntosh JR
PROVIDER: S-EPMC7808341 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Simultaneous electroencephalography-functional MRI (EEG-fMRI) is a technique that combines temporal (largely from EEG) and spatial (largely from fMRI) indicators of brain dynamics. It is useful for understanding neuronal activity during many different event types, including spontaneous epileptic discharges, the activity of sleep stages, and activity evoked by external stimuli and decision-making tasks. However, EEG recorded during fMRI is subject to imaging, pulse, environment and motion artifact, causing noise many times greater than the neuronal signals of interest. Therefore, artifact removal methods are essential to ensure that artifacts are accurately removed, and EEG of interest is retained. This paper presents a systematic review of methods for artifact reduction in simultaneous EEG-fMRI from literature published since 1998, and an additional systematic review of EEG-fMRI studies published since 2016. The aim of the first review is to distill the literature into clear guidelines for use of simultaneous EEG-fMRI artifact reduction methods, and the aim of the second review is to determine the prevalence of artifact reduction method use in contemporary studies. We find that there are many published artifact reduction techniques available, including hardware, model based, and data-driven methods, but there are few studies published that adequately compare these methods. In contrast, recent EEG-fMRI studies show overwhelming use of just one or two artifact reduction methods based on literature published 15-20 years ago, with newer methods rarely gaining use outside the group that developed them. Surprisingly, almost 15% of EEG-fMRI studies published since 2016 fail to adequately describe the methods of artifact reduction utilized. We recommend minimum standards for reporting artifact reduction techniques in simultaneous EEG-fMRI studies and suggest that more needs to be done to make new artifact reduction techniques more accessible for the researchers and clinicians using simultaneous EEG-fMRI.
| S-EPMC7991907 | biostudies-literature
Project description:Objective.Simultaneous electroencephalography-functional magnetic resonance imaging (EEG-fMRI) recordings offer a high spatiotemporal resolution approach to study human brain and understand the underlying mechanisms mediating cognitive and behavioral processes. However, the high susceptibility of EEG to MRI-induced artifacts hinders a broad adaptation of this approach. More specifically, EEG data collected during fMRI acquisition are contaminated with MRI gradients and ballistocardiogram artifacts, in addition to artifacts of physiological origin. There have been several attempts for reducing these artifacts with manual and time-consuming pre-processing, which may result in biasing EEG data due to variations in selecting steps order, parameters, and classification of artifactual independent components. Thus, there is a strong urge to develop a fully automatic and comprehensive pipeline for reducing all major EEG artifacts. In this work, we introduced an open-access toolbox with a fully automatic pipeline for reducing artifacts from EEG data collected simultaneously with fMRI (refer to APPEAR).Approach.The pipeline integrates average template subtraction and independent component analysis to suppress both MRI-related and physiological artifacts. To validate our results, we tested APPEAR on EEG data recorded from healthy control subjects during resting-state (n= 48) and task-based (i.e. event-related-potentials (ERPs);n= 8) paradigms. The chosen gold standard is an expert manual review of the EEG database.Main results.We compared manually and automated corrected EEG data during resting-state using frequency analysis and continuous wavelet transformation and found no significant differences between the two corrections. A comparison between ERP data recorded during a so-called stop-signal task (e.g. amplitude measures and signal-to-noise ratio) also showed no differences between the manually and fully automatic fMRI-EEG-corrected data.Significance.APPEAR offers the first comprehensive open-source toolbox that can speed up advancement of EEG analysis and enhance replication by avoiding experimenters' preferences while allowing for processing large EEG-fMRI cohorts composed of hundreds of subjects with manageable researcher time and effort.
| S-EPMC10696919 | biostudies-literature
Project description:In simultaneous EEG-fMRI, the EEG recordings are severely contaminated by ballistocardiogram (BCG) artifacts, which are caused by cardiac pulsations. To reconstruct and remove the BCG artifacts, one promising method is to measure the artifacts in the absence of EEG signal by placing a group of electrodes (BCG electrodes) on a conductive layer (reference layer) insulated from the scalp. However, current BCG reference layer (BRL) methods either use a customized EEG cap composed of electrode pairs, or need to construct the custom reference layer through additional model-building experiments for each EEG-fMRI experiment. These requirements have limited the versatility and efficiency of BRL. The aim of this study is to propose a more practical and efficient BRL method and compare its performance with the most popular BCG removal method, the optimal basis sets (OBS) algorithm.By designing the reference layer as a permanent and reusable cap, the new BRL method is able to be used with a standard EEG cap, and no extra experiments and preparations are needed to use the BRL in an EEG-fMRI experiment.The BRL method effectively removed the BCG artifacts from both oscillatory and evoked potential scalp recordings and recovered the EEG signal.Compared to the OBS, this new BRL method improved the contrast-to-noise ratios of the alpha-wave, visual, and auditory evoked potential signals by 101%, 76%, and 75%, respectively, employing 160 BCG electrodes. Using only 20 BCG electrodes, the BRL improved the EEG signal by 74%/26%/41%, respectively.The proposed method can substantially improve the EEG signal quality compared with traditional methods.
| S-EPMC4126606 | biostudies-literature
Project description:EEG data recorded during simultaneous fMRI are contaminated by large voltages generated by time-varying magnetic field gradients. Correction of the resulting gradient artifact (GA) generally involves low-pass filtering to attenuate the high-frequency voltage fluctuations of the GA, followed by subtraction of a GA template produced by averaging over repeats of the artifact waveforms. This average artifact subtraction (AAS) process relies on the EEG amplifier having a large enough dynamic range to characterize the artifact voltages and on invariance of the artifact waveform over repeated image acquisitions. Saturation of the amplifiers and changes in subject position can leave unwanted residual GA after AAS. Previous modeling work suggested that modifying the lead layout and the exit position of the cable bundle on the EEG cap could reduce the GA amplitude. Here, we used simulations and experiments to evaluate the effect of modifying the lead paths on the magnitude of the GA and on the residual artifact after AAS. The modeling work showed that for wire paths following great circles, the smallest overall GA occurs when the leads converge at electrode Cz. The performance of this new cap design was compared with a standard cap in experiments on a spherical agar phantom and human subjects. Using gradient pulses applied separately along the three Cartesian axes, we found that the GA due to the foot-head gradient was most significantly reduced relative to a standard cap for the phantom, whereas the anterior-posterior GA was most attenuated for human subjects. In addition, there was an overall 37% reduction in the RMS GA amplitude produced by a standard EPI sequence when comparing the two caps on the phantom. In contrast, the subjects showed an 11% increase in the average RMS of the GA. This work shows that the optimal design reduces the GA on a spherical phantom however; these gains are not translated to human subjects, probably due to the differences in geometry.
| S-EPMC4114285 | biostudies-literature
Project description:Sensorimotor coordination requires orchestrated network activity in the brain, mediated by inter- and intra-hemispheric interactions that may be affected by aging-related changes. We adopted a theoretical model, according to which intra-hemispheric inhibition from premotor to primary motor cortex is mandatory to compensate for inter-hemispheric excitation through the corpus callosum. To test this as a function of age we acquired electroencephalography (EEG) simultaneously with functional magnetic resonance imaging (fMRI) in two groups of healthy adults (younger N = 13: 20-25 year and older N = 14: 59-70 year) while learning a unimanual motor task. On average, quality of performance of older participants stayed significantly below that of the younger ones. Accompanying decreases in motor-event-related EEG β-activity were lateralized toward contralateral motor regions, albeit more so in younger participants. In this younger group, the mean β-power during motor task execution was significantly higher in bilateral premotor areas compared to the older adults. In both groups, fMRI blood oxygen level dependent (BOLD) signals were positively correlated with source-reconstructed β-amplitudes: positive in primary motor and negative in premotor cortex. This suggests that β-amplitude modulation is associated with primary motor cortex "activation" (positive BOLD response) and premotor "deactivation" (negative BOLD response). Although the latter results did not discriminate between age groups, they underscore that enhanced modulation in primary motor cortex may be explained by a β-associated excitatory crosstalk between hemispheres.
| S-EPMC8996364 | biostudies-literature
Project description:Combining electroencephalogram (EEG) recording and functional magnetic resonance imaging (fMRI) offers the potential for imaging brain activity with high spatial and temporal resolution. This potential remains limited by the significant ballistocardiogram (BCG) artifacts induced in the EEG by cardiac pulsation-related head movement within the magnetic field. We model the BCG artifact using a harmonic basis, pose the artifact removal problem as a local harmonic regression analysis, and develop an efficient maximum likelihood algorithm to estimate and remove BCG artifacts. Our analysis paradigm accounts for time-frequency overlap between the BCG artifacts and neurophysiologic EEG signals, and tracks the spatiotemporal variations in both the artifact and the signal. We evaluate performance on: simulated oscillatory and evoked responses constructed with realistic artifacts; actual anesthesia-induced oscillatory recordings; and actual visual evoked potential recordings. In each case, the local harmonic regression analysis effectively removes the BCG artifacts, and recovers the neurophysiologic EEG signals. We further show that our algorithm outperforms commonly used reference-based and component analysis techniques, particularly in low SNR conditions, the presence of significant time-frequency overlap between the artifact and the signal, and/or large spatiotemporal variations in the BCG. Because our algorithm does not require reference signals and has low computational complexity, it offers a practical tool for removing BCG artifacts from EEG data recorded in combination with fMRI.
| S-EPMC4761527 | biostudies-literature
Project description:Background: Potentially curative epilepsy surgery can be offered if a single, discrete epileptogenic zone (EZ) can be identified. For individuals in whom there is no clear concordance between clinical localization, scalp EEG, and imaging data, intracranial EEG (icEEG) may be needed to confirm a predefined hypothesis regarding irritative zone (IZ), seizure onset zone (SOZ), and EZ prior to surgery. However, icEEG has limited spatial sampling and may fail to reveal the full extent of epileptogenic network if predefined hypothesis is not correct. Simultaneous icEEG-fMRI has been safely acquired in humans and allows exploration of neuronal activity at the whole-brain level related to interictal epileptiform discharges (IED) captured intracranially. Methods: We report icEEG-fMRI in eight patients with refractory focal epilepsy who had resective surgery and good postsurgical outcome. Surgical resection volume in seizure-free patients post-surgically reflects confirmed identification of the EZ. IEDs on icEEG were classified according to their topographic distribution and localization (Focal, Regional, Widespread, and Non-contiguous). We also divided IEDs by their location within the surgical resection volume [primary IZ (IZ1) IED] or outside [secondary IZ (IZ2) IED]. The distribution of fMRI blood oxygen level-dependent (BOLD) changes associated with individual IED classes were assessed over the whole brain using a general linear model. The concordance of resulting BOLD map was evaluated by comparing localization of BOLD clusters with surgical resection volume. Additionally, we compared the concordance of BOLD maps and presence of BOLD clusters in remote brain areas: precuneus, cuneus, cingulate, medial frontal, and thalamus for different IED classes. Results: A total of 38 different topographic IED classes were identified across the 8 patients: Focal (22) and non-focal (16, Regional = 9, Widespread = 2, Non-contiguous = 5). Twenty-nine IEDs originated from IZ1 and 9 from IZ2. All IED classes were associated with BOLD changes. BOLD maps were concordant with the surgical resection volume for 27/38 (71%) IED classes, showing statistical global maximum BOLD cluster or another cluster in the surgical resection volume. The concordance of BOLD maps with surgical resection volume was greater (p < 0.05) for non-focal (87.5%, 14/16) as compared to Focal (59%, 13/22) IED classes. Additionally, BOLD clusters in remote cortical and deep brain areas were present in 84% (32/38) of BOLD maps, more commonly (15/16; 93%) for non-focal IED-related BOLD maps. Conclusions: Simultaneous icEEG-fMRI can reveal BOLD changes at the whole-brain level for a wide range of IEDs on icEEG. BOLD clusters within surgical resection volume and remote brain areas were more commonly seen for non-focal IED classes, suggesting that a wider hemodynamic network is at play.
| S-EPMC8490636 | biostudies-literature
Project description:EEG data acquired in an MRI scanner are heavily contaminated by gradient artifacts that can significantly compromise signal quality. We developed two new methods based on independent component analysis (ICA) for reducing gradient artifacts from spiral in-out and echo-planar pulse sequences at 3 T, and compared our algorithms with four other commonly used methods: average artifact subtraction (Allen, P., Josephs, O., Turner, R., 2000. A method for removing imaging artifact from continuous EEG recorded during functional MRI. NeuroImage 12, 230-239.), principal component analysis (Niazy, R., Beckmann, C., Iannetti, G., Brady, J., Smith, S., 2005. Removal of FMRI environment artifacts from EEG data using optimal basis sets. NeuroImage 28, 720-737.), Taylor series ( Wan, X., Iwata, K., Riera, J., Kitamura, M., Kawashima, R., 2006. Artifact reduction for simultaneous EEG/fMRI recording: adaptive FIR reduction of imaging artifacts. Clin. Neurophysiol. 117, 681-692.) and a conventional temporal ICA algorithm. Models of gradient artifacts were derived from simulations as well as a water phantom and performance of each method was evaluated on datasets constructed using visual event-related potentials (ERPs) as well as resting EEG. Our new methods recovered ERPs and resting EEG below the beta band (<12.5 Hz) with high signal-to-noise ratio (SNR>4). Our algorithms outperformed all of these methods on resting EEG in the theta and alpha bands (SNR>4); however, for all methods, signal recovery was modest (SNR approximately 1) in the beta band and poor (SNR<0.3) in the gamma band and above. We found that the conventional ICA algorithm performed poorly with uniformly low SNR (<0.1). Taken together, our new ICA-based methods offer a more robust technique for gradient artifact reduction when scanning at 3 T using spiral in-out and echo-planar pulse sequences. We provide new insights into the strengths and weaknesses of each method using a unified subspace framework.
| S-EPMC2745974 | biostudies-other
Project description:In recent years, deep learning models have demonstrated an inherently better ability to tackle non-linear classification tasks, due to advances in deep learning architectures. However, much remains to be achieved, especially in designing deep convolutional neural network (CNN) configurations. The number of hyper-parameters that need to be optimized to achieve accuracy in classification problems increases with every layer used, and the selection of kernels in each CNN layer has an impact on the overall CNN performance in the training stage, as well as in the classification process. When a popular classifier fails to perform acceptably in practical applications, it may be due to deficiencies in the algorithm and data processing. Thus, understanding the feature extraction process provides insights to help optimize pre-trained architectures, better generalize the models, and obtain the context of each layer's features. In this work, we aim to improve feature extraction through the use of a texture amortization map (TAM). An algorithm was developed to obtain characteristics from the filters amortizing the filter's effect depending on the texture of the neighboring pixels. From the initial algorithm, a novel geometric classification score (GCS) was developed, in order to obtain a measure that indicates the effect of one class on another in a classification problem, in terms of the complexity of the learnability in every layer of the deep learning architecture. For this, we assume that all the data transformations in the inner layers still belong to a Euclidean space. In this scenario, we can evaluate which layers provide the best transformations in a CNN, allowing us to reduce the weights of the deep learning architecture using the geometric hypothesis.
| S-EPMC8870362 | biostudies-literature
Project description:Ballistocardiogram (BCG) artifact remains a major challenge that renders electroencephalographic (EEG) signals hard to interpret in simultaneous EEG and functional MRI (fMRI) data acquisition. Here, we propose an integrated learning and inference approach that takes advantage of a commercial high-density EEG cap, to estimate the BCG contribution in noisy EEG recordings from inside the MR scanner. To estimate reliably the full-scalp BCG artifacts, a near-optimal subset (20 out of 256) of channels first was identified using a modified recording setup. In subsequent recordings inside the MR scanner, BCG-only signal from this subset of channels was used to generate continuous estimates of the full-scalp BCG artifacts via inference, from which the intended EEG signal was recovered. The reconstruction of the EEG was performed with both a direct subtraction and an optimization scheme. We evaluated the performance on both synthetic and real contaminated recordings, and compared it to the benchmark Optimal Basis Set (OBS) method. In the challenging non-event-related-potential (non-ERP) EEG studies, our reconstruction can yield more than fourteen-fold improvement in reducing the normalized RMS error of EEG signals, compared to OBS.
| S-EPMC4114198 | biostudies-literature