(289 days)
The Cerebra Sleep System is an integrated diagnostic platform that acquires, transmits, analyzes, and displays physiological signals from adult patients, and then provides for scoring (automatic and manual), editing, and generating reports. The system uses polysomnography (PSG) to record the electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), electromyogram (EMG), accelerometry, acoustic signals, nasal airflow, thoracic and abdomen respiratory effort, pulse rate, and oxyhemoglobin saturation, depending on the sleep study configuration. The Cerebra Sleep System is for prescription use in a home or healthcare facility.
The Cerebra Sleep System is intended to be used as a support tool by physicians and PSG technologists to aid in the evaluation and diagnosis of sleep disorders. It is intended to provide sleep-related information that is interpreted by a qualified physician to render findings and/or diagnosis, but it does not directly generate a diagnosis.
The Cerebra Sleep System is an integrated diagnostic platform that acquires, transmits, analyzes, and displays physiological signals from adult patients, and then provides for scoring (automatic and manual), editing, and generating reports. It is for prescription use in a home or healthcare facility and is used by physicians and polysomnographic (PSG) technologists as a support tool to aid in the evaluation and diagnosis of sleep disorders. A PSG tecnologist must edit, score, and review the data before sleep reports are generated.
The Cerebra Sleep System is capable of collecting data required for Level 2 PSG and Level 3 HSAT studies. A Level 3 HSAT study is a home sleep apnea test with a minimum of 4 channels that include oxygen saturation, electrocardiogram (ECG) or heart rate, airflow (e.g., nasal flow), and respiratory effort (e.g., chest band). A Level 2 PSG study is an unattended sleep test with a minimum of 7 channels that include electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG) or heart rate, chin electromyogram (EMG), and all signals from Level 3.
The Cerebra Sleep System (CSS) is comprised of three main areas:
- Prodigy: This is a PSG recorder capable of performing Level 2 and Level 3 sleep studies. It includes the Head Mounted Unit (HMU), which is worn on the patient's head, the Chest Mounted Unit (CMU), which is affixed to a chest effort belt, the Table Top Unit (TTU), which receives data wirelessly, and third party accessories including an oximeter.
- Cerebra Analytics Suite (CAS): The CAS has 4 components Web Processing (for signal processing of data), Web Scoring (for generating scoring results, which encompasses autoscoring), Cerebra Viewer (for viewing and editing PSG studies and scoring results) and Web Reporting (for generating reports). A PSG technician must edit, score, and review the data before reports are generated. These components utilize well-defined file formats to enable communication; communication with each component is done through the internet, via the Cerebra Portal.
- Cerebra Portal: All areas of the CSS product are managed by the Portal software. The Portal is used to configure a study on the TTU and allows sleep analysis service providers to manage inventory, patients, and sleep studies. It also configures Prodigy system hardware for an individual sleep study to be performed by a patient.
The Cerebra Sleep System includes an "autoscoring" module, the performance of which was evaluated against acceptance criteria.
1. Acceptance Criteria and Reported Device Performance
| Scoring Function | Acceptance Criteria (Overall % Agreement) | Reported Device Performance (Overall % Agreement) | Acceptance Criteria (Kappa) | Reported Device Performance (Kappa) |
|---|---|---|---|---|
| SLEEP STAGING | Not explicitly stated, but compared to Michele and Alice 5. Michele: 82.6%, Alice 5: 30.5% | 79.9% | Not explicitly stated, but compared to Michele and Alice 5. Michele: 0.77, Alice 5: 0.06 | 0.72 |
| Awake | Not explicitly stated | 82.32% APPA / 97.15% ANPA | Not explicitly stated | Not applicable |
| N1 | Not explicitly stated | 56.80% APPA / 91.64% ANPA | Not explicitly stated | Not applicable |
| N2 | Not explicitly stated | 82.95% APPA / 88.05% ANPA | Not explicitly stated | Not applicable |
| N3 | Not explicitly stated | 84.51% APPA / 97.16% ANPA | Not explicitly stated | Not applicable |
| REM | Not explicitly stated | 79.41% APPA / 98.84% ANPA | Not explicitly stated | Not applicable |
| AROUSALS | Not explicitly stated, but compared to Michele and Alice 5. Michele: 89.9%, Alice 5: 57.9% | 86.36% | Not explicitly stated, but compared to Michele and Alice 5. Michele: 0.54, Alice 5: 0.10 | 0.48 |
| Yes | Not explicitly stated | 64.79% APPA / 89.74% ANPA | Not explicitly stated | Not applicable |
| PLMs | Not explicitly stated, but compared to Michele and Alice 5. Michele: 95.7%, Alice 5: 88.3% | 95.99% | Not explicitly stated, but compared to Michele and Alice 5. Michele: 0.69, Alice 5: 0.38 | 0.69 |
| Yes | Not explicitly stated | 64.96% APPA / 98.59% ANPA | Not explicitly stated | Not applicable |
| RESPIRATORY EVENTS | Not explicitly stated, but compared to Michele and Alice 5. Michele: 94.0%, Alice 5: 78.0% | 89.18% | Not explicitly stated, but compared to Michele and Alice 5. Michele: 0.74, Alice 5: 0.25 | 0.54 |
| Overall Apneas | Not explicitly stated | 62.60% APPA / 93.03% ANPA | Not explicitly stated | Not applicable |
| Hypopneas & RERAs | Not explicitly stated | 66.52% APPA / 98.65% ANPA | Not explicitly stated | Not applicable |
Note: While specific numerical acceptance criteria for the Cerebra Sleep System's autoscoring performance are not explicitly stated as strict cut-offs, the comparison tables (Table 3 and Table 5) strongly imply that the acceptance criteria were based on achieving "moderate to substantial agreement" with manual scoring and demonstrating performance similar to or better than the identified predicate devices (Michele and Alice 5 autoscoring software systems).
2. Sample Size and Data Provenance
- Test Set Sample Size: 138 randomly selected pre-existing sleep studies.
- Data Provenance: The studies were from both sleep laboratory (Level 1 sleep test) and in-home (Level 2 or 3 sleep test) use environments. The document does not specify the country of origin, but given the FDA submission, it's likely to encompass data relevant to US regulatory standards. The data was "pre-existing," indicating a retrospective study design.
3. Number of Experts and Qualifications
- Number of Experts: Three.
- Qualifications of Experts: Board-certified registered polysomnographic technologists (RPSGT) with a minimum of 13 years of experience scoring sleep studies.
4. Adjudication Method
- Method: Consensus manual scoring. The document states that the autoscoring software was compared to "the consensus manual scoring." For "No Consensus" epochs in sleep staging (1654 epochs) and respiratory events (833 epochs), it's explicitly mentioned that "all three technologists gave different scores," indicating that if the three experts did not agree, those epochs were labeled as "No Consensus" and likely excluded from the primary agreement calculations against the consensus. This suggests a form of 3-way agreement leading to "consensus."
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
- No, a direct MRMC comparative effectiveness study involving human readers improving with AI vs without AI assistance was not explicitly described in the provided text for the autoscoring module. The study focuses on evaluating the standalone performance of the autoscoring algorithm against manual scoring ("gold standard").
6. Standalone (Algorithm Only) Performance
- Yes, a standalone performance study was done. The "Software Verification and Validation Testing - Autoscoring" section details the comparison of the "Cerebra Sleep System (CSS) autoscoring software" to the "consensus manual scoring," which evaluates the algorithm's performance without a human-in-the-loop.
7. Type of Ground Truth Used
- Type of Ground Truth: Expert consensus manual scoring. The text states: "The testing consisted of comparing the autoscoring software with the current gold standard (manual scoring) for key sleep variables." And further, "Sleep recordings were manually scored by three board-certified registered polysomnographic technologists (RPSGT)..."
8. Sample Size for the Training Set
- The document does not explicitly state the sample size used for the training set of the autoscoring module. It only specifies the test set size of 138 studies.
9. How Ground Truth for Training Set was Established
- The document does not provide details on how the ground truth for the training set was established. It focuses solely on the validation/test set.
§ 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).