(125 days)
The ANNE Chest is a wearable, wireless sensor intended for the measurement of electrocardioaraphy (ECG) waveforms, heart rate, respiratory rate, activity, fall detection, body position, and skin temperature. The ANNE Chest sensor is not intended to monitor or measure respiratory rate while the patient undergoes significant motion or is active. The ANNE Chest sensor communicates with compatible software applications for the display, storage, and analysis of data. The device is intended to provide continuous physiological information as an aid to diagnosis and treatment by healthcare professionals in general care patients who are 12 years of age or older in clinical and home environments. The device is not intended for use on critical care patients.
The ANNE Chest Sensor is a skin-mounted, bio-integrated sensor that collects real-time biosignals including electrocardiography (ECG), 3-axis accelerometry, and temperature to measure vital signs such as heart rate, respiratory rate, body position, fall detection, and skin temperature. The sensor communicates via Bluetooth to the Sibel SDK, which may be integrated within software applications for the display and storage of data. The ECG signal obtained by the ANNE Chest sensor is not intended for manual discrimination of any arrhythmias or cardiac conditions.
The provided document, primarily an FDA 510(k) clearance letter and summary, details the "ANNE Chest" device and its substantial equivalence to predicate devices. It includes performance data, but it does not contain the level of detail typically found in a comprehensive clinical study report for evaluating acceptance criteria and proving device performance for areas like AI/ML algorithms.
Specifically, the document lacks the following information crucial for a detailed response on acceptance criteria and study proving device meets them for AI/ML components:
- Explicit Acceptance Criteria Tables: While performance specifications are listed for heart rate, respiratory rate, and skin temperature, these are not presented as explicit "acceptance criteria" against which the device passed.
- Detailed Study Design for each feature: Only a brief summary of a clinical validation study for respiratory rate accuracy is provided. Details on ECG waveform analysis, activity, fall detection, and body position, if they incorporate AI/ML, are not detailed in terms of their dedicated performance studies.
- Sample Size for Test Set and Data Provenance for all features: Only n=40 is mentioned for respiratory rate. Data provenance (country, retrospective/prospective) is not specified.
- Expert Ground Truth Details: The number and qualifications of experts for establishing ground truth are not mentioned for any of the features.
- Adjudication Method: No information on adjudication is provided.
- MRMC Study Details: No information on multi-reader multi-case studies or effect sizes of human reader improvement with AI assistance is provided.
- Standalone AI Performance: The document describes the device as a sensor measuring physiological signals. It mentions software for display, storage, and analysis, but doesn't explicitly refer to AI/ML algorithms that operate in a "standalone" fashion where their performance metrics (e.g., sensitivity, specificity, AUC) against a ground truth would be relevant. The respiratory rate accuracy is presented as a direct comparison to a reference device.
- Ground Truth Type for all features: Only "End Tidal Carbon Dioxide (EtCO2) monitor reference" is mentioned for respiratory rate.
- Training Set Sample Size and Ground Truth Establishment: No information about training sets or how their ground truth was established is provided, suggesting that the device's functions might rely more on signal processing rather than intricate AI/ML models requiring large training datasets with defined ground truth methods.
Based on the limited information provided in the document, here's what can be extracted and inferred:
Acceptance Criteria and Reported Device Performance
The document provides performance specifications for certain physiological measurements. While not explicitly stated as "acceptance criteria," these are the performance targets the device claims to meet.
Table 1: Reported Device Performance
| Parameter | Acceptance Criteria / Specification | Reported Device Performance (from text) |
|---|---|---|
| Heart Rate | 30 - 270 bpm | (the greater of ±10% or ±5bpm) |
| Respiratory Rate | 8 - 35 bpm | Mean absolute error (MAE) of 1.27 breaths per minute (against capnography reference) at 8, 13, 23, 27, and 35 bpm. (±3 bpm RMSE) |
| Skin Temperature | 73.4°F - 109.4°F (23°C - 43°C) | ±0.54°F (±0.3°C) |
| Activity | (Not specified) | Accelerometer-based |
| Body Position | (Not specified) | Body Position |
| Fall Detection | (Not specified) | Sensor collects 3-axis accelerometry |
| ECG Waveform | Compliant to IEC standards | Compliant to IEC 60601-2-27 and IEC 60601-2-47 |
| ECG Sampling Freq. | (Not specified) | 512 Hz |
| ECG Streaming Freq. | (Not specified) | 256 Hz |
| ECG Resolution | (Not specified) | 18 bit |
Note: The document explicitly states: "The ECG signal obtained by the ANNE Chest sensor is not intended for manual discrimination of any arrhythmias or cardiac conditions." This suggests the ECG waveform is for general display and rate calculation, not for specific diagnostic interpretation that might involve AI/ML for abnormality detection.
Study Details Proving Device Meets Acceptance Criteria
1. Sample Size for Test Set and Data Provenance:
* Respiratory Rate: n=40 healthy adult and adolescent subjects.
* Data Provenance: Not specified (e.g., country of origin). The study is described as "a clinical validation study," implying it was prospective for this evaluation.
* Other features (ECG, Heart Rate, Skin Temperature, Activity, Fall Detection, Body Position): No specific sample sizes for clinical validation studies are mentioned, only adherence to standards (e.g., IEC 60601-2-27, IEC 60601-2-47 for ECG) and "performance testing." This suggests these features might have been validated through bench testing or engineering verification rather than clinical studies of the type described for respiratory rate.
2. Number of Experts used to establish Ground Truth and Qualifications:
* Not specified. The ground truth for respiratory rate was an "End Tidal Carbon Dioxide (EtCO2) monitor reference," which is an objective measurement, not expert consensus. For other features, no details about expert involvement for ground truth are provided.
3. Adjudication Method for the Test Set:
* None specified. Given the use of an objective reference device (EtCO2 monitor) for respiratory rate, human adjudication would not be applicable for this specific metric.
4. If a Multi-Reader Multi-Case (MRMC) comparative effectiveness study was done:
* No. The document does not describe any MRMC studies or human reader improvement with AI assistance. The device focuses on physiological measurements rather than interpretive tasks for human readers.
5. If a Standalone (i.e., algorithm only without human-in-the-loop performance) was done:
* Implied for Respiratory Rate accuracy. The study directly compared the ANNE Chest sensor's respiratory rate measurement against a capnography reference, which aligns with standalone performance evaluation for that specific function. For other parameters like heart rate and skin temperature, this is also implied through performance specifications. The device itself is a measurement tool, outputs direct physiological values, and does not describe AI intended for interpretation or decision support that would typically have a human-in-the-loop component.
6. The Type of Ground Truth Used:
* Respiratory Rate: Objective reference standard: "End Tidal Carbon Dioxide (EtCO2) monitor reference."
* Other features: Not explicitly stated, but likely objective measurements from reference devices/standards (e.g., test ECG signals for ECG compliance, calibrated thermometers for skin temperature).
7. The Sample Size for the Training Set:
* Not specified. The document does not indicate the use of machine learning models that would require a distinct training set. The descriptions of the device functions point towards signal processing and known algorithms for extracting vital signs from raw sensor data, rather than complex AI requiring large, labeled training datasets.
8. How the Ground Truth for the Training Set was Established:
* Not applicable based on available information. Since no training set or complex AI model development requiring one is mentioned, this information is not provided.
§ 870.2910 Radiofrequency physiological signal transmitter and receiver.
(a)
Identification. A radiofrequency physiological signal transmitter and receiver is a device used to condition a physiological signal so that it can be transmitted via radiofrequency from one location to another, e.g., a central monitoring station. The received signal is reconditioned by the device into its original format so that it can be displayed.(b)
Classification. Class II (performance standards).