Search Results
Found 1 results
510(k) Data Aggregation
(189 days)
Nelli is indicated for use as an adjunct to seizure monitoring of patients aged 18 years and above in healthcare facilities during periods of rest. The device utilizes automated analysis of audio and video (media) to identify epileptic and non-epileptic seizure events with a positive motor component. The system provides prioritization of identified major seizures and other motor events for review and media playback by a qualified healthcare professional.
Nelli is a seizure detection and monitoring system based upon the analysis of audio/video recording which does not require the patient to be fitted with any body-worn accessory which may interfere with sleep and/or rest. Nelli uses machine learning algorithms to detect and categorize into review priority classes events indicative of seizure activity with a positive motor component. The identified events can be accessed and viewed in the interactive report as a scattergram through the web-based interface.
Here is a summary of the acceptance criteria and study information for the Nelli Seizure Monitoring System, based on the provided FDA 510(k) Clearance Letter:
1. Table of Acceptance Criteria and Reported Device Performance
| Seizure Category | Performance Metric | Acceptance Criteria | Reported Device Performance |
|---|---|---|---|
| I: High Priority (Convulsive) (Tonic-Clonic, Bilateral Clonic) | Sensitivity (Positive Percent Agreement) | Lower bound of the 95% CI for Sensitivity > 70% | 84.5% [74.6%, 94.2%] |
| False Detection Rate (FDR) (False Positives per hour) | Upper bound of the 95% CI for FDR < 0.14 FP/hour | 0.050 FP/hour [0.0376, 0.0636] | |
| II: Medium Priority (Other Major Motor) (Tonic, Unilateral Clonic, Hyperkinetic (>10s), Other Motor (>30s)) | Sensitivity (Positive Percent Agreement) | Lower bound of the 95% CI for Sensitivity > 70% | 84.3% [75.2%, 91.9%] |
| False Detection Rate (FDR) (False Positives per hour) | Upper bound of the 95% CI for FDR < 8 FP/hour | 6.08 FP/hour [5.63, 6.54] |
Overall Success Criterion: Fulfillment of both primary objectives (High-Priority Convulsive seizure sensitivity and FDR). The reported results indicate that Nelli met all specified performance goals for both primary and secondary objectives.
2. Sample Size Used for the Test Set and Data Provenance
- Test Set Sample Size: 172 completed cases (subjects).
- Data Provenance: The clinical investigation was conducted in three sites located within the United States. The study is described as a "single-arm clinical investigation with performance goals," implying a prospective data collection for validation. Data combines two datasets from Phase I and Phase II.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts
The document states that the performance was measured "against an expert panel's review of video EEG." However, it does not specify the number of experts on this panel or their specific qualifications (e.g., "Radiologist with 10 years of experience"). It only mentions "expert panel."
4. Adjudication Method for the Test Set
The document does not explicitly describe an adjudication method like 2+1 or 3+1. It mentions "an expert panel's review of video EEG" as the gold standard for ground truth. This suggests that the panel collectively established the ground truth, but the specific process for consensus or conflict resolution among multiple experts is not detailed.
5. If a Multi Reader Multi Case (MRMC) Comparative Effectiveness Study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance
No, a Multi-Reader Multi-Case (MRMC) comparative effectiveness study comparing human readers with AI assistance versus without AI assistance was not explicitly mentioned in the provided document. The study design described is a single-arm clinical investigation evaluating the standalone performance of the Nelli system against a video EEG ground truth. Nelli is indicated as an adjunct to seizure monitoring, providing prioritization for review by a qualified healthcare professional, but the study did not measure the improvement of human readers using the system.
6. If a Standalone (i.e. algorithm only without human-in-the-loop performance) was done
Yes, a standalone performance study was conducted. The clinical investigation directly evaluated Nelli's ability to "function as an assessment aid in the monitoring of seizure-related activity" by measuring its event-level Sensitivity and False Detection Rate against video EEG. The system automatically analyzes audio and video to identify events and categorize them for review; the reported metrics reflect this automated detection capability.
7. The Type of Ground Truth Used
The ground truth used for the clinical investigation was "an expert panel's review of video EEG (the current gold standard for seizure monitoring)."
8. The Sample Size for the Training Set
The document mentions "Training of detection algorithms using Artificial Intelligence (AI)," but it does not specify the sample size used for the training set. It only refers to documentation per GMLP (Good Machine Learning Practice).
9. How the Ground Truth for the Training Set Was Established
The document does not detail how the ground truth for the training set was established. It only states, "Training of detection algorithms using Artificial Intelligence (AI) (documentation per GMLP²)." It is reasonable to infer that the training ground truth would also be based on expert review of video EEG data, given its status as the "current gold standard" for the test set, but this is not explicitly stated for the training data.
Ask a specific question about this device
Page 1 of 1