Search Filters

Search Results

Found 2 results

510(k) Data Aggregation

    K Number
    K231546
    Device Name
    Somfit
    Date Cleared
    2023-11-30

    (184 days)

    Product Code
    Regulation Number
    868.2375
    Reference & Predicate Devices
    Why did this record match?
    Reference Devices :

    K191031, K130013, K183625

    AI/MLSaMDIVD (In Vitro Diagnostic)TherapeuticDiagnosticis PCCP AuthorizedThirdpartyExpeditedreview
    Intended Use

    The Somfit system is a non-invasive prescription device for home use with patients suspected to have sleep-related breathing disorders. The Somfit is a diagnostic aid for the detection of sleep-related breathing disorders, sleep staging (REM, N1, N2, N3, Wake), and snoring level. The Somfit system acquires electrical data from three frontal electrodes, triaxial accelerometer data, acoustic and plethysmographic data. The Somfit calculates and reports to clinicians EEG/EOG channels, Sleep Stages, SpO2, Peripheral Arterial Tonometry (PAT) signal, pulse rate, and snoring level. The Somfit calculates and reports to clinicians derived parameters such as PAT-derived Apnea Hypopnea Index, Obstructive Desaturation Index; and hypnogram-derived indices such as time in each sleep stage. Somfit data is not intended to be used as the sole or primary basis for diagnosing any sleep-related breathing disorder, prescribing treatment, or determining whether additional diagnostic assessment is warranted. The Somfit is not intended for use as life support equipment, for example vital signs monitoring in intensive care unit. The Somfit is a prescription device indicated for adult patients aged 21 years and over.

    Device Description

    The Somfit is a home-based sleep monitoring device which records signals from the patient's forehead and surrounding environment. Somfit is a wearable, low voltage, battery operated device which is attached to subject forehead via a self-adhesive and disposable skin electrode patch. The electrodes are placed on the anterior Prefrontal cortex (PFC) at the Fp1 and Fp2 positions according to the 10/20 EEG system. The device allows for recording of two frontal EEG signals, pulse rate, SPO2, Peripheral Arterial Tonometry (PAT), PPG, motion, and snore. Somfit uses a mobile phone application to acquire data wirelessly via Bluetooth LE technology, then transfer into a secure cloud, for management, storage and postprocessing. The software reports measured parameters in a format compatible with AASM guidelines, including sleep time, pAHI and conventional graphical displays such as a hypnogram.

    AI/ML Overview

    The Somfit device underwent rigorous testing to establish its performance against predefined acceptance criteria, particularly for its ability to detect sleep stages and related parameters.

    Here's a breakdown of the acceptance criteria and the supporting study:

    1. Table of Acceptance Criteria and Reported Device Performance

    The document does not explicitly present a formal "acceptance criteria" table with specific thresholds for each metric. However, the clinical performance section implicitly defines acceptance by demonstrating non-inferiority to the predicate device (Watch-PAT300) and superior or comparable performance against the gold standard (PSG) for key metrics.

    Based on the provided text, here's a table summarizing the performance metrics that act as implicit acceptance criteria:

    Implicit Acceptance Criteria & Reported Device Performance for Somfit

    MetricImplicit Acceptance Criteria (Comparative)Reported Somfit PerformanceReported Watch-PAT300 PerformanceGround Truth (PSG)
    PAT-derived AHI (pAHI)Non-inferiority to predicate (Watch-PAT300) compared to PSG.Mean difference with Watch-PAT300: 0.294 (95% CI: -2.661, 2.074).
    Mean difference with PSG: 0.658 (CI: -1.012, 2.326).Mean difference with PSG: 2.499 (CI: 0.732, 4.265).Polysomnography (PSG)
    Obstructive Desaturation Index (ODI)Superior or comparable accuracy to predicate.Mean difference with PSG: 0.658 (CI: -1.012, 2.326).Mean difference with PSG: 2.499 (CI: 0.732, 4.265).Polysomnography (PSG)
    N1/N2 Merged Sleep Stages AgreementHigh agreement with PSG; superior to predicate.80.92% agreement (95% CI: 79.96, 81.89)61.77% agreement (95% CI: 60.99, 62.55)Polysomnography (PSG) (consensus of three scoring sleep technologists)
    NREM/REM/Wake Stage Differential AgreementHigh agreement with PSG; superior to predicate.86.79% agreement (95% CI: 85.76, 87.81)74.42% agreement (95% CI: 73.51, 75.32)Polysomnography (PSG) (consensus of three scoring sleep technologists)
    Sleep/Wake Determination AgreementHigh agreement with PSG; superior to predicate.89.45% agreement (95% CI: 88.53, 90.38)82.09% agreement (95% CI: 81.37, 82.82)Polysomnography (PSG) (consensus of three scoring sleep technologists)
    SpO2 Accuracy (ARMS)Maintain or outperform predicate, satisfy ISO 80601-2-61.1.46%1.88% (Watch-PAT300)Gold standard SaO2 (invasive controlled desaturation study)
    Pulse Rate AccuracySatisfy ISO 80601-2-61.1.91 BPM (vs Masimo), 1.81 BPM (vs Nellcor)Not explicitly compared hereMasimo and Nellcor reference oximeters

    2. Sample Size and Data Provenance

    • Test Set Sample Size: 92 subjects for the comparative clinical study involving Somfit, Watch-PAT300, and PSG.
    • Data Provenance: The subjects were recruited from 3 site locations in Australia. The study was a prospective clinical study conducted to compare the devices' performance.

    3. Number of Experts and Qualifications for Ground Truth

    • Number of Experts: Three scoring sleep technologists.
    • Qualifications of Experts: The document states they were "scoring sleep technologists." Specific educational background or years of experience are not provided, but their role implies expertise in polysomnography scoring.

    4. Adjudication Method for the Test Set

    The adjudication method for sleep staging ground truth was expert consensus. The PSG epochs were classified based on the "consensus of three scoring sleep technologists." This implies a form of 3-way consensus. Further details on how discrepancies were resolved (e.g., majority rule, discussion to reach agreement) are not provided.

    5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study

    No, a traditional MRMC comparative effectiveness study was not explicitly described in the provided text in the context of human readers improving with AI vs. without AI assistance. The study evaluated the device's performance (Somfit and Watch-PAT300) against a human-scored gold standard (PSG), rather than measuring the improvement of human readers assisted by AI. The Somfit system is described as a diagnostic aid that calculates and reports parameters to clinicians, implying that clinicians review the output, but the study design was not focused on measuring changes in human reader performance.

    6. Standalone (Algorithm Only) Performance

    Yes, the study primarily evaluated the standalone performance of the Somfit algorithm (and Watch-PAT300) in deriving parameters like pAHI and sleep stages, comparing its output directly to the gold standard PSG and the predicate device. There was no human-in-the-loop component as part of the evaluated performance for regulatory submission.

    7. Type of Ground Truth Used

    The primary ground truth used for the clinical study was Polysomnography (PSG), which is considered the gold standard technology for diagnosing sleep disorders. Specifically, for sleep staging, it was expert consensus-scored PSG data. For SpO2, it was gold standard SaO2 data obtained from an invasively controlled desaturation study.

    8. Sample Size for the Training Set

    The document does not provide information on the sample size for the training set for the Somfit algorithms. It focuses solely on the clinical validation/test set.

    9. How the Ground Truth for the Training Set Was Established

    As the training set details are not provided, the method for establishing its ground truth is also not mentioned in the document.

    Ask a Question

    Ask a specific question about this device

    K Number
    K222579
    Date Cleared
    2023-02-23

    (182 days)

    Product Code
    Regulation Number
    868.2375
    Reference & Predicate Devices
    Why did this record match?
    Reference Devices :

    K211407, K183625

    AI/MLSaMDIVD (In Vitro Diagnostic)TherapeuticDiagnosticis PCCP AuthorizedThirdpartyExpeditedreview
    Intended Use

    The Belun Sleep System BLS-100 is a wearable device intended to record, analyze, display, export, and store biophysical parameters to aid in evaluating moderate to severe sleep-related breathing disorders suspected of sleep apnea. The device is intended for use in clinical and home settings under the direction of a Healthcare Professional (HCP).

    Device Description

    The Belun Sleep System BLS-100 comprises a sensor that is worn on the proximal phalanx of index finger (Belun Ring) over-night whilst the subject is sleeping and a stand-alone analysis software (Belun Sleep AI). The Belun Ring has a small biocompatible enclosure. The sensor has 2 LEDs, one in the red spectrum and the other in the infrared spectrum, and an accelerometer. The sensor is placed on the proximal phalanx of the index finger, with the sensor window applied against the palmar side of the proximal phalanx of the index finger. The sensor measures the reflected red/infrared signals to record the photoplethysmograph (PPG) signal. The accelerometer is used to detect movement. The data recorded by the Belun Ring is stored in device on-board memory. The data is exported when the Belun Ring is returned to the prescribing HCP via USB or Bluetooth and passed to the Belun Sleep AI Software, which is standalone PC software. The Belun Sleep Al loads and processes the signal from the exported data and generates the apnea-hypopnea index (bAHI) and sleep staging identification (bSTAGES).

    AI/ML Overview

    Let's break down the information regarding the acceptance criteria and the study that proves the device meets them for the Belun Sleep System BLS-100.

    1. Table of Acceptance Criteria and Reported Device Performance

    The acceptance criteria are implicitly defined by the clinical study results being presented as sufficient evidence for clearance. While explicit "acceptance criteria" are not listed as pass/fail thresholds in a formal table, the provided performance metrics represent the device's demonstrated capabilities.

    Here's a table summarizing the reported device performance, which the FDA accepted as evidence of substantial equivalence:

    Metric (Implicit Acceptance Criteria)Performance (Belun Sleep System BLS-100)
    AHI Accuracy (at cutoff 15)0.877
    AHI Sensitivity (at cutoff 15)0.898
    AHI Specificity (at cutoff 15)0.860
    AHI Accuracy (at cutoff 30)0.925
    AHI Sensitivity (at cutoff 30)0.840
    AHI Specificity (at cutoff 30)0.951
    Sleep Stage Accuracy (Wake)0.885
    Sleep Stage Sensitivity (Wake)0.604
    Sleep Stage Specificity (Wake)0.961
    Sleep Stage Accuracy (REM)0.908
    Sleep Stage Sensitivity (REM)0.712
    Sleep Stage Specificity (REM)0.944
    Sleep Stage Accuracy (NREM)0.827
    Sleep Stage Sensitivity (NREM)0.904
    Sleep Stage Specificity (NREM)0.695
    Mean difference between bTST and PSG-TST21.8 minutes
    Standard deviation of difference between bTST and PSG-TST41.6 minutes
    Mean absolute difference between bTST and PSG-TST30.8 minutes

    2. Sample Size and Data Provenance

    • Sample Size for Test Set: 106 patients suspected of obstructive sleep apnea (OSA).
    • Data Provenance: The study compared the device's performance against overnight polysomnography (PSG) studies conducted in a sleep laboratory. The location of the sleep laboratory (country of origin) is not explicitly stated in the provided text. The study design implies this was a prospective collection of data for this evaluation, as it describes patients going through a study with both the Belun device and PSG.

    3. Number of Experts and Qualifications for Ground Truth

    • Number of Experts: At least two experts, as the text states "a senior sleep tech scorer and reviewed by a board-certified sleep physician."
    • Qualifications of Experts:
      • One "senior sleep tech scorer."
      • One "board-certified sleep physician."

    4. Adjudication Method for the Test Set

    The adjudication method used to establish the ground truth for the test set was: "All sleep studies were manually scored based on the AASM scoring manual (version 2.4) by a senior sleep tech scorer and reviewed by a board-certified sleep physician." This indicates a two-step process where one scorer performs the primary scoring, and a physician provides a review, implying a form of consensus or verification, though not a multi-reader disagreement resolution specifically.

    5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study

    • No, a multi-reader multi-case (MRMC) comparative effectiveness study was not done. The study focuses on the standalone performance of the device's AI algorithms (bAHI and bSTAGES) compared to PSG ground truth, not on how human readers' performance improves with or without AI assistance.

    6. Standalone Performance

    • Yes, a standalone (algorithm only without human-in-the-loop performance) study was done. The clinical study section directly reports the accuracy, sensitivity, and specificity of the Belun Sleep System BLS-100's AHI and sleep staging calculations compared to PSG results, indicating the performance of the device's algorithms themselves. The statement "All investigators, sleep lab team, and scorers were blinded to the results until statistical analysis was performed" further supports that the device's output was generated independently.

    7. Type of Ground Truth Used

    • The type of ground truth used was expert consensus based on Polysomnography (PSG) studies, manually scored according to the American Academy of Sleep Medicine (AASM) guidelines (version 2.4) by a senior sleep tech scorer and reviewed by a board-certified sleep physician.

    8. Sample Size for the Training Set

    • The document does not explicitly state the sample size for the training set used for the Belun Sleep AI's deep-learning algorithms. It only provides details for the clinical validation (test) set.

    9. How the Ground Truth for the Training Set Was Established

    • The document does not explicitly state how the ground truth for the training set was established. It mentions that the clinical evaluation "confirmed that the Belun Sleep System deep-learning algorithms calculating the Belun Apnea Hypopnea Index (bAHI) and Belun Sleep Stage (bSTAGES) generate comparable output to human manual scoring of an Apnea Hypopnea Index (AHI) from Polysomnography (PSG) studies, using American Academy of Sleep Medicine (AASM) scoring guidelines for adult patients". While this describes the validation against PSG ground truth, it doesn't detail the ground truth establishment process for the data used to train the AI models.
    Ask a Question

    Ask a specific question about this device

    Page 1 of 1