West syndrome (WS), is a severe developmental and epileptic encephalopathy that occurs in infancy or early childhood, affecting 2–5/10,000 newborns (1). It is characterized by clustered epileptic spasms, interictal hypsarrhythmia on scalp electroencephalogram (EEG), and neurodevelopmental arrest or psychomotor delay (2). WS is drug-resistant with only a third of patients experiencing seizure remission after treatement. Cognitive impairments persist, and fewer than 20% of patients can attend school normally, while approximately 20% of patients develop autism spectrum disorder (3). At least 20% of WS patients do not survive to adulthood (4). Early diagnosis and appropriate intervention are crucial for achieving a positive outcome for WS patients.

Hypsarrhythmia, as first introduced by Gibbs and Gibbs (5), is regarded as the EEG hallmark of WS. It is characterized by disorganized high-amplitude slow waves and multi-focal spikes without interhemispheric synchronization. Many variants of hypsarrhythmia, including typical hypsarrhythmia and modified hypsarrhythmia, have been described (6). Despite its common use, the term “hypsarrhythmia” lacks quantifiable definitions, therefore, clinicians rely on visual interpretation of electrographic findings for the diagnosis of WS. Yet, due to the wide variability in temporal and spatial patterns, the interrater reliability is remarkedly low (7, 8).

Advances in digital EEG recording and signal processing techniques have enabled in-depth investigation of brain signals. The derivation of a variety of quantifiable features not only enhances our understanding of neurophysiology in health and diseases (9) but also provides alternative or supplementary biomarkers for clinical evaluations (10). For instance, recent studies revealed that WS patients exhibit stronger delta/alpha activity (11) and enhanced high-frequency oscillation (12, 13). It has been also proposed that a stronger coupling between high-frequency oscillation with slow wave activity could serve as a predictor of drug responsiveness in WS patients (14). Additionally, WS patients demonstrate abnormal functional connectivity (15), and reduced EEG complexity, i.e., lower entropy (10, 16). However, further studies are necessary to validate whether these observations could be repeated in larger cohorts and whether they could be used for differential diagnosis in clinical practices.

The etiology of WS is diverse and can be attributed to genetic, structural, infectious, metabolic, and immunologic abnormalities. Accurate etiologic determination is crucial for selecting the appropriate intervention strategy and achieving positive outcomes. However, in approximately one-third of WS patients, the etiological cause remains unknown even after extensive investigation (17). Furthermore, different factors may act in association and cannot be reliably distinguished. It is also unclear whether different etiological causes could result in distinctive EEG patterns in WS.

In this study, we aimed to investigate whether there are specific interictal EEG biomarkers of WS independent of hypsarrhythmia and whether these EEG patterns could be attributed to structural or genetic abnormalities. To this end, we analyzed EEG recordings from 31 WS patients with structural brain aberrancies and/or genetic abnormalities and compared them to 20 age-matched controls. Our focus was on linear features including power spectrum, cross-frequency coupling, bi-channel coherences as well as non-linear characteristics: the entropy of EEG data. Subsequently, we explored the correlation of these features with structural and genetic abnormalities.

In the present study, we observed a significant increase in global theta activity and a decrease in delta activity coupled with increased alpha activity in occipital regions in interictal EEG recordings of WS patients compared to controls. Additionally, beta-band activity was significantly lower in WS patients, although no specific region was identified in the subsequent channel-wise analysis. Delta activities are proposed to be associated with functional cortical inhibition (21), the attenuation of delta power in the occipital lobe could therefore reflect impaired inhibition specifically within the visual cortex. This disinhibition might also contribute to the excessive neuronal firing and the spread of seizure activity observed in WS. Our findings differ from previous reports. Burroughs et al. reported strengthened delta activity and attenuated theta, beta, and gamma activity in WS patients (15), while Smith et al. reported that WS patients showed global and broadband strengthening of the EEG power (10). The discrepancy between the present study and previous reports may be attributed to the exclusion of hypsarrhythmia activity in the present study, as hypsarrhythmia EEG patterns in WS patients have been associated with increased broadband amplitude and strengthened delta/alpha activities (11). Converging evidence suggests that theta activities are associated with cognitive conflict or errors (22), and the association begins as early as infancy (23). Dysregulation of theta activities in WS patients may reflect a continuous uncertain state, potentially leading to anxiety (24) and cognitive deficiency (25). High-gamma oscillations (40–150 Hz) have been associated with epileptogenicity and pathological brain lesions in WS patients (13, 26). Unfortunately, these were not assessed in the present study due to the limited sampling rate.

Cross-frequency coupling between neuronal oscillations plays a crucial role in neural functioning (27). High-frequency brain activities are associated with local processing, while low-frequency rhythms can travel across distant brain regions. Therefore, cross-frequency coupling facilitates the exchange of information between large-scale brain networks and local cortical processes, enabling functional integration across different spatiotemporal scales and effective computation (28). A previous study demonstrated that the coupling between interictal fast ripples (100–600 Hz) and the phase of slow wave activities (2–3 Hz) could differentiate treatment refractoriness in WS patients (14). In our present study, we observed an overall decrease in cross-frequency coupling across all assessed band pairs in WS patients, although no significant difference was identified at the single-channel level. This finding indicates that WS patients face significant challenges in coordinating global and local neural computations, which aligns with the neuro-psychomotor deficiencies commonly seen in WS patients.

Coherence between two brain regions indicates brain connectivity and reflects the functional integrity of the brain. Burroughs et al. reported that short-distance bi-channel coherence is decreased in frontal regions, whereas longer-distance bi-channel coherence is increased at broadband in the EEG of WS patients (15). The authors therefore proposed aberrant coherences may contribute to the developmental delay in WS. Moreover, Smith et al. recently reported that more than 80% of inter-channel connections at the delta band were increased in WS patients compared to controls (10). In our present study, we found that bi-channel coherence in EEG recordings of WS patients was significantly elevated at lower frequencies (i.e., delta-theta band), while it was decreased at higher frequencies (i.e., beta band). Despite methodological differences, increased inter-channel connectivity at the beta band in EEG recordings of WS patients was consistent among previous and present studies. This indicates that increased connectivity at the beta-band could serve as a biomarker for WS patients, whereas further investigations are warranted to elucidate the specific mechanism of connectivity alterations in WS and their functional and therapeutic implications.

Nonlinear characteristics are of particular significance in the investigation of intricate systems such as the human brain. Notably, entropy analysis of EEG signals has proven effective in the classification of various brain states, including distinguishing between awake and anesthetized states (29), as well as identifying healthy and epileptic conditions (30). A recent study by Chu et al. revealed a global reduction in EEG complexity, as assessed by multiscale entropy, in WS patients, which was found to be reversible with anti-epileptic drugs (16). Smith et al. observed that WS patients exhibited significantly higher Shannon entropy at the delta band, but decreased permutation entropy (PeEn) at both the delta and theta bands (10). In the present study, we investigated four entropy estimation methods—Approximate Entropy (ApEn), Sample Entropy (SaEn), Permutation Entropy (PeEn), and Wavelet Entropy (WaEn)—in interictal EEG recordings of WS patients. Interestingly, PeEn emerged as the most effective nonlinear feature in differentiating WS patients from controls, displaying minimal inter-subject variation and aligning with previous findings. While both ApEn and SaEn did not demonstrate significant differences, WaEn exhibited a slight increase in WS patients. PeEn assesses signal complexity by analyzing patterns of relative change between neighboring values, rather than absolute distances within an embedding space as with ApEn and SaEn (31), rendering it more robust when applied to noisy data such as EEG recordings (29). Although further research is necessary to validate these findings, the results suggest that permutation entropy holds significant promise as a nonlinear EEG biomarker for WS.

Variant patterns of hypsarrhythmia have been implicated in the etiology of WS (32). However, the correlation of computational EEG features with the etiology of WS has not been previously reported. One of the objectives of this study was to address this gap by examining the impact of genetic and structural aberrancy on abnormal linear and non-linear interictal EEG features in WS patients. Our findings revealed only 2 out of 969 features to be linked to genetic brain abnormalities in WS, and none were associated with causative structural variants. These results suggest that diverse etiological factors may give rise to converging pathology, leading to common symptomatic manifestations in individuals with WS.

The present study is subjected to several limitations. Firstly, the relatively small size of the patient and control cohorts may result in low statistical power. Secondly, the limited sampling rate prevented the analysis of EEG components with higher frequency. Thirdly, the present study only analyzed data from the pre-treatment state, thus further exploration is needed to understand how these findings may reflect drug response or prognosis.

In summary, interictal EEG patterns in WS patients revealed prominent global elevation of theta activity and decreased delta activity, coupled with increased alpha activity in occipital regions. Cross-frequency coupling was weaker in WS patients, and they also exhibited minor bi-channel correlation attenuation. Among the entropy estimation methods studies, PeEn, but not ApEn, SaEn, or WaEn, demonstrated a global reduction of complexity of interictal EEG of WS patients. While structural aberrancies were not found to contribute to alterations in interictal EEG in WS patients, genetic abnormalities were associated with subtle shifts in the power spectrum and inter-channel coherences. This study proposes that a combined global strengthening of theta activity and global reduction of permutation entropy could serve as a computational EEG biomarker of West syndrome.