Encevis

• K Number: K211452
• Device Name: Encevis
• Manufacturer: Austrian Institute of Technology GmbH
• Date Cleared: 2021-12-02 ( 206 days)
• Product Code: OMB , OLT , OMA
• Regulation Number: 882.1400
• Type: Traditional
• Panel: NE
• Reference & Predicate Devices: K132306 , K171720

Intended Use

1. encevis is intended for the review, monitoring and analysis of EEG recordings made by electroencephalogram (EEG) devices using scalp electrodes and to aid neurologists in the assessment of EG. This device is intended to be used by qualified medical practitioners who will exercise professional judgment in using the information.

2. The seizure detection component of encevis is intended to mark previously acquired sections of adult (greater than or equal to 18 years) EEG recordings that may correspond to electrographic seizures, in order to assist qualified clinical practitioners in the assessment of EEG traces. EEG recordings should be obtained with a full scalp montage according to the standard 10/20-system.

3. The spike detection component of encevis is intended to mark previously acquired sections of the patients EEG recordings that may correspond to spikes, in order to assist qualified clinical practitioners in the assessment of EEG traces. The Spike Detection component is intended to be used in adult patients greater than or equal to 18 years. encevis Spike Detection performance has not been assessed for intracranial recordings.

4. encevis includes the calculation and display of a set of quantitative measures intended to monitor and analyze the EEG waveform. These include frequency bands, rhythmic and periodic patterns and burst suppression. These quantitative EEG measures should always be interpreted in conjunction with review of the original EEG waveforms.

5. The aEEG functionality included in encevis is intended to monitor the state of the brain.

6. encevis provides notifications on an on-screen display for seizure detection, quantitative EEG and aEEG that can be used when processing a record during acquisition. Delays of up to several minutes can occur between the beginning of a seizure, the occurrence of a spike or detection of quantitative EEG features and when the encevis notifications will be shown to a user. encevis notifications cannot be used as a substitute for real time monitoring of the underlying EEG by a trained expert.

7. encevis PureEEG (Artifact Reduction) is intended to reduce EMG and electrode artifacts in a standard 10-20 EEG recording. PureEEG does not remove the entifact signal, and is not effective for other types of artifacts. PureEEG may modify portions of waveforms representing cerebral activity. Waveforms must still be read by a qualified medical practitioner trained in recognizing artifact, and any interpretation or diagnosis must be made with reference to the original waveforms.

8. This device does not provide any diagnostic conclusion about the patient's condition to the user.

Device Description

encevis combines several modalities for viewing and analyzing EEG data in one integrated software package. Encevis consists of the following modalities: encevis EEG-Viewer, encevis artifact reduction (PureEEG), encevis seizure detection, encevis spike detection (EpiSpike), encevis rhythmic and periodic patterns, encevis aEEG, encevis frequency bands, and encevis Burst Suppression.

AI/ML Overview

Here's a detailed breakdown of the acceptance criteria and study that proves the device meets them, based on the provided text:

Acceptance Criteria and Device Performance

For Seizure Detection:

Criteria Type	Acceptance Criteria (Non-inferiority margins)	Reported Device Performance (encevis vs. Predicate Persyst)
Positive Percentage Agreement (PPA)	PPA non-inferior to Predicate (margin 10%)	encevis PPA: 86.52% vs. Persyst PPA: 75.94% (non-inferior)
Negative Disagreement Rate (NDR)	NDR non-inferior to Predicate (margin 4 false detections/24h)	encevis NDR: 11.2 false detections/24h vs. Persyst NDR: 10.61 false detections/24h (non-inferior), lower is better

For Spike Detection:

Criteria Type	Acceptance Criteria (Non-inferiority margins)	Reported Device Performance (encevis vs. Predicate Persyst)
Positive Percentage Agreement (PPA)	PPA non-inferior to Predicate (margin 3%)	encevis PPA: 79.15% vs. Persyst PPA: 8.7% (non-inferior)
Negative Percentage Agreement (NPA)	NPA non-inferior to Predicate (margin 3%)	encevis NPA: 97.64% vs. Persyst NPA: 99.69% (non-inferior)
Positive Localization Percentage Agreement (PLPA)	PLPA non-inferior to Predicate (margin 3%)	encevis PLPA: 93.39% vs. Persyst PLPA: 93.97% (non-inferior)

For Artifact Reduction (PureEEG):

Criteria Type	Acceptance Criteria (Non-inferiority margins)	Reported Device Performance (encevis vs. Predicate Persyst)
Relative Suppression of Clean EEG (dB)	Non-inferior to Predicate (margin 1 dB)	encevis suppression lower than Persyst in 73/127 cases; 95% delta CI=[-0.07, -0.02], margin=0.01 (non-inferior)
Signal-to-Noise Ratio (SNR) after artifact removal (dB)	Non-inferior to Predicate (margin 1 dB)	encevis SNR higher than Persyst in 83/93 cases; 95% delta CI=[4.37,5.88], margin=0.01 (non-inferior)

For Rhythmic and Periodic Patterns (NeuroTrend):

Pattern Type	Sensitivity (%)	Specificity (%)
ANY	81.86 (79.9-83.8)	83.80 (83.1-84.5)
PD	69.73 (67.2-72.3)	95.89 (95.5-96.3)
ARA (RTA, RAA, RDA+S)	89.40 (84.2-94.6)	94.85 (94.5-95.3)
RDA	91.73 (86.4-97.1)	86.05 (85.4-86.7)
Localization (ACNS Main Term 1)	Cohen's Kappa: 0.51 (Moderate agreement) compared to expert consensus

For Burst Suppression:

Performance Metric	Value (%)
Sensitivity (SE)	87 (84.7-89.9)
Specificity (SP)	92 (91.4-92.9)
Positive Predictive Value (PPV)	61 (57.9-64.3)
Negative Predictive Value (NPV)	98 (97.7-98.5)

For aEEG (Amplitude-integrated EEG) and Frequency Bands:

These features were validated by comparing their output to the proposed method of Zhang and Ding (2013) and to the predicate device Persyst. The frequency response of aEEG was checked, and the outputs of both aEEG and frequency bands were found to be in good accordance with the reference and predicate, respectively. Specific numerical acceptance criteria were not explicitly stated beyond "good accordance" and "correct assignment," but the validation aimed to demonstrate consistency and accuracy.

---

Study Details

2. Sample sizes and data provenance for the test sets:

• Seizure Detection:
  • Sample Size: 55 subjects (scalp-EEG recordings). 50 patients with epilepsy, 5 subjects without epilepsy.
  • Data Provenance: Retrospective, from video-EEG monitoring in an epilepsy monitoring unit. Country of origin not specified, but the text mentions "different centers" for artifact reduction, rhythmic patterns, and burst suppression, implying multi-center data collection.
  • EEG Hours: A total of 1619 hours of EEG data were used. A maximum of 30 hours of continuous EEG data from each subject. For subjects without epilepsy, the first available 30 hours were included.
• Spike Detection:
  • Sample Size: 23 patients (scalp-EEG recordings). 18 subjects with spike events, 5 subjects without epilepsy.
  • Data Provenance: Retrospective, from video-EEG monitoring in an epilepsy monitoring unit. Country of origin not specified.
• Artifact Reduction (PureEEG):
  • Sample Size: 128 EEG data records from different patient groups (60 from epilepsy monitoring units, 65 from ICU patients).
  • Data Provenance: Retrospective. Includes 31 EEG segments from 31 subjects for seizure EEGs, 33 EEG segments from 6 subjects for spikes, and 65 EEG segments from 65 subjects for ICU patients. Country of origin not specified, but the text mentions "different centers".
• Rhythmic and Periodic Patterns (NeuroTrend):
  • Sample Size: 83 long-term EEGs from ICU patients, resulting in 11935 common annotation segments (each minute split into three 20-second segments).
  • Data Provenance: Prospective, from two different centers. Country of origin not specified.
• Burst Suppression:
  • Sample Size: 83 long-term EEGs from intensive care patients, resulting in 3978 valid annotation segments.
  • Data Provenance: Retrospective, from two different centers. Country of origin not specified.

3. Number of experts and qualifications for ground truth:

• Seizure Detection: 3 independent neurologists for blinded review.
• Spike Detection: 3 independent neurologists for blinded review.
• Artifact Reduction (PureEEG): 3 independent epileptologists or neurologists for blinded review.
• Rhythmic and Periodic Patterns (NeuroTrend): 2 clinical neurophysiologists, who were naive to these EEGs.
• Burst Suppression: 2 clinical EEG experts (reviewers).

Specific years of experience or detailed qualifications beyond "neurologist," "epileptologist," and "clinical neurophysiologist" were not provided in the text.

4. Adjudication method for the test set:

• Seizure Detection: An event was considered "true seizure" if the time interval of two out of three reviewers overlapped by at least 1 second. The seizure epoch was then defined as the overlapping time range of these two reviewers.
• Spike Detection: An event was considered as "true spike" only if the time interval of two out of three reviewers overlapped. (Localization information was based on the 3D-coordinates of the electrode next to the spike maximum, averaged over reviewers).
• Artifact Reduction (PureEEG): The text mentions that three EEG experts were engaged to identify "clean, pure EEG patterns and artifacts of different types," implying a consensus-based approach, but a specific adjudication rule (e.g., 2+1) is not explicitly detailed for this section.
• Rhythmic and Periodic Patterns (NeuroTrend): Annotations had to be consistent between both reviewers to be used in sensitivity and specificity measurement.
• Burst Suppression: Detection performance was analyzed for consensus annotations of the two reviewers, meaning only annotation segments where both reviewers showed the same decision about Burst Suppression pattern were used.

5. Multi-reader multi-case (MRMC) comparative effectiveness study:

• No, a full MRMC comparative effectiveness study was not explicitly described as per the traditional definition (i.e., measuring how human readers improve with AI assistance vs. without AI assistance).
• The studies primarily focused on standalone algorithm performance compared to a human-established ground truth and, in many cases, compared to a predicate device (Persyst).
• For Rhythmic and Periodic Patterns and Burst Suppression, the device's performance (sensitivity, specificity) was measured against an expert consensus, implying the device is intended to assist, but there's no direct evaluation of human performance with vs. without the device.
• Therefore, no effect size for human readers improving with AI vs. without AI assistance was reported.

6. Standalone (algorithm only without human-in-the-loop performance) study:

• Yes, standalone performance was done for all described features.
• For Seizure Detection, Spike Detection, and Artifact Reduction, the device's detection results were compared to the expert-established ground truth.
• For Rhythmic and Periodic Patterns and Burst Suppression, sensitivity and specificity were calculated based on the algorithm's detections against the expert consensus.
• For aEEG and Frequency Bands, the algorithm's output was compared to theoretical models and a predicate device.

7. Type of ground truth used:

• Expert Consensus:
  • Seizure Detection: Consensus annotations from three independent neurologists.
  • Spike Detection: Blinded review sessions from three neurologists.
  • Artifact Reduction (PureEEG): Annotations of clean EEG recordings and artifacts by three independent epileptologists or neurologists.
  • Rhythmic and Periodic Patterns (NeuroTrend): Annotations from two clinical neurophysiologists, requiring consistency between them.
  • Burst Suppression: Annotations from two clinical EEG experts, requiring consensus.
• Theoretical/Predicate Comparison:
  • aEEG: Compared to the proposed method of Zhang and Ding (2013) and compared to Persyst.
  • Frequency Bands: Tested against sinusoidal test data and manually selected representative EEG recordings from epilepsy/ICU patients, then compared to Persyst.

8. The sample size for the training set:

• The document does not explicitly state the sample size for the training set for any of the features. The provided information solely pertains to verification and validation activities, which typically use separate, unseen datasets for testing.

9. How the ground truth for the training set was established:

• Since the training set size and details are not provided, the method for establishing ground truth for any potential training data is also not described in the submission. The document focuses on the test data ground truth establishment.

Regulation Number and Section

§ 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).