Normal variants and artifacts: Importance in EEG interpretation

Computer-assisted analysis of routine EEG to identify hidden biomarkers of epilepsy: A systematic review

Background

Computational analysis of routine electroencephalogram (rEEG) could improve the accuracy of epilepsy diagnosis. We aim to systematically assess the diagnostic performances of computed biomarkers for epilepsy in individuals undergoing rEEG.

Methods

We searched MEDLINE, EMBASE, EBM reviews, IEEE Explore and the grey literature for studies published between January 1961 and December 2022. We included studies reporting a computational method to diagnose epilepsy based on rEEG without relying on the identification of interictal epileptiform discharges or seizures. Diagnosis of epilepsy as per a treating physician was the reference standard. We assessed the risk of bias using an adapted QUADAS-2 tool.

Results

We screened 10 166 studies, and 37 were included. The sample size ranged from 8 to 192 (mean=54). The computed biomarkers were based on linear (43%), non-linear (27%), connectivity (38%), and convolutional neural networks (10%) models. The risk of bias was high or unclear in all studies, more commonly from spectrum effect and data leakage. Diagnostic accuracy ranged between 64% and 100%. We observed high methodological heterogeneity, preventing pooling of accuracy measures.

Conclusion

The current literature provides insufficient evidence to reliably assess the diagnostic yield of computational analysis of rEEG.

Significance

We provide guidelines regarding patient selection, reference standard, algorithms, and performance validation.

Generalizability of electroencephalographic interpretation using artificial intelligence: An external validation study

OBJECTIVE

The automated interpretation of clinical electroencephalograms (EEGs) using artificial intelligence (AI) holds the potential to bridge the treatment gap in resource-limited settings and reduce the workload at specialized centers. However, to facilitate broad clinical implementation, it is essential to establish generalizability across diverse patient populations and equipment. We assessed whether SCORE-AI demonstrates diagnostic accuracy comparable to that of experts when applied to a geographically different patient population, recorded with distinct EEG equipment and technical settings.

METHODS

We assessed the diagnostic accuracy of a "fixed-and-frozen" AI model, using an independent dataset and external gold standard, and benchmarked it against three experts blinded to all other data. The dataset comprised 50% normal and 50% abnormal routine EEGs, equally distributed among the four major classes of EEG abnormalities (focal epileptiform, generalized epileptiform, focal nonepileptiform, and diffuse nonepileptiform). To assess diagnostic accuracy, we computed sensitivity, specificity, and accuracy of the AI model and the experts against the external gold standard.

RESULTS

We analyzed EEGs from 104 patients (64 females, median age = 38.6 [range = 16-91] years). SCORE-AI performed equally well compared to the experts, with an overall accuracy of 92% (95% confidence interval [CI] = 90%-94%) versus 94% (95% CI = 92%-96%). There was no significant difference between SCORE-AI and the experts for any metric or category. SCORE-AI performed well independently of the vigilance state (false classification during awake: 5/41 [12.2%], false classification during sleep: 2/11 [18.2%]; p = .63) and normal variants (false classification in presence of normal variants: 4/14 [28.6%], false classification in absence of normal variants: 3/38 [7.9%]; p = .07).

SIGNIFICANCE

SCORE-AI achieved diagnostic performance equal to human experts in an EEG dataset independent of the development dataset, in a geographically distinct patient population, recorded with different equipment and technical settings than the development dataset.

Reaching the ceiling? Empirical scaling behaviour for deep EEG pathology classification

Machine learning techniques, particularly deep convolutional neural networks (ConvNets), are increasingly being used to automate clinical EEG analysis, with the potential to reduce the clinical burden and improve patient care. However, further research is required before they can be used in clinical settings, particularly regarding the impact of the number of training samples and model parameters on their testing error. To address this, we present a comprehensive study of the empirical scaling behaviour of ConvNets for EEG pathology classification. We analysed the testing error with increasing the training samples and model size for four different ConvNet architectures. The focus of our experiments is width scaling, and we have increased the number of parameters to up to 1.8 million. Our evaluation was based on two publicly available datasets: the Temple University Hospital (TUH) Abnormal EEG Corpus and the TUH Abnormal Expansion Balanced EEG Corpus, which together contain 10,707 training samples. The results show that the testing error follows a saturating power-law with both model and dataset size. This pattern is consistent across different datasets and ConvNet architectures. Furthermore, empirically observed accuracies saturate at 85%-87%, which may be due to an imperfect inter-rater agreement on the clinical labels. The empirical scaling behaviour of the test performance with dataset and model size has significant implications for deep EEG pathology classification research and practice. Our findings highlight the potential of deep ConvNets for high-performance EEG pathology classification, and the identified scaling relationships provide valuable recommendations for the advancement of automated EEG diagnostics.

Overcoming traps and pitfalls leading to misinterpretation of normal EEG variants and variation of the background activity

Normal EEG variants, especially the epileptiform variants, can be challenging to interpret because they often have sharp contours and may be confused with "epileptic" interictal activities. However, they can be recognized by the fact that "most spikes or sharp wave discharges of clinical import are followed by a slow wave or a series of slow deflections" (Maulsby, 1971). If there is no wave after the spike, electroencephalographers should be suspicious of artifacts and normal EEG variants. Most normal EEG variants display a single rhythm with the same frequency within the pattern and the morphology remains stable throughout the entire EEG recording with repetition of the same pattern. In case of doubt or difficulties with a standard EEG, it is recommended to undergo an EEG that includes sleep stages with or without sleep deprivation. Finally, epileptiform is an ambiguous term corresponding to an electroencephalographic trait. Epileptiform does not imply a pathological condition, including epilepsy. The clinical context remains the most paramount in the diagnosis of epilepsy. In this article, we propose a set of rules and guidelines to identify normal EEG variants in EEG tracings and normal variation of the background activity. It is not easy to accurately assign a specific/precise name to all EEG activity, but with an orderly approach to EEG that involves using a set of criteria, nonepileptic activity can be identified.

The effects of slow wave sleep characteristics on semantic, episodic, and procedural memory in people with epilepsy

Slow wave sleep (SWS) is highly relevant for verbal and non-verbal/spatial memory in healthy individuals, but also in people with epilepsy. However, contradictory findings exist regarding the effect of seizures on overnight memory retention, particularly relating to procedural and non-verbal memory, and thorough examination of episodic memory retention with ecologically valid tests is missing. This research explores the interaction of SWS duration with epilepsy-relevant factors, as well as the relation of spectral characteristics of SWS on overnight retention of procedural, verbal, and episodic memory. In an epilepsy monitoring unit, epilepsy patients (N = 40) underwent learning, immediate and 12 h delayed testing of memory retention for a fingertapping task (procedural memory), a word-pair task (verbal memory), and an innovative virtual reality task (episodic memory). We used multiple linear regression to examine the impact of SWS duration, spectral characteristics of SWS, seizure occurrence, medication, depression, seizure type, gender, and epilepsy duration on overnight memory retention. Results indicated that none of the candidate variables significantly predicted overnight changes for procedural memory performance. For verbal memory, the occurrence of tonic-clonic seizures negatively impacted memory retention and higher psychoactive medication load showed a tendency for lower verbal memory retention. Episodic memory was significantly impacted by epilepsy duration, displaying a potential nonlinear impact with a longer duration than 10 years negatively affecting memory performance. Higher drug load of anti-seizure medication was by tendency related to better overnight retention of episodic memory. Contrary to expectations longer SWS duration showed a trend towards decreased episodic memory performance. Analyses on associations between memory types and EEG band power during SWS revealed lower alpha-band power in the frontal right region as significant predictor for better episodic memory retention. In conclusion, this research reveals that memory modalities are not equally affected by important epilepsy factors such as duration of epilepsy and medication, as well as SWS spectral characteristics.

Can epilepsy affect normal EEG variants? A comparative study between subjects with and without epilepsy

OBJECTIVES

To compare the prevalence of benign EEG variants (BEVs) between epileptic and non-epileptic subjects.

METHODS

A prospective, observational EEG study of 1,163 consecutive patients, using the 10-20 international system with systematically two additional anterior/inferior temporal electrodes. The video-EEG monitoring duration was between 24 h and eight days.

RESULTS

We identified 917 (78.9%) epileptic patients (mean age: 33.42 ± 15.5 years; females: 53.4%) and 246 (21.2%) non-epileptic patients (mean age: 35.6 ± 18.75 years; females: 54.9%). Despite a shorter mean duration of the EEG recordings, the prevalence of BEVs was higher in non-epileptic vs. epileptic patients (73.2% vs. 57.8%, p = 0.000011). This statistical difference was confirmed for lambda waves (23.6% in the non-epilepsy group vs. 14.8% in the epilepsy group, p = 0.001), POSTs (50.8% vs. 32.5%, p < 0.000001), wicket spikes (20.3% vs. 13.6%, p = 0.009) in particular in NREM and REM sleep, and 14- and 6-Hz positive bursts (13% vs. 7.1% p = 0.003). Mu rhythm was observed at the same frequency in both groups (21.1% in the non-epilepsy group vs. 22.7% in the epilepsy group). There was no difference between the two groups for rarer rhythms, such as rhythmic mid-temporal theta burst of drowsiness, small sharp spikes, and midline theta rhythm.

CONCLUSIONS

There was no increase in any of the BEVs in the epilepsy group. On the contrary, BEVs were more frequent and diversified in the non-epilepsy group. Epilepsy may negatively affect the occurrence of the most common BEVs, with the exception of the mu rhythm, which is present in about one-fifth of the population with or without epilepsy.

Models of care and relevance of territorial management in assisting persons with epilepsy

Epilepsy is a widespread social disease that affects people of all ages and often involves both diagnostic and therapeutic difficulties. Beyond seizure control, it is necessary to ensure people with epilepsy a good quality of life and respect for human rights, seeking to increase self-management capacity and break down stigma. People with epilepsy should have privileged access to specialized epilepsy centers, where multidisciplinary care is possible. These centers, organized by different levels of complexity, should be uniformly distributed throughout the country and networked together. The scientific community and health care organizations must therefore design all necessary strategies so that knowledge about epilepsy improves among the general population and the most effective pathways of care are effectively implemented.

Pitfalls in scalp EEG: Current obstacles and future directions

Although electroencephalography (EEG) serves a critical role in the evaluation and management of seizure disorders, it is commonly misinterpreted, resulting in avoidable medical, social, and financial burdens to patients and health care systems. Overinterpretation of sharply contoured transient waveforms as being representative of interictal epileptiform abnormalities lies at the core of this problem. However, the magnitude of these errors is amplified by the high prevalence of paroxysmal events exhibited in clinical practice that compel investigation with EEG. Neurology training programs, which vary considerably both in the degree of exposure to EEG and the composition of EEG didactics, have not effectively addressed this widespread issue. Implementation of competency-based curricula in lieu of traditional educational approaches may enhance proficiency in EEG interpretation amongst general neurologists in the absence of formal subspecialty training. Efforts in this regard have led to the development of a systematic, high-fidelity approach to the interpretation of epileptiform discharges that is readily employable across medical centers. Additionally, machine learning techniques hold promise for accelerating accurate and reliable EEG interpretation, particularly in settings where subspecialty interpretive EEG services are not readily available. This review highlights common diagnostic errors in EEG interpretation, limitations in current educational paradigms, and initiatives aimed at resolving these challenges.

Shorter duration of slow wave sleep is related to symptoms of depression in patients with epilepsy

Slow wave sleep duration and spectral abnormalities are related to both epilepsy and depression, but it is unclear how depressive symptoms in patients with epilepsy are affected by slow wave sleep duration and clinical factors, and how the spectral characteristics of slow wave sleep reflect a potential interaction of epilepsy and depression. Long-term video-EEG monitoring was conducted in 51 patients with focal epilepsy, 13 patients with generalized epilepsy, and 9 patients without epilepsy. Slow wave sleep segments were manually marked in the EEG and duration as well as EEG power spectra were extracted. Depressive symptoms were documented with the Beck Depression Inventory (BDI). At least mild depressive symptoms (BDI > 9) were found among 23 patients with focal epilepsy, 5 patients with generalised epilepsy, and 6 patients who had no epilepsy diagnosis. Slow wave sleep duration was shorter for patients with at least mild depressive symptoms (p =.004), independently from epilepsy diagnosis, antiseizure medication, age, and sex. Psychoactive medication was associated with longer slow wave sleep duration (p =.008). Frontal sigma band power (13-15 Hz) during slow wave sleep was higher for patients without epilepsy and without depressive symptoms as compared to patients without depressive symptoms but with focal epilepsy (p =.005). Depressive symptoms affect slow wave sleep duration of patients with epilepsy similarly as in patients without epilepsy. Since reduced slow wave sleep can increase the likelihood of seizure occurrence, these results stress the importance of adequate treatment for patients with epilepsy who experience depressive symptoms.

Interictal epileptiform discharges in Alzheimer's disease: prevalence, relevance, and controversies

Alzheimer's disease (AD) is the most common type of dementia and remains an incurable, progressive disease with limited disease-modifying interventions available. In patients with AD, interictal epileptiform discharges (IEDs) have been identified in up to 54% of combined cohorts of mild cognitive impairment (MCI) or mild dementia and are a marker of a more aggressive disease course. Studies assessing the role of IEDs in AD are limited by the lack of standardization in the definition of IEDs or the different neurophysiologic techniques used to capture them. IEDs are an appealing treatment target given the availability of EEG and anti-seizure medications. There remains uncertainty regarding when to treat IEDs, the optimal drug and dose for treatment, and the impact of treatment on disease course. This review covers the state of knowledge of the field of IEDs in AD, and the steps needed to move the field forward.