Pitfalls in ICU EEG Interpretation: A Retrospective Case Series

Abstract

Background: Electroencephalography (EEG) is increasingly used in intensive care units (ICUs), primarily for seizure detection. However, the complex clinical context of critically ill patients and the dynamic ICU environment can complicate interpretation. Purpose: To highlight common pitfalls in ICU EEG interpretation and strategies to distinguish epileptic seizures from mimics. Research Design: Retrospective case series. Study Sample: Five ICU patients undergoing continuous video EEG (cvEEG) for altered mental status, status epilepticus, or paroxysmal events, with underlying conditions including malignancy, autoimmune encephalitis, neuromodulation therapy, and drug-resistant epilepsy. Results: Physiological artifacts, such as non-epileptic rhythmic movements, and device-related artifacts (e.g., pacemakers, ventilators) can mimic epileptic seizures. Plateau waves from elevated intracranial pressure may produce deficits resembling epileptic seizures. Accurate differentiation requires attention to clinical context, EEG features, and video correlation. Maintaining a broad differential and avoiding diagnostic anchoring are also essential to prevent misdiagnosis and unnecessary antiseizure medication therapy. Conclusions: ICU EEG interpretation is challenged by artifacts and non-epileptic movements that mimic epileptic seizures. Awareness of these issues, along with a thoughtful, multidisciplinary approach, is critical for improving diagnostic accuracy and optimizing patient outcomes during cvEEG monitoring.

Introduction

Continuous video EEG (cvEEG) plays a vital role in the detection and management of epileptic seizures in critically ill patients. In the intensive care unit (ICU) setting, cvEEG is essential for identifying nonconvulsive seizures, evaluating the effectiveness of antiseizure medications (ASMs), monitoring sedation levels, and characterizing paroxysmal events, including both epileptic and nonepileptic motor symptoms. ¹ Additionally, cvEEG is used to monitor for cerebral ischemia following subarachnoid hemorrhage. ²

The adoption of cvEEG in ICUs has grown substantially in recent years, contributing to improved seizure detection and more targeted ASM therapy, which has been associated with reduced in-hospital mortality. ³ However, increased utilization has also led to challenges, including the identification of electrographic patterns that may lack clinical significance. Misinterpretation of these findings can result in diagnostic errors and unnecessary interventions. Careful patient selection is crucial when considering cvEEG, not only to optimize diagnostic yield but also to minimize technical challenges that may lead to misinterpretation. Ideal candidates include patients with unexplained persistent coma as well as those with clinical or subclinical seizures. To support appropriate use of cvEEG, risk stratification tools such as the 2HELPS2B score have been developed. ⁴ This algorithm incorporates five electrographic features along with a single clinical factor—a history of acute or remote seizures— to estimate short-term seizure risk and guide more efficient use of cvEEG.

In this retrospective case series, we reviewed five cases to highlight common pitfalls in ICU EEG interpretation and propose strategies to aid in their identification and prevention.

Case 1

A 72-year-old female with B-cell lymphoma presented with altered mental status. A STAT EEG revealed 1.5-2 Hz generalized periodic discharges (GPDs) (Figure 1A) without evolution into a definitive electrographic seizure. Given the concern for non-convulsive status epilepticus (NCSE), the electrodes from the STAT study were left in place to initiate cvEEG monitoring, in accordance with institutional protocol. While the cvEEG showed a similar pattern, the patient also exhibited continuous large-amplitude rhythmic movements in the left upper extremity. These movements were not consistently time-locked to the GPDs. After administration of 2 mg intravenous lorazepam, the GPDs resolved. However, the movements persisted intermittently, and the patient remained confused.

Figure 1.

Title. EEG Epochs Captured During Paroxysmal Events. 15-s EEG Epochs in a Bilateral Longitudinal Montage. (A) – EEG Demonstrating 1.5-2 Hz GPDs Without a Clearly Localizable Maximum (LFF: 1 Hz, HFF: 70 Hz, Sensitivity: 7 µV/mm). Note the Variable Morphology, Amplitude, and Frequency, Alternating Wave to Wave but not Following a Temporal Pattern of Change Along the Epoch. (B) – 6.5 Hz Rhythmic Discharges Involving the Bilateral Posterior Regions, More Prominent on the Right (LFF: 1 Hz, HFF: 70 Hz, Sensitivity: 7 µV/mm). Note that the Sterotyped Morphology and Frequency do not Change Along the Epoch, which is Consistent With an Artifactual Source of Waveform. (C) – Low Amplitude Continuous Generalized Slowing With Overlying Sweat Artifact and Slow Drifts due to Loose Electrode Contacts (LFF: 1 Hz, HFF: 70 Hz, Sensitivity: 5 µV/mm). In the Midline Region where These Artifacts are not Seen, Suppression of Background Activity can be Better Appreciated. On the Right Upper Corner, the Video Image Shows Increased Tonus in the Right Neck Resulting in Turning to This Side Along With the Left Shoulder Adduction. (D)1-2 Hz Generalized Rhythmic Delta Activity With Superimposed Fast Activity Mainly on the Downslope of Delta Waves (LFF: 1 Hz, HFF: 70 Hz, Sensitivity: 7 µV/mm)

This case highlights the critical distinction between epileptic motor seizures and non-epileptic movements, such as tremors, dystonia, or myoclonus, particularly in the ICU setting where the majority of the presumed motor seizures may actually be non-epileptic. Table 1 provides an overview of abnormal movements that may resemble epileptic motor seizures in critically ill patients. It is important to recognize these movements as they are often mistakenly treated as seizures. A 2010 study by Benbadis et al. demonstrated that only 27% of abnormal motor events were true seizures, with the remainder being tremor-like movements, myoclonus without electrographic changes, or other abnormal movements. ⁵ A response to benzodiazepines does not always indicate a seizure, as they can also alleviate various movement disorders. ⁶ Key distinguishing features, such as movement laterality while diffuse EEG abnormalities are seen, and lack of a clear electrographic seizure pattern can help differentiate non-epileptic from epileptic events.

Table 1.

Abnormal Movements Mimicking Epileptic Motor Seizures in Critically Ill Patients

Abnormal movement	Definition	Etiology
Chorea/Hemiballismus	Chorea: Involuntary, purposeless, non-rhythmic, non-sustained movements that flow from one body part to the other	Inflammatory/infectious; toxic-metabolic; vascular
	Hemiballismus: A severe form of chorea, is characterized by vigorous irregular high amplitude movements on one side of the body
Clonus	Rhythmic involuntary muscular contractions and relaxations	Vascular; post-anoxic injury
Dystonia	Sustained twisting movements that are often frequent and progresses to prolonged abnormal postures	Vascular; post-anoxic injury; autoimmune (NMDAR-AE); anesthetics; TBI; medications
Myoclonus	Sudden, brief involuntary movements which may be caused by muscle contractions (positive myoclonus)	Vascular; post-anoxic injury; medications; CJD; toxic-metabolic causes
	Asterixis is considered a negative myoclonus secondary to sudden loss of tone
Paroxysmal posturing	Involuntary flexor or extensor posturing on one side or bilateral spontaneously or with pain	Vascular; TBI; increased intracranial pressure (eg, brain tumors)
	Opisthotonus posturing refers to hyperextension of the neck and back “arching position”
Shivering	High frequency involuntary muscular contractions involving one group or more of muscles	Post-anoxic injury; hypothermia protocol; infections/sepsis
Tremor	Oscillatory rhythmic movement that affects one or more parts of the body	Toxic-metabolic causes; vascular; post-anoxic injury; medications

(CJD, creutzfeldt-jakob disease; NMDAR-AE: NMDA receptor autoimmune encephalitis; TBI, traumatic brain injury).

Nevertheless, the absence of an EEG correlate does not definitively exclude epileptic origin. Focal motor seizures arising from very small cortical areas (less than 10 cm²) or from deeper brain foci may generate electrical signals too weak to be consistently detected by scalp electrodes. ⁷ Given these limitations, comprehensive clinical evaluation—including a detailed history and careful observation of seizure semiology—remains essential for accurate diagnosis and management in such cases.

Case 2

A 21-year-old patient with drug-resistant epilepsy on multiple ASMs and neuromodulation therapies (vagus nerve stimulator [VNS] and deep brain stimulation [DBS]), presented with status epilepticus (SE) and was admitted to the ICU. Initially intubated and on a propofol drip, the cvEEG revealed brief runs of generalized electrographic seizures, which ceased after increasing the propofol infusion rate. However, the EEG also showed rhythmic activity predominantly in the posterior regions, with a right-greater-than-left distribution, occurring in a highly periodic pattern every 5 min (Figure 1B). The ICU team was concerned that this was breakthrough seizure activity. However, further EEG analysis suggested this rhythmic activity was likely not an electrographic seizure but rather an artifact from the patient’s neuromodulation devices, as the pattern was stereotyped and perfectly periodic. Additionally, the discharges were monomorphic, lacked a consistent spatial field, and demonstrated no temporal or spatial evolution. Moreover, their duration and intervals correlated with the on–off settings of the patient’s DBS.

This case highlights the importance of recognizing EEG artifacts caused by implantable devices such as VNS, responsive neurostimulation (RNS), and DBS. ⁸ In the ICU, where patients are frequently connected to multiple electronic devices, these artifacts can obscure true electrographic seizures and may produce patterns that closely mimic ictal activity. A summary of common artifact sources in the ICU is provided in Table 2.

Table 2.

Frequently Seen Physiologic and Non-physiologic Artifacts in the ICU Setting

Physiologic artifacts

• Sweat

• Chewing, teeth grinding

• Tremor and other movement artifacts

• Pulse artifact

• ECG artifact

Non-physiologic artifacts

• Mechanical ventilation (including water in tube)

• 60 Hz artifact (50 Hz in some countries)

• Intravenous drips

• Electrode artifact

• Implanted devices (eg, VNS, DBS, RNS, pacemaker)

• Suctioning

• Chest percussions

DBS, deep brain stimulator; ECG, electrocardiogram; RNS, responsive neurostimulator; VNS, vagal nerve stimulator.

Certain features may help distinguish artifacts from epileptic seizures. These include periodicity, highly stereotyped or monomorphic waveforms, and the absence of a physiological electric field. Artifact-related activity often fails to localize to physiologic brain regions and may exhibit atypical features, such as double phase reversals or activity isolated to a single electrode. When available, video monitoring can be diagnostic—especially when the artifact source (eg, chewing, toothbrushing, rhythmic patting, or chest percussion) is captured and shown to be time-locked with the suspicious EEG discharges.

Case 3

A 47-year-old patient with epilepsy due to a left frontal anaplastic oligodendroglioma status post resection presented to the ICU with SE. Video EEG showed recurrent episodes of staring, left upper extremity stiffening, and slow eye deviation with the head turning to the right. The EEG revealed moderate slowing of background activity, with severe suppression but no clear seizure pattern (Figure 1C). Despite intravenous lorazepam, these episodes did not resolve.

In ICU settings, distinguishing between tonic seizures and non-epileptic tonic events (NTE) can be challenging. NTEs, which are often caused by increased intracranial pressure or brainstem injury, typically lack an EEG correlate and are often accompanied by diffuse background slowing or suppression. ⁹ A recent study by Katyal and colleagues outlined differences between tonic seizures and NTEs, noting that NTEs often involve proximal muscle groups with internal rotation and adduction, rather than distal muscles or abduction. ¹⁰ In this case, the lack of an EEG correlate, the pattern of muscle involvement, and the patient’s brain tumor history raised suspicion for NTE.

NTEs in ICU patients can sometimes be associated with plateau waves caused by elevated intracranial pressure (ICP), often indicating poor prognosis, particularly in cases of aggressive brain tumors. ¹¹ In our case, the EEG showed severe diffuse slowing followed by generalized suppression. Similarly, a 2022 study by Sheikh et al. reported EEG changes presumed to reflect ICP fluctuations in seven patients, with patterns ranging from generalized theta/delta slowing to diffuse suppression. ¹² These cyclic patterns and varying degrees of attenuation often preceded clinical deterioration related to intracranial hypertension. Future systematic studies may help validate cvEEG as an early, noninvasive marker of elevated ICP.

Case 4

A 29-year-old patient with anti-N-methyl-d-aspartate (NMDA) receptor encephalitis and super-refractory SE was on cvEEG for several days while receiving multiple ASMs, including intravenous anesthetics. The EEG demonstrated continuous generalized rhythmic delta activity with superimposed fast activity in a stereotyped relationship to the delta waves (Figure 1D)—also known as extreme delta brush, which is seen in approximately one-third of patients with this condition.^13,14 The cvEEG also demonstrated runs of rhythmic 2-3 Hz generalized delta activity that evolved in morphology, frequency, and temporospatial distribution, lasting 40-110 s in duration. These had no clinical correlate and were interpreted as electrographic seizures, which resolved following escalation of ASM therapy. Subsequently, cvEEG revealed abnormal orofacial twitching movements, for which the team initially escalated treatment. However, these movements did not have an electrographic correlation.

NMDA receptor encephalitis is an autoimmune disorder characterized by IgG antibodies targeting the NR1 subunit of NMDA receptors. ¹⁵ The clinical course typically begins with a prodromal phase resembling a viral illness, followed by the onset of psychiatric symptoms such as delusions and psychosis. ¹⁵ During the acute phase, patients often develop movement disorders, including orofacial dyskinesias, dystonia, myoclonus, and tremors.^1,15 These abnormal movements can closely mimic epileptic seizures, making diagnosis and management particularly challenging. ¹⁶

This case underscores the difficulty in distinguishing between epileptic seizures and movement disorders in the context of NMDA receptor encephalitis. Given the typical etiology and symptomatology of the condition, careful evaluation of EEG findings, alongside comparison with prior seizure semiology and electrographic patterns, is essential for accurate diagnosis and to avoid unnecessary treatments. In this patient, the absence of an EEG correlate and the marked difference in semiology compared to previously documented seizures suggest that the observed episodes are more consistent with dyskinetic events rather than epileptic seizures.

Case 5

A 61-year-old female with relapsed high-risk myelodysplastic syndrome, status post stem cell transplantation, was hospitalized due to progression of disease and transformation to acute myeloid leukemia. During her hospitalization, the patient developed altered mental status. cvEEG showed generalized periodic discharges at approximately 2 Hz with a triphasic morphology (Figure 2A). Lorazepam was administered, which resolved the periodic discharges (Figure 2B), but the patient developed hypotension and hypoxia, requiring increased oxygen support. She continued to deteriorate, and 10 h later, the EEG demonstrated a severe degree of background suppression. Unfortunately, she subsequently succumbed. Of note, the initial EEG pattern did not meet the criteria for seizures, as the frequency was below 2.5 Hz and lacked temporal evolution. ¹⁴ The pattern was consistent with toxic metabolic encephalopathy, where triphasic GPDs can occur.

Figure 2.

Title. EEG Epochs of Case 5. (A) – 2 Hz GPDs With Triphasic Morphology (LFF: 1 Hz, HFF: 70 Hz, Sensitivity: 7 µV/mm) B – Polymorphic Background Activity after Lorazepam Administration (LFF: 1 Hz, HFF: 70 Hz, Sensitivity: 7 µV/mm; GPDs: Generalized Periodic Discharges)

This case underscores the principle of “do no harm,” highlighting that overly aggressive treatment may be unwarranted when EEG patterns do not meet established criteria for definite seizures. In this instance, the patient’s EEG findings were likely consistent with the ictal–interictal continuum (IIC)—for example, rhythmic or periodic EEG patterns concerning for ongoing electrographic seizures but not meeting formal electrographic criteria to be definitively characterized as ictal. ¹⁴ Although an EEG pattern within the IIC spectrum is potentially ictal, careful consideration of the clinical presentation—including neurological examination, laboratory studies, and medication review—is of utmost importance. ¹⁷

It is possible that our patient experienced further neurological decompensation and background suppression due to multiple factors, which were exacerbated by benzodiazepine use. In such cases, deferring sedating ASMs in an obtunded patient with a tenuous airway—when there is no evidence of definitive seizures—may be preferable, allowing focus on timely treatment of the underlying medical cause. The risks of empiric treatment, particularly the use of sedating medications in patients with impending respiratory compromise, should also be carefully weighed against potential benefits given the clinical context.^18,19 Notably, GPDs in the setting of toxic or metabolic encephalopathy have been associated with a reduced risk of neuronal damage. In these situations, avoiding overtreatment is especially prudent, as it may be not only unnecessary but potentially harmful. ¹⁸

Nonetheless, it must be recognized that overtreatment is sometimes identified only in retrospect. In this case, the patient presented with unexplained altered mental status and an IIC pattern on EEG characterized by 2-Hz GPDs with triphasic morphology. Empiric treatment was initiated; although the EEG findings improved, the patient’s clinical status did not, and benzodiazepine therapy may have produced adverse effects. The literature on IIC remains controversial. ²⁰ EEG interpreters show variable agreement in classifying GPDs as triphasic waves, and emerging data suggest that GPDs, regardless of triphasic morphology, may carry similar seizure risk.^6,21 Moreover, some patients with “triphasic waves” demonstrate both clinical and electrographic improvement with ASM therapy. ²¹ This has led some experts to shift focus away from triphasic morphology alone and to avoid conflating triphasic waves exclusively with metabolic encephalopathy.

Conclusions

cvEEG monitoring is a vital tool in the ICU, but interpretation is often challenged by artifacts and non-epileptic movements that can mimic seizures. The complex ICU environment increases the risk of misinterpreting EEG patterns, especially in cases involving tremors, dystonia/dyskinesa, or device-related artifacts. Accurate differentiation from true epileptic activity requires careful attention to clinical context, EEG features, and video correlation. IIC patterns must also be interpreted cautiously, as not all suggest seizures or carry a high risk of neuronal injury. A thoughtful, multidisciplinary approach is essential for accurate interpretation and optimal care of patients undergoing cvEEG monitoring. By nature, cvEEG is a collaborative effort, requiring close communication between the EEG interpreter and the clinical team to ensure accurate clinical correlation and interpretation.

ORCID iD

Anteneh M. Feyissa https://orcid.org/0000-0002-9318-3947