Channel-independent recreation of artefactual signals in chronically recorded local field potentials using machine learning

Acquisition of neuronal signals involves a wide range of devices with specific electrical properties. Combined with other physiological sources within the body, the signals sensed by the devices are often distorted. Sometimes these distortions are visually identifiable, other times, they overlay with the signal characteristics making them very difficult to detect. To remove these distortions, the recordings are visually inspected and manually processed. However, this manual annotation process is time-consuming and automatic computational methods are needed to identify and remove these artefacts. Most of the existing artefact removal approaches rely on additional information from other recorded channels and fail when global artefacts are present or the affected channels constitute the majority of the recording system. Addressing this issue, this paper reports a novel channel-independent machine learning model to accurately identify and replace the artefactual segments present in the signals. Discarding these artifactual segments by the existing approaches causes discontinuities in the reproduced signals which may introduce errors in subsequent analyses. To avoid this, the proposed method predicts multiple values of the artefactual region using long–short term memory network to recreate the temporal and spectral properties of the recorded signal. The method has been tested on two open-access data sets and incorporated into the open-access SANTIA (SigMate Advanced: a Novel Tool for Identification of Artefacts in Neuronal Signals) toolbox for community use.

Methods for inferring neural circuit interactions and neuromodulation from local field potential and electroencephalogram measures

Electrical recordings of neural mass activity, such as local field potentials (LFPs) and electroencephalograms (EEGs), have been instrumental in studying brain function. However, these aggregate signals lack cellular resolution and thus are not easy to be interpreted directly in terms of parameters of neural microcircuits. Developing tools for a reliable estimation of key neural parameters from these signals, such as the interaction between excitation and inhibition or the level of neuromodulation, is important for both neuroscientific and clinical applications. Over the years, we have developed tools based on neural network modeling and computational analysis of empirical data to estimate neural parameters from aggregate neural signals. This review article gives an overview of the main computational tools that we have developed and employed to invert LFPs and EEGs in terms of circuit-level neural phenomena, and outlines future challenges and directions for future research.

Nonlinear reconfiguration of network edges, topology and information content during an artificial learning task

Here, we combine network neuroscience and machine learning to reveal connections between the brain’s network structure and the emerging network structure of an artificial neural network. Specifically, we train a shallow, feedforward neural network to classify hand-written digits and then used a combination of systems neuroscience and information-theoretic tools to perform ‘virtual brain analytics’ on the resultant edge weights and activity patterns of each node. We identify three distinct phases of network reconfiguration across learning, each of which are characterized by unique topological and information-theoretic signatures. Each phase involves aligning the connections of the neural network with patterns of information contained in the input dataset or preceding layers (as relevant). We also observe a process of low-dimensional category separation in the network as a function of learning. Our results offer a systems-level perspective of how artificial neural networks function—in terms of multi-stage reorganization of edge weights and activity patterns to effectively exploit the information content of input data during edge-weight training—while simultaneously enriching our understanding of the methods used by systems neuroscience.

Integrated space–frequency–time domain feature extraction for MEG-based Alzheimer’s disease classification

Magnetoencephalography (MEG) has been combined with machine learning techniques, to recognize the Alzheimer’s disease (AD), one of the most common forms of dementia. However, most of the previous studies are limited to binary classification and do not fully utilize the two available MEG modalities (extracted using magnetometer and gradiometer sensors). AD consists of several stages of progression, this study addresses this limitation by using both magnetometer and gradiometer data to discriminate between participants with AD, AD-related mild cognitive impairment (MCI), and healthy control (HC) participants in the form of a three-class classification problem. A series of wavelet-based biomarkers are developed and evaluated, which concurrently leverage the spatial, frequency and time domain characteristics of the signal. A bimodal recognition system based on an improved score-level fusion approach is proposed to reinforce interpretation of the brain activity captured by magnetometers and gradiometers. In this preliminary study, it was found that the markers derived from gradiometer tend to outperform the magnetometer-based markers. Interestingly, out of the total 10 regions of interest, left-frontal lobe demonstrates about 8% higher mean recognition rate than the second-best performing region (left temporal lobe) for AD/MCI/HC classification. Among the four types of markers proposed in this work, the spatial marker developed using wavelet coefficients provided the best recognition performance for the three-way classification. Overall, the proposed approach provides promising results for the potential of AD/MCI/HC three-way classification utilizing the bimodal MEG data.

3D convolutional neural networks-based multiclass classification of Alzheimer’s and Parkinson’s diseases using PET and SPECT neuroimaging modalities

Background Alzheimer’s disease (AD) is a neurodegenerative brain pathology formed due to piling up of amyloid proteins, development of plaques and disappearance of neurons. Another common subtype of dementia like AD, Parkinson’s disease (PD) is determined by the disappearance of dopaminergic neurons in the region known as substantia nigra pars compacta located in the midbrain. Both AD and PD target aged population worldwide forming a major chunk of healthcare costs. Hence, there is a need for methods that help in the early diagnosis of these diseases. PD subjects especially those who have confirmed postmortem plaque are a strong candidate for a second AD diagnosis. Modalities such as positron emission tomography (PET) and single photon emission computed tomography (SPECT) can be combined with deep learning methods to diagnose these two diseases for the benefit of clinicians. Result In this work, we deployed a 3D Convolutional Neural Network (CNN) to extract features for multiclass classification of both AD and PD in the frequency and spatial domains using PET and SPECT neuroimaging modalities to differentiate between AD, PD and Normal Control (NC) classes. Discrete Cosine Transform has been deployed as a frequency domain learning method along with random weak Gaussian blurring and random zooming in/out augmentation methods in both frequency and spatial domains. To select the hyperparameters of the 3D-CNN model, we deployed both 5- and 10-fold cross-validation (CV) approaches. The best performing model was found to be AD/NC(SPECT)/PD classification with random weak Gaussian blurred augmentation in the spatial domain using fivefold CV approach while the worst performing model happens to be AD/NC(PET)/PD classification without augmentation in the frequency domain using tenfold CV approach. We also found that spatial domain methods tend to perform better than their frequency domain counterparts. Conclusion The proposed model provides a good performance in discriminating AD and PD subjects due to minimal correlation between these two dementia types on the clinicopathological continuum between AD and PD subjects from a neuroimaging perspective.

Recurrence eigenvalues of movements from brain signals

The ability to characterize muscle activities or skilled movements controlled by signals from neurons in the motor cortex of the brain has many useful implications, ranging from biomedical perspectives to brain–computer interfaces. This paper presents the method of recurrence eigenvalues for differentiating moving patterns in non-mammalian and human models. The non-mammalian models of Caenorhabditis elegans have been studied for gaining insights into behavioral genetics and discovery of human disease genes. Systematic probing of the movement of these worms is known to be useful for these purposes. Study of dynamics of normal and mutant worms is important in behavioral genetic and neuroscience. However, methods for quantifying complexity of worm movement using time series are still not well explored. Neurodegenerative diseases adversely affect gait and mobility. There is a need to accurately quantify gait dynamics of these diseases and differentiate them from the healthy control to better understand their pathophysiology that may lead to more effective therapeutic interventions. This paper attempts to explore the potential application of the method for determining the largest eigenvalues of convolutional fuzzy recurrence plots of time series for measuring the complexity of moving patterns of Caenorhabditis elegans and neurodegenerative disease subjects. Results obtained from analyses demonstrate that the largest recurrence eigenvalues can differentiate phenotypes of behavioral dynamics between wild type and mutant strains of Caenorhabditis elegans; and walking patterns among healthy control subjects and patients with Parkinson’s disease, Huntington’s disease, or amyotrophic lateral sclerosis.

Person authentication based on eye-closed and visual stimulation using EEG signals

The study of Electroencephalogram (EEG)-based biometric has gained the attention of researchers due to the neurons’ unique electrical activity representation of an individual. However, the practical application of EEG-based biometrics is not currently widespread and there are some challenges to its implementation. Nowadays, the evaluation of a biometric system is user driven. Usability is one of the concerning issues that determine the success of the system. The basic elements of the usability of a biometric system are effectiveness, efficiency and user satisfaction. Apart from the mandatory consideration of the biometric system’s performance, users also need an easy-to-use and easy-to-learn authentication system. Thus, to satisfy these user requirements, this paper proposes a reasonable acquisition period and employs a consumer-grade EEG device to authenticate an individual to identify the performances of two acquisition protocols: eyes-closed (EC) and visual stimulation. A self-collected database of eight subjects was utilized in the analysis. The recording process was divided into two sessions, which were the morning and afternoon sessions. In each session, the subject was requested to perform two different tasks: EC and visual stimulation. The pairwise correlation of the preprocessed EEG signals of each electrode channel was determined and a feature vector was formed. Support vector machine (SVM) was then used for classification purposes. In the performance analysis, promising results were obtained, where EC protocol achieved an accuracy performance of 83.70–96.42% while visual stimulation protocol attained an accuracy performance of 87.64–99.06%. These results have demonstrated the feasibility and reliability of our acquisition protocols with consumer-grade EEG devices.

EEG-based human emotion recognition using entropy as a feature extraction measure

Many studies on brain–computer interface (BCI) have sought to understand the emotional state of the user to provide a reliable link between humans and machines. Advanced neuroimaging methods like electroencephalography (EEG) have enabled us to replicate and understand a wide range of human emotions more precisely. This physiological signal, i.e., EEG-based method is in stark comparison to traditional non-physiological signal-based methods and has been shown to perform better. EEG closely measures the electrical activities of the brain (a nonlinear system) and hence entropy proves to be an efficient feature in extracting meaningful information from raw brain waves. This review aims to give a brief summary of various entropy-based methods used for emotion classification hence providing insights into EEG-based emotion recognition. This study also reviews the current and future trends and discusses how emotion identification using entropy as a measure to extract features, can accomplish enhanced identification when using EEG signal.

Modeling brain connectivity dynamics in functional magnetic resonance imaging via particle filtering

Interest in the studying of functional connections in the brain has grown considerably in the last decades, as many studies have pointed out that alterations in the interaction among brain areas can play a role as markers of neurological diseases. Most studies in this field treat the brain network as a system of connections stationary in time, but dynamic features of brain connectivity can provide useful information, both on physiology and pathological conditions of the brain. In this paper, we propose the application of a computational methodology, named Particle Filter (PF), to study non-stationarities in brain connectivity in functional Magnetic Resonance Imaging (fMRI). The PF algorithm estimates time-varying hidden parameters of a first-order linear time-varying Vector Autoregressive model (VAR) through a Sequential Monte Carlo strategy. On simulated time series, the PF approach effectively detected and enabled to follow time-varying hidden parameters and it captured causal relationships among signals. The method was also applied to real fMRI data, acquired in presence of periodic tactile or visual stimulations, in different sessions. On these data, the PF estimates were consistent with current knowledge on brain functioning. Most importantly, the approach enabled to detect statistically significant modulations in the cause-effect relationship between brain areas, which correlated with the underlying visual stimulation pattern presented during the acquisition.

Convolutional neural network optimizes the application of diffusion kurtosis imaging in Parkinson’s disease

Objectives The literature regarding the use of diffusion-tensor imaging-derived metrics in the evaluation of Parkinson’s disease (PD) is controversial. This study attempted to assess the feasibility of a deep-learning-based method for detecting alterations in diffusion kurtosis measurements associated with PD. Methods A total of 68 patients with PD and 77 healthy controls were scanned using scanner-A (3 T Skyra) (DATASET-1). Meanwhile, an additional five healthy volunteers were scanned with both scanner-A and an additional scanner-B (3 T Prisma) (DATASET-2). Diffusion kurtosis imaging (DKI) of DATASET-2 had an extra b shell compared to DATASET-1. In addition, a 3D-convolutional neural network (CNN) was trained from DATASET-2 to harmonize the quality of scalar measures of scanner-A to a similar level as scanner-B. Whole-brain unpaired t test and Tract-Based Spatial Statistics (TBSS) were performed to validate the differences between the PD and control groups using the model-fitting method and CNN-based method, respectively. We further clarified the correlation between clinical assessments and DKI results. Results An increase in mean diffusivity (MD) was found in the left substantia nigra (SN) in the PD group. In the right SN, fractional anisotropy (FA) and mean kurtosis (MK) values were negatively correlated with Hoehn and Yahr (H&Y) scales. In the putamen (Put), FA values were positively correlated with the H&Y scales. It is worth noting that these findings were only observed with the deep learning method. There was neither a group difference nor a correlation with clinical assessments in the SN or striatum exceeding the significance level using the conventional model-fitting method. Conclusions The CNN-based method improves the robustness of DKI and can help to explore PD-associated imaging features.