Seizure forecasting: Where do we stand?

Artificial intelligence applied to electroencephalography in epilepsy

Artificial intelligence (AI) is progressively transforming all fields of medicine, promising substantial changes in clinical practice. In the context of epilepsy, electroencephalography (EEG), a technique used for over a century, has historically been resistant to automated analysis due to the complexity of the signals and the challenges posed by artifact management. While the human eye excels at recognizing patterns, algorithms have demonstrated superior capabilities in detecting and characterizing specific features, such as long-term dynamics and synchrony. Furthermore, the advent of wearable EEG devices has led to an exponential increase in data volume, surpassing the limits of visual interpretation. AI algorithms are now being developed to address these limitations, offering enhanced efficiency in both identifying subtle signal features and managing massive datasets. This review explores the fundamental principles of AI and its transformative potential in the field of EEG. It discusses the implications and the current limitations, including improvements limited to aggregation of already known knowledge, for epilepsy diagnosis, medical and surgical treatment, and innovative approaches to patient monitoring, including seizure forecasting, highlighting how AI is poised to redefine the management of epilepsy.

Seizure Detection Devices

Goals of automated detection of epileptic seizures using wearable devices include objective documentation of seizures, prevention of sudden unexpected death in epilepsy (SUDEP) and seizure-related injuries, obviating both the unpredictability of seizures and potential social embarrassment, and finally to develop seizure-triggered on-demand therapies. Automated seizure detection devices are based on the analysis of EEG signals (scalp-EEG, subcutaneous EEG and intracranial EEG), of motor manifestations of seizures (surface EMG, accelerometry), and of physiologic autonomic changes caused by seizures (heart and respiration rate, oxygen saturation, sweat secretion, body temperature). While the detection of generalized tonic-clonic and of focal to bilateral tonic-clonic seizures can be achieved with high sensitivity and low false alarm rates, the detection of focal seizures is still suboptimal, especially in the everyday ambulatory setting. Multimodal seizure detection devices in general provide better performance than devices based on single measurement parameters. Long-term use of seizure detection devices in home environments helps to improve the accuracy of seizure diaries and to reduce seizure-related injuries, while evidence for prevention of SUDEP is still lacking. Automated seizure detection devices are generally well accepted by patients and caregivers.

Necessary for seizure forecasting outcome metrics: Seizure frequency and benchmark model

BACKGROUND

This study aims to illustrate the connection between seizure frequency (SF) and performance metrics in seizure forecasting, and to compare the effectiveness of a moving average (MA) model versus the commonly used permutation benchmark.

METHODS

Metrics of calibration and discrimination were computed for each dataset, comparing MA and permutation performance across SF values. Three datasets were used: (1) self-reported seizure diaries from 3994 Seizure Tracker patients, (2) automatically detected and sometimes manually reported or edited generalized tonic-clonic seizures from 2350 Empatica Embrace 2 and Mate App users, and (3) simulated datasets with varying SFs.

RESULTS

Most metrics were found to depend on SF. The MA model outperformed or matched the permutation model in all cases. These more advanced metrics show that comparison to permutation will falsely elevate poor forecasting models.

CONCLUSIONS

The findings highlight SF's role in seizure forecasting accuracy and the MA model's suitability as a benchmark. This study underscores the need for considering patient SF in forecasting studies and suggests the MA model may provide a better standard for evaluating future seizure forecasting models.

Epileptic seizure forecasting with wearable-based nocturnal sleep features

OBJECTIVE

Non-invasive biomarkers have recently shown promise for seizure forecasting in people with epilepsy. In this work, we developed a seizure-day forecasting algorithm based on nocturnal sleep features acquired using a smart shirt.

METHODS

Seventy-eight individuals with epilepsy admitted to the Centre hospitalier de l'Université de Montréal epilepsy monitoring unit wore the Hexoskin biometric smart shirt during their stay. The shirt continuously measures electrocardiography, respiratory, and accelerometry activity. Ten sleep features, including sleep efficiency, sleep latency, sleep duration, time spent in non-rapid eye movement sleep (NREM) and rapid eye movement sleep (REM), wakefulness after sleep onset, average heart and breathing rates, high-frequency heart rate variability, and the number of position changes, were automatically computed using the Hexoskin sleep algorithm. Each night's features were then normalized using a reference night for each patient. A support vector machine classifier was trained for pseudo-prospective seizure-day forecasting, with forecasting horizons of 16- and 24-h to include both diurnal and nocturnal seizures (24-h) or diurnal seizures only (16-h). The algorithm's performance was assessed using a nested leave-one-patient-out cross-validation approach.

RESULTS

Improvement over chance (IoC) performances were achieved for 48.7% and 40% of patients with the 16- and 24-h forecasting horizons, respectively. For patients with IoC performances, the proposed algorithm reached mean IoC, sensitivity and time in warning of 34.3%, 86.0%, and 51.7%, respectively for the 16-h horizon, and 34.2%, 64.4% and 30.2%, respectively, for the 24-h horizon.

SIGNIFICANCE

Smart shirt-based nocturnal sleep analysis holds promise as a non-invasive approach for seizure-day forecasting in a subset of people with epilepsy. Further investigations, particularly in a residential setting with long-term recordings, could pave the way for the development of innovative and practical seizure forecasting devices.

PLAIN LANGUAGE SUMMARY

Seizure forecasting with wearable devices may improve the quality of life of people living with epilepsy who experience unpredictable, recurrent seizures. In this study, we have developed a seizure forecasting algorithm using sleep characteristics obtained from a smart shirt worn at night by a large number of hospitalized patients with epilepsy (78). A daily seizure forecast was generated following each night using machine learning methods. Our results show that around half of people with epilepsy may benefit from such an approach.

Hippocampal network activity forecasts epileptic seizures

Seizures in people with epilepsy were long thought to occur at random, but recent methods for seizure forecasting enable estimation of the likelihood of seizure occurrence over short horizons. These methods rely on days-long cyclical patterns of brain electrical activity and other physiological variables that determine seizure likelihood and that require measurement through long-term, multimodal recordings. In this retrospective cohort study of 15 adults with bitemporal epilepsy who had a device that provides chronic intracranial recordings, functional connectivity of hippocampal networks fluctuated in multiday cycles with patterns that mirrored cycles of seizure likelihood. A functional connectivity biomarker of seizure likelihood derived from 90-s recordings of background hippocampal activity generalized across individuals and forecasted 24-h seizure likelihood as accurately as cycle-based models requiring months-long baseline recordings. Larger, prospective studies are needed to validate this approach, but our results have the potential to make reliable seizure forecasts accessible to more people with epilepsy.

SzCORE: Seizure Community Open-Source Research Evaluation framework for the validation of electroencephalography-based automated seizure detection algorithms

The need for high-quality automated seizure detection algorithms based on electroencephalography (EEG) becomes ever more pressing with the increasing use of ambulatory and long-term EEG monitoring. Heterogeneity in validation methods of these algorithms influences the reported results and makes comprehensive evaluation and comparison challenging. This heterogeneity concerns in particular the choice of datasets, evaluation methodologies, and performance metrics. In this paper, we propose a unified framework designed to establish standardization in the validation of EEG-based seizure detection algorithms. Based on existing guidelines and recommendations, the framework introduces a set of recommendations and standards related to datasets, file formats, EEG data input content, seizure annotation input and output, cross-validation strategies, and performance metrics. We also propose the EEG 10-20 seizure detection benchmark, a machine-learning benchmark based on public datasets converted to a standardized format. This benchmark defines the machine-learning task as well as reporting metrics. We illustrate the use of the benchmark by evaluating a set of existing seizure detection algorithms. The SzCORE (Seizure Community Open-Source Research Evaluation) framework and benchmark are made publicly available along with an open-source software library to facilitate research use, while enabling rigorous evaluation of the clinical significance of the algorithms, fostering a collective effort to more optimally detect seizures to improve the lives of people with epilepsy.

Passive and active markers of cortical excitability in epilepsy

Electroencephalography (EEG) has been the primary diagnostic tool in clinical epilepsy for nearly a century. Its review is performed using qualitative clinical methods that have changed little over time. However, the intersection of higher resolution digital EEG and analytical tools developed in the past decade invites a re-exploration of relevant methodology. In addition to the established spatial and temporal markers of spikes and high-frequency oscillations, novel markers involving advanced postprocessing and active probing of the interictal EEG are gaining ground. This review provides an overview of the EEG-based passive and active markers of cortical excitability in epilepsy and of the techniques developed to facilitate their identification. Several different emerging tools are discussed in the context of specific EEG applications and the barriers we must overcome to translate these tools into clinical practice.