Monitoring the Burden of Seizures and Highly Epileptiform Patterns in Critical Care with a Novel Machine Learning Method

Abstract

Introduction: Current electroencephalography (EEG) practice relies on interpretation by expert neurologists, which introduces diagnostic and therapeutic delays that can impact patients' clinical outcomes. As EEG practice expands, these experts are becoming increasingly limited resources. A highly sensitive and specific automated seizure detection system would streamline practice and expedite appropriate management for patients with possible nonconvulsive seizures. We aimed to test the performance of a recently FDA-cleared machine learning method (Claritγ, Ceribell Inc.) that measures the burden of seizure activity in real time and generates bedside alerts for possible status epilepticus (SE).

Methods: We retrospectively identified adult patients (n = 353) who underwent evaluation of possible seizures with Rapid Response EEG system (Rapid-EEG, Ceribell Inc.). Automated detection of seizure activity and seizure burden throughout a recording (calculated as the percentage of ten-second epochs with seizure activity in any 5-min EEG segment) was performed with Claritγ, and various thresholds of seizure burden were tested (≥ 10% indicating ≥ 30 s of seizure activity in the last 5 min, ≥ 50% indicating ≥ 2.5 min of seizure activity, and ≥ 90% indicating ≥ 4.5 min of seizure activity and triggering a SE alert). The sensitivity and specificity of Claritγ's real-time seizure burden measurements and SE alerts were compared to the majority consensus of at least two expert neurologists.

Results: Majority consensus of neurologists labeled the 353 EEGs as normal or slow activity (n = 249), highly epileptiform patterns (HEP, n = 87), or seizures [n = 17, nine longer than 5 min (e.g., SE), and eight shorter than 5 min]. The algorithm generated a SE alert (≥ 90% seizure burden) with 100% sensitivity and 93% specificity. The sensitivity and specificity of various thresholds for seizure burden during EEG recordings for detecting patients with seizures were 100% and 82% for ≥ 50% seizure burden and 88% and 60% for ≥ 10% seizure burden. Of the 179 EEG recordings in which the algorithm detected no seizures, seizures were identified by the expert reviewers in only two cases, indicating a negative predictive value of 99%.

Discussion: Claritγ detected SE events with high sensitivity and specificity, and it demonstrated a high negative predictive value for distinguishing nonepileptiform activity from seizure and highly epileptiform activity.

Conclusions: Ruling out seizures accurately in a large proportion of cases can help prevent unnecessary or aggressive over-treatment in critical care settings, where empiric treatment with antiseizure medications is currently prevalent. Claritγ's high sensitivity for SE and high negative predictive value for cases without epileptiform activity make it a useful tool for triaging treatment and the need for urgent neurological consultation.

Keywords: Electroencephalography; Machine learning method; Neurology; Seizure burden; Status epilepticus.

Fig. 1

Rapid Response EEG system. The Rapid Response EEG system (Rapid-EEG) consists of a portable EEG recorder and a disposable electrode headband. Recorded EEG tracings are shown on the device screen (1) and sonified when needed (2) by the bedside recorder. HIPAA-compliant secure Wi-Fi connection enables real-time transfer of the data to the cloud where the EEG tracings can be reviewed by expert neurologists using the remote portal for EEG review (3). Machine learning computations (by Claritγ algorithm) are performed on the cloud portal (4) interfacing in real time with the bedside device. As such, the system is meant to provide not only easy and fast access to EEG acquisition, but also a reliable and actionable diagnostic information for risk stratification using four different modes of triage

Fig. 2

Computation of seizure burden. The output of the Claritγ algorithm was a continuous quantitative trend of seizure burden values, which represented the percentage of 10-second long bins of EEG data in a 5-min period that contained seizure activity. Seizure burden values updated every 10 s; therefore, consecutive seizure burden values (e.g., value 1 and 2, as shown, offset by 10 s) could represent the evolution of the patient’s seizure prevalence over the course of the recording. Seizure burden thresholds were adapted from American Clinical Neurophysiology Society guidelines [22], such that “frequent” seizure activity was defined as 10% seizure burden (i.e., 30 s of seizure activity within a 5-min period), “abundant” seizure activity was defined as 50% seizure burden (i.e., 2.5 min of seizure activity within a 5-min period), and “continuous” seizure activity was defined as 90% seizure burden (i.e., 4.5 min of seizure activity within a 5-min period). An alert was presented to the user when seizure burden reached a threshold of 90%, which indicated a high risk of status epilepticus and the impending need for urgent clinical intervention

Fig. 3

Samples of EEG recorded with Ceribell Rapid Response EEG System. Each EEG is displayed in a ten-second epoch with filter settings of 1–30 Hz. The line plot under each EEG shows the Claritγ algorithm output. The top image shows seizure activity approaching the 90% threshold to trigger a status epilepticus alert, and the bottom image shows lateralized periodic discharges that go undetected by the algorithm

Fig. 4

Summary of Claritγ Performance. Performance of Claritγ algorithm at the group level suggests that the algorithm can be seen as a reliable triage tool to help detect cases of status epilepticus with the highest sensitivity (while overcalling about one-fourth of highly epileptiform patterns as possible status epilepticus). It also performs as a reliable triage tool to help physicians avoid over-aggressive treatments in majority of EEG cases where the overwhelming pattern is either slowing or normal. HEP highly epileptiform patterns, NL normal activity, RDA rhythmic delta activity, SE status epilepticus, SL slow activity, SZ seizure