(219 days)
The Olympic Brainz Monitor (OBM) is a three channel electroencephalograph (EEG) acquisition system intended to be used in a hospital environment to record, collect, display and facilitate manual marking of aEEG recordings.
- The signals acquired from P3-P4, C3-P3 and C4-P4 channels are intended for use only with neonatal patients (defined as from birth to 28 days post-delivery, and corresponding to a postconceptual age of 24 to 46 weeks) to display aEEG for monitoring the state of the brain.
- The signals acquired from P3-P4 channel is intended to assist in the assessment of Hypoxic-Ischemic Encephalopathy severity and long-term outcome, in full term neonates (postconceptual age of 37-46 weeks) who have suffered a hypoxic-ischemic event.
- The RecogniZe seizure detection algorithm is intended to mark sections of EEG/aEEG that may correspond to electrographic seizures in only the centro-parietal regions of full term neonates (defined as from birth to 28 days post-delivery, and corresponding to a postconceptual age of 37 to 46 weeks). EEG recordings should be obtained from centro-parietal electrodes (located at P3, P4, C3 and C4 according to 10/20 system). The output of the Recognize algorithm is intended to assist in post hoc assessment of EEG/aEEG traces by qualified clinical practitioners, who will exercise professional judgment in using the information.
The Olympic Brainz Monitor does not provide any diagnostic conclusion about the patient's condition.
The Olympic Brainz Monitor is a three-channel electroencephalograph (EEG) system, as per 21 CFR §882.1400: a device used to measure and record the electrical activity of the patient's brain by placing two or more electrodes on the head. The device does not introduce, transfer or deliver any type of energy to the patient. As any other electroencephalograph the device passively record the electroencephalographic activity from the patient trough the hydrogel electrodes and then process the signal for display, analysis and archiving.
The Olympic Brainz Monitor system consists of the following:
- Data Acquisition Box (DAB)
- Touchscreen Monitor
- Roll Stand or optional Desktop Stand
- 9 Neonatal Sensor set (K033010)
- Software
These components have equivalent configuration and functions to those described in K093949 for the OBM Monitor. The Neonatal Sensor set (cleared on K033010) is an accessory to the device that is the only part that enters into contact with the patient. The sensor guarantees acquisition of the electroencephalographic signal and passively transfers it to the main unit. This is a set of five hydrogel skin electrodes that are attached to the patient's head at one extreme and to the Data Acquisition Box at the other extreme using standard touch-proof connectors.
The device allows practitioners to acquired, store, review and archive EEG activity from 4 centroparietal locations corresponding to C3, C4. P3 and P4 of the international 10-20 System. The device displays the recorded activity in form of the raw EEG and as amplitude integrated EEG (aEEG).
In addition the device now includes a seizure detection algorithm (i.e RecogniZe) to allow automated analysis of the recorded EEG. The RecogniZe Seizure Detection Algorithm identifies sections of the EEG trace where seizure activity is detected. The algorithm comprises filtering of the EEG signal, fragmentation of EEG signal into waves, wave-feature extraction, and elementary, preliminary and final detection. The main idea behind the algorithm is to detect heightened regularity in EEG wave sequences using wave intervals, amplitudes and shapes, as increased regularity is the major distinguishing feature of seizure discharges.
Here's a breakdown of the acceptance criteria and the study that proves the device meets them, based on the provided text:
Acceptance Criteria and Device Performance for RecogniZe Seizure Detection Algorithm
1. Table of Acceptance Criteria and Reported Device Performance
| Metric | Acceptance Criteria (Predicate Devices - IndenEvent K092039, Persyst Reveal K011397) | Reported Device Performance (RecogniZe K123079) |
|---|---|---|
| Positive Percent Agreement (PPA) | 74% - 79.5% (observed in predicate devices) | 61% (95% CI: 52 - 68) |
| False Detection Rate (FDR) | 0.08 - 0.3 FP/h (observed in predicate devices) | 0.5 FP/h (95% CI: 0.4 - 0.7) |
Note: The document argues that RecogniZe is "substantially equivalent to the performance of medical experts confronted with similar task and amount of data" and therefore substantially equivalent to the predicate devices, despite the numerical differences compared to the predicate devices themselves. The comparison is made against expert inter-rater agreement for the specific limited-channel montage.
2. Sample Size Used for the Test Set and Data Provenance
- Number of Events: 421 seizure events
- Total Number of Patients: 82
- Number of Hours (of EEG recordings): 621 hours
- Data Provenance: Retrospective clinical evaluation from neonatal patients seen for routine clinical evaluation at the Neonatal Intensive Care Unit of St. Louis Children's Hospital, USA.
The study included recordings from full term neonates (post-conceptual age of 37 to 46 weeks, defined as from birth to 28 days post-delivery).
To avoid over-weighting, a maximum of 13 events per limited-channel recording were permitted.
3. Number of Experts Used to Establish Ground Truth for the Test Set and Their Qualifications
- Number of Experts: 3
- Qualifications of Experts: Board certified neurophysiologists.
4. Adjudication Method for the Test Set
The document describes how the ground truth was established by experts:
- Experts independently, blindly, and manually marked seizures (no seizure detection algorithm was allowed) in the same manner they would normally do in clinical practice.
- Initially, experts reviewed the full cohort of standard montage recordings (157 of them) marking seizure onset and topography.
- After a 4-week wash-out period, the same reviewers were provided with the limited-channel (C3-P3, C4-P4, and P3-P4) recordings for marking.
- The ground truth used for comparison with the algorithm was the outcome of the expert review on these limited-channel recordings.
- For the inter-rater agreement of experts themselves, individual expert markings were compared against each other (e.g., Rater 1 vs Rater 2, Rater 1 vs Rater 3, Rater 2 vs Rater 3). It isn't explicitly stated if a consensus (e.g., 2+1, 3+1) was used to define the final "gold standard" truth for the algorithm comparison, but rather "the gold standard, defined as seizures detected by a panel of 3 EEG board certified medical professionals" was used. The reported PPA and FDR for the algorithm are compared to the average inter-rater agreement of these experts on the limited-channel montage.
5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done, and Effect Size of Human Improvement with AI vs. Without AI Assistance
A classic MRMC comparative effectiveness study, directly comparing human reader performance with AI assistance versus without AI assistance, was not explicitly described for the RecogniZe module.
The study compared the standalone performance of the RecogniZe algorithm against the performance of human experts (who were themselves establishing the ground truth) on the limited-channel montage. It also reported inter-rater agreement among the human experts.
The document states: "RecogniZe is intended as a tool to aid in the assessment of long EEG recordings to help reduced the amount of time devoted to review." However, it does not quantify this reduction or demonstrate increased accuracy of humans when using the AI.
6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) Was Done
Yes, a standalone performance study was done for the RecogniZe algorithm. The algorithm's PPA and FDR were calculated by comparing its output directly against the "gold standard" established by the panel of 3 experts on the limited-channel EEG recordings.
7. The Type of Ground Truth Used
The ground truth used was expert consensus / expert marking. Specifically, it was defined as "seizures detected by a panel of 3 EEG board certified medical professionals" who independently marked seizures on de-identified and randomized EEG recordings.
8. The Sample Size for the Training Set
The document does not report the sample size used for the training set for the RecogniZe algorithm. It only details the "Testing Dataset."
9. How the Ground Truth for the Training Set Was Established
The document does not describe how the ground truth for the training set was established, as it does not provide details on the training set itself. The information provided pertains solely to the clinical validation (testing) dataset.
§ 882.1400 Electroencephalograph.
(a)
Identification. An electroencephalograph is a device used to measure and record the electrical activity of the patient's brain obtained by placing two or more electrodes on the head.(b)
Classification. Class II (performance standards).