REMI AI Discrete Detection Module

The REMI-AI Discrete Detection Module is an assistive software tool that analyzes 4-channel REMI EEG recordings to automatically mark potential electrographic seizures lasting at least 10 seconds in adult and pediatric patients (6+ years), aiding physician review. Its scientific basis lies in a machine learning algorithm trained and validated against expert epileptologist consensus on concurrently acquired EEG data, demonstrating specific performance in terms of sensitivity and false alarm rate for seizure event identification.

The REMI-AI Discrete Detection Module (REMI-AI DDM) is a software as a medical device (SaMD) intended for the analysis of electroencephalogram (EEG) recordings from the REMI Remote EEG Monitoring System. Its purpose is to assist qualified physicians by automatically identifying and annotating sections of previously acquired 4-channel REMI EEG data that may correspond to potential electrographic seizures lasting at least 10 seconds. The system analyzes recordings from bilateral, bipolar scalp electrodes at frontal and temporoparietal regions and is indicated for use in adult and pediatric patients (6 years and older). Importantly, REMI-AI DDM does not operate in real-time and is not intended to provide diagnostic conclusions, but rather to serve as an aid to a physician's professional judgment during EEG review. The REMI-AI DDM's algorithm was trained and validated using EEG data where a reference standard was established through a consensus review by a panel of three independent expert epileptologists evaluating concurrently acquired standard 19+channel wired EEG. The clinical validation involved a dataset of 50 patient records, including 31 with seizures and 19 without. Performance testing demonstrated that the REMI-AI DDM achieved an event-level sensitivity with a 95% confidence interval lower bound of 79.5% and a False Alarm Rate (FAR) with a 95% confidence interval upper bound of 0.221 False Positives (FP)/hr. Across all patients with seizures, the event-level sensitivity was 86.2%, and the mean per-patient FAR across all 50 patients was 0.162 FP/hr. The device includes an authorized Predetermined Change Control Plan (PCCP) for future algorithm performance improvements.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of the 4-channel REMI-AI DDM compare to seizure detection algorithms using full-montage (e.g., 19+ channel) EEG, particularly for subtle or focal seizures not well represented by the 4-channel configuration?; What specific electrographic features or patterns does the REMI-AI DDM algorithm prioritize for identifying potential seizures, and how does its performance vary across different seizure types (focal, focal evolving to generalized, generalized) and durations?; Considering the False Alarm Rate (FAR) of 0.162 FP/hr, what is the projected impact on clinician workflow efficiency and review burden, especially in continuous long-term EEG monitoring settings?; How does the Predetermined Change Control Plan (PCCP) specifically address the potential for dataset shift or drift as new patient data is incorporated, and what are the re-validation triggers for algorithm modifications?; What are the documented limitations of the REMI-AI DDM in detecting seizures in the presence of common EEG artifacts or for patient populations with characteristics underrepresented in the validation dataset (e.g., specific age subgroups within pediatrics, rare epileptic syndromes)?