Semi-supervised Seizure Prediction with Generative Adversarial Networks

Combining EEG Features and Convolutional Autoencoder for Neonatal Seizure Detection

Neonatal epilepsy is a common emergency phenomenon in neonatal intensive care units (NICUs), which requires timely attention, early identification, and treatment. Traditional detection methods mostly use supervised learning with enormous labeled data. Hence, this study offers a semi-supervised hybrid architecture for detecting seizures, which combines the extracted electroencephalogram (EEG) feature dataset and convolutional autoencoder, called Fd-CAE. First, various features in the time domain and entropy domain are extracted to characterize the EEG signal, which helps distinguish epileptic seizures subsequently. Then, the unlabeled EEG features are fed into the convolutional autoencoder (CAE) for training, which effectively represents EEG features by optimizing the loss between the input and output features. This unsupervised feature learning process can better combine and optimize EEG features from unlabeled data. After that, the pre-trained encoder part of the model is used for further feature learning of labeled data to obtain its low-dimensional feature representation and achieve classification. This model is performed on the neonatal EEG dataset collected at the University of Helsinki Hospital, which has a high discriminative ability to detect seizures, with an accuracy of 92.34%, precision of 93.61%, recall rate of 98.74%, and F1-score of 95.77%, respectively. The results show that unsupervised learning by CAE is beneficial to the characterization of EEG signals, and the proposed Fd-CAE method significantly improves classification performance.

Supervised and Unsupervised Deep Learning Approaches for EEG Seizure Prediction

Epilepsy affects more than 50 million people worldwide, making it one of the world's most prevalent neurological diseases. The main symptom of epilepsy is seizures, which occur abruptly and can cause serious injury or death. The ability to predict the occurrence of an epileptic seizure could alleviate many risks and stresses people with epilepsy face. We formulate the problem of detecting preictal (or pre-seizure) with reference to normal EEG as a precursor to incoming seizure. To this end, we developed several supervised deep learning approaches model to identify preictal EEG from normal EEG. We further develop novel unsupervised deep learning approaches to train the models on only normal EEG, and detecting pre-seizure EEG as an anomalous event. These deep learning models were trained and evaluated on two large EEG seizure datasets in a person-specific manner. We found that both supervised and unsupervised approaches are feasible; however, their performance varies depending on the patient, approach and architecture. This new line of research has the potential to develop therapeutic interventions and save human lives.

An effective fusion model for seizure prediction: GAMRNN

The early prediction of epileptic seizures holds paramount significance in patient care and medical research. Extracting useful spatial-temporal features to facilitate seizure prediction represents a primary challenge in this field. This study proposes GAMRNN, a novel methodology integrating a dual-layer gated recurrent unit (GRU) model with a convolutional attention module. GAMRNN aims to capture intricate spatial-temporal characteristics by highlighting informative feature channels and spatial pattern dynamics. We employ the Lion optimization algorithm to enhance the model's generalization capability and predictive accuracy. Our evaluation of GAMRNN on the widely utilized CHB-MIT EEG dataset demonstrates its effectiveness in seizure prediction. The results include an impressive average classification accuracy of 91.73%, sensitivity of 88.09%, specificity of 92.09%, and a low false positive rate of 0.053/h. Notably, GAMRNN enables early seizure prediction with a lead time ranging from 5 to 35 min, exhibiting remarkable performance improvements compared to similar prediction models.

A Novel Deep Learning Method Based on an Overlapping Time Window Strategy for Brain-Computer Interface-Based Stroke Rehabilitation

Globally, stroke is a leading cause of death and disability. The classification of motor intentions using brain activity is an important task in the rehabilitation of stroke patients using brain-computer interfaces (BCIs). This paper presents a new method for model training in EEG-based BCI rehabilitation by using overlapping time windows. For this aim, three different models, a convolutional neural network (CNN), graph isomorphism network (GIN), and long short-term memory (LSTM), are used for performing the classification task of motor attempt (MA). We conducted several experiments with different time window lengths, and the results showed that the deep learning approach based on overlapping time windows achieved improvements in classification accuracy, with the LSTM combined vote-counting strategy (VS) having achieved the highest average classification accuracy of 90.3% when the window size was 70. The results verified that the overlapping time window strategy is useful for increasing the efficiency of BCI rehabilitation.

Big data analysis and artificial intelligence in epilepsy - common data model analysis and machine learning-based seizure detection and forecasting

There has been significant interest in big data analysis and artificial intelligence (AI) in medicine. Ever-increasing medical data and advanced computing power have enabled the number of big data analyses and AI studies to increase rapidly. Here we briefly introduce epilepsy, big data, and AI and review big data analysis using a common data model. Studies in which AI has been actively applied, such as those of electroencephalography epileptiform discharge detection, seizure detection, and forecasting, will be reviewed. We will also provide practical suggestions for pediatricians to understand and interpret big data analysis and AI research and work together with technical expertise.

Data Augmentation for Deep Neural Networks Model in EEG Classification Task: A Review

Classification of electroencephalogram (EEG) is a key approach to measure the rhythmic oscillations of neural activity, which is one of the core technologies of brain-computer interface systems (BCIs). However, extraction of the features from non-linear and non-stationary EEG signals is still a challenging task in current algorithms. With the development of artificial intelligence, various advanced algorithms have been proposed for signal classification in recent years. Among them, deep neural networks (DNNs) have become the most attractive type of method due to their end-to-end structure and powerful ability of automatic feature extraction. However, it is difficult to collect large-scale datasets in practical applications of BCIs, which may lead to overfitting or weak generalizability of the classifier. To address these issues, a promising technique has been proposed to improve the performance of the decoding model based on data augmentation (DA). In this article, we investigate recent studies and development of various DA strategies for EEG classification based on DNNs. The review consists of three parts: what kind of paradigms of EEG-based on BCIs are used, what types of DA methods are adopted to improve the DNN models, and what kind of accuracy can be obtained. Our survey summarizes the current practices and performance outcomes that aim to promote or guide the deployment of DA to EEG classification in future research and development.

A Semi-Supervised Few-Shot Learning Model for Epileptic Seizure Detection

In the past decade, the rapid development of machine learning has dramatically improved the performance of epileptic detection with Electroencephalography (EEG). However, only a small amount of labeled epileptic data is available for training because labeling requires numerous neurologists. This paper proposes a one-step semi-supervised epilepsy detection system to reduce the labeling cost by fully utilizing the unlabeled data. The proposed neural network training strategy enables a more robust and accurate decision boundary by forcing the consistency of the double predictions on the same unlabeled data. The results show that the Area Under Receiver Operating Characteristic (AUROC) curves of our proposed model are 10.3% and 4.9% higher than the supervised methods on CHB-MIT and Kaggle datasets, respectively.

Closed-Loop Neural Prostheses With On-Chip Intelligence: A Review and a Low-Latency Machine Learning Model for Brain State Detection

The application of closed-loop approaches in systems neuroscience and therapeutic stimulation holds great promise for revolutionizing our understanding of the brain and for developing novel neuromodulation therapies to restore lost functions. Neural prostheses capable of multi-channel neural recording, on-site signal processing, rapid symptom detection, and closed-loop stimulation are critical to enabling such novel treatments. However, the existing closed-loop neuromodulation devices are too simplistic and lack sufficient on-chip processing and intelligence. In this paper, we first discuss both commercial and investigational closed-loop neuromodulation devices for brain disorders. Next, we review state-of-the-art neural prostheses with on-chip machine learning, focusing on application-specific integrated circuits (ASIC). System requirements, performance and hardware comparisons, design trade-offs, and hardware optimization techniques are discussed. To facilitate a fair comparison and guide design choices among various on-chip classifiers, we propose a new energy-area (E-A) efficiency figure of merit that evaluates hardware efficiency and multi-channel scalability. Finally, we present several techniques to improve the key design metrics of tree-based on-chip classifiers, both in the context of ensemble methods and oblique structures. A novel Depth-Variant Tree Ensemble (DVTE) is proposed to reduce processing latency (e.g., by 2.5× on seizure detection task). We further develop a cost-aware learning approach to jointly optimize the power and latency metrics. We show that algorithm-hardware co-design enables the energy- and memory-optimized design of tree-based models, while preserving a high accuracy and low latency. Furthermore, we show that our proposed tree-based models feature a highly interpretable decision process that is essential for safety-critical applications such as closed-loop stimulation.

An epileptic seizure prediction model based on a time-wise attention simulation module and a pretrained ResNet

Epilepsy is a neurological disorder that affects approximately 1% of the world's populations. Epilepsy prediction has been of great interest as it can identify and warn of an upcoming seizure, and to reduce the burden of the unpredictability of seizures. In this paper, we proposed a new seizure prediction model, TASM_ResNet, based on a time-wise attention simulation module and a pre-trained ResNet, using intracranial EEG signals. The simulation module with a time-wise attention was designed to convert EEG data into image like data and extract temporal features from raw data. Pre-trained ResNet was applied to reduce the amount of training data without initial training. Moreover, since the data is extremely imbalanced, we used an improved focal loss (FL) instead of the cross-entropy loss and investigated the optimal parameters for FL. Compared with a state-of-art CNN model, our proposed model achieved a better average AUC of 0.877. Moreover, our results demonstrated that EEG signals can be migrated to the image network which was pre-trained on large data set through a simulation module.

A Survey on Deep Learning-Based Short/Zero-Calibration Approaches for EEG-Based Brain-Computer Interfaces

Brain-computer interfaces (BCIs) utilizing machine learning techniques are an emerging technology that enables a communication pathway between a user and an external system, such as a computer. Owing to its practicality, electroencephalography (EEG) is one of the most widely used measurements for BCI. However, EEG has complex patterns and EEG-based BCIs mostly involve a cost/time-consuming calibration phase; thus, acquiring sufficient EEG data is rarely possible. Recently, deep learning (DL) has had a theoretical/practical impact on BCI research because of its use in learning representations of complex patterns inherent in EEG. Moreover, algorithmic advances in DL facilitate short/zero-calibration in BCI, thereby suppressing the data acquisition phase. Those advancements include data augmentation (DA), increasing the number of training samples without acquiring additional data, and transfer learning (TL), taking advantage of representative knowledge obtained from one dataset to address the so-called data insufficiency problem in other datasets. In this study, we review DL-based short/zero-calibration methods for BCI. Further, we elaborate methodological/algorithmic trends, highlight intriguing approaches in the literature, and discuss directions for further research. In particular, we search for generative model-based and geometric manipulation-based DA methods. Additionally, we categorize TL techniques in DL-based BCIs into explicit and implicit methods. Our systematization reveals advances in the DA and TL methods. Among the studies reviewed herein, ~45% of DA studies used generative model-based techniques, whereas ~45% of TL studies used explicit knowledge transferring strategy. Moreover, based on our literature review, we recommend an appropriate DA strategy for DL-based BCIs and discuss trends of TLs used in DL-based BCIs.

Deep Convolutional Neural Network Based Interictal-Preictal Electroencephalography Prediction: Application to Focal Cortical Dysplasia Type-II

We aimed to differentiate between the interictal and preictal states in epilepsy patients with focal cortical dysplasia (FCD) type-II using deep learning-based classifiers based on intracranial electroencephalography (EEG). We also investigated the practical conditions for high interictal-preictal discriminability in terms of spatiotemporal EEG characteristics and data size efficiency. Intracranial EEG recordings of nine epilepsy patients with FCD type-II (four female, five male; mean age: 10.7 years) were analyzed. Seizure onset and channel ranking were annotated by two epileptologists. We performed three consecutive interictal-preictal classification steps by varying the preictal length, number of electrodes, and sampling frequency with convolutional neural networks (CNN) using 30 s time-frequency data matrices. Classification performances were evaluated based on accuracy, F1 score, precision, and recall with respect to the above-mentioned three parameters. We found that (1) a 5 min preictal length provided the best classification performance, showing a remarkable enhancement of >13% on average compared to that with the 120 min preictal length; (2) four electrodes provided considerably high classification performance with a decrease of only approximately 1% on average compared to that with all channels; and (3) there was minimal performance change when quadrupling the sampling frequency from 128 Hz. Patient-specific performance variations were noticeable with respect to the preictal length, and three patients showed above-average performance enhancements of >28%. However, performance enhancements were low with respect to both the number of electrodes and sampling frequencies, and some patients showed at most 1–2% performance change. CNN-based classifiers from intracranial EEG recordings using a small number of electrodes and efficient sampling frequency are feasible for predicting the interictal-preictal state transition preceding seizures in epilepsy patients with FCD type-II. Preictal lengths affect the predictability in a patient-specific manner; therefore, pre-examinations for optimal preictal length will be helpful in seizure prediction.

Artificial Intelligence and Computational Approaches for Epilepsy

Studies on treatment of epilepsy have been actively conducted in multiple avenues, but there are limitations in improving its efficacy due to between-subject variability in which treatment outcomes vary from patient to patient. Accordingly, there is a growing interest in precision medicine that provides accurate diagnosis for seizure types and optimal treatment for an individual epilepsy patient. Among these approaches, computational studies making this feasible are rapidly progressing in particular and have been widely applied in epilepsy. These computational studies are being conducted in two main streams: 1) artificial intelligence-based studies implementing computational machines with specific functions, such as automatic diagnosis and prognosis prediction for an individual patient, using machine learning techniques based on large amounts of data obtained from multiple patients and 2) patient-specific modeling-based studies implementing biophysical in-silico platforms to understand pathological mechanisms and derive the optimal treatment for each patient by reproducing the brain network dynamics of the particular patient per se based on individual patient's data. These computational approaches are important as it can integrate multiple types of data acquired from patients and analysis results into a single platform. If these kinds of methods are efficiently operated, it would suggest a novel paradigm for precision medicine.