SleepStageML is intended for assisting the diagnostic evaluation by a qualified clinician to assess sleep quality from level 1 polysomnography (PSG) recordings in a clinical environment in patients aged 18 and older.

SleepStageML is a software-only medical device to be used to analyze physiological signals and automatically score sleep stages. All outputs are subject to review by a qualified clinician.

Device Description

SleepStageML is an Artificial Intelligence/Machine Learning (Al/ML)-enabled software-only medical device that analyzes polysomnography (PSG) recordings and automatically scores sleep stages. It is intended for assisting the diagnostic evaluation by a qualified clinician to assess sleep quality in patients aged 18 and older.

Qualified clinicians (also referred to as clinical users) such as sleep physicians, sleep technicians, or registered PSG technologists (RPSGTs) who are qualified to review PSG studies, provide PSG recordings in European Data Format (EDF) file format through a secure file transfer system to Beacon Biosignals. SleepStageML automatically analyzes the provided PSG recording and return an EDF file containing the original PSG recording with software-generated sleep stage annotations (i.e., Wake (W), non-REM 1 (N1), non-REM 2 (N2), non-REM 3 (N3), and REM (R)) back to the clinical user. The EDF files containing PSG signals as well as sleep stage annotations are referred to as EDF+. The returned EDF+ files can then be reviewed by the qualified clinicians via the users' PSG viewing software. The recordings processed by SleepStageML are level-1 PSG recordings obtained in an attended setting in accordance with American Association of Sleep Medicine (AASM) recommendations with respect to minimum sampling rate, electroencephalography (EEG) channels, and EEG locations. SleepStageML only uses the EEG signals in provided PSGs and does not consider electromyography (EMG) or electrooculography (EOG) signals when performing sleep staging. The sleep stage outputs of SleepStageML are intended to be comparable to sleep stages as defined by AASM guidelines. SleepStageML software outputs are subject to qualified clinician's review.

AI/ML Overview

Here's a breakdown of the acceptance criteria and the study proving the device meets them, based on the provided FDA 510(k) summary for SleepStageML:

Acceptance Criteria and Reported Device Performance

Sleep Staging Comparisons	Acceptance Criteria (Predicate Reference: Sleep Profiler, K153412, N=43 subjects)	Reported Device Performance (SleepStageML, N=100 subjects)
Overall Agreement (OA)
W	89%	96.1% (95% CI: 95.4%, 96.8%)
N1	89%	94.5% (95% CI: 93.7%, 95.2%)
N2	81%	87.1% (95% CI: 85.9%, 88.3%)
N3	91%	92.9% (95% CI: 91.8%, 93.8%)
R	95%	97.3% (95% CI: 96.7%, 97.9%)
Positive Agreement (PA)
W	73%	88.9% (95% CI: 86.5%, 91.2%)
N1	25%	58.4% (95% CI: 54.2%, 62.4%)
N2	77%	79.8% (95% CI: 77.7%, 81.8%)
N3	76%	93.0% (95% CI: 89.8%, 95.7%)
R	74%	93.1% (95% CI: 91.5%, 94.5%)
Negative Agreement (NA)
W	94%	98.5% (95% CI: 98.2%, 98.8%)
N1	93%	96.2% (95% CI: 95.4%, 96.9%)
N2	84%	94.2% (95% CI: 93.2%, 95.0%)
N3	94%	92.9% (95% CI: 91.7%, 93.9%)
R	97%	98.0% (95% CI: 97.3%, 98.6%)
Multi-stage Agreement	Not explicitly stated for predicate in a comparable way, but implied.	84.02% (Calculated from N=100 subjects total epochs: 86,983 overall, 2,289 no consensus)

Study Details:

Sample sizes used for the test set and data provenance:
- Test Set Sample Size: 100 patients.
- Data Provenance: Retrospective pivotal validation study using previously collected clinical polysomnography (PSG) recordings. The recordings were randomly selected from three Level 1 clinical PSG data sources. The document does not specify the country of origin of the data.
Number of experts used to establish the ground truth for the test set and their qualifications:
- Number of Experts: Three (3) registered PSG technologists (RPSGTs).
- Qualifications: Each RPSGT had at least 5 years of experience in clinical scoring of sleep studies.
Adjudication method for the test set:
- Method: 2/3 majority scoring. Expert consensus sleep stages were constructed using the stage per epoch where at least 2 of the 3 experts agreed. Epochs where all 3 RPSGTs disagreed were excluded.
If a multi-reader multi-case (MRMC) comparative effectiveness study was done:
- No, an MRMC study comparing human readers with AI vs. without AI assistance was not explicitly detailed. The study focused on the standalone performance of the AI algorithm against human expert consensus to demonstrate non-inferiority to a predicate device. The device's indication for use explicitly states, "All outputs are subject to review by a qualified clinician," indicating a human-in-the-loop design, but the described performance study is primarily a standalone evaluation.
If a standalone (i.e., algorithm only without human-in-the-loop performance) was done:
- Yes, the clinical validation test evaluated the SleepStageML software's performance "against the expert consensus sleep stages" in a standalone manner. The device's outputs are intended to be reviewed by a clinician, but the performance metrics reported are for the algorithm's direct output compared to ground truth.
The type of ground truth used:
- Type: Expert Consensus. The ground truth was established by three RPSGTs, with a 2/3 majority rule for consensus.
The sample size for the training set:
- The document states, "SleepStageML uses a deep learning algorithm based on convolutional neural networks, which was trained on a large and diverse set of PSG recordings with sleep staging labels." However, a specific sample size for the training set is not provided in the summary.
How the ground truth for the training set was established:
- The document states the training was on "PSG recordings with sleep staging labels." It does not explicitly detail the method for establishing ground truth for the training set (e.g., if it was also expert consensus, single expert, or another method). However, given the nature of sleep staging, it is highly likely that these labels were also derived from expert annotations, similar to the test set, though possibly not with the same rigorous 3-expert consensus and adjudication for every record.

Summary

{0}------------------------------------------------

March 8, 2024

Image /page/0/Picture/1 description: The image shows the logo of the U.S. Food and Drug Administration (FDA). On the left is the Department of Health & Human Services logo. To the right of that is the FDA logo, which is a blue square with the letters "FDA" in white. To the right of the blue square is the text "U.S. FOOD & DRUG ADMINISTRATION" in blue.

Beacon Biosignals, Inc. Delphine Lemoine Senior Regulatory Affairs Specialist 22 Boston Wharf Road 7th Floor, Unit 41 Boston, Massachusetts 02210

Re: K233438

Trade/Device Name: SleepStageML Regulation Number: 21 CFR 882.1400 Regulation Name: Electroencephalograph Regulatory Class: Class II Product Code: OLZ Dated: February 9, 2024 Received: February 9, 2024

Dear Delphine Lemoine:

We have reviewed your section 510(k) premarket notification of intent to market the device referenced above and have determined the device is substantially equivalent (for the indications for use stated in the enclosure) to legally marketed predicate devices marketed in interstate commerce prior to May 28, 1976, the enactment date of the Medical Device Amendments, or to devices that have been reclassified in accordance with the provisions of the Federal Food, Drug, and Cosmetic Act (the Act) that do not require approval of a premarket approval application (PMA). You may, therefore, market the device, subject to the general controls provisions of the Act. Although this letter refers to your product as a device, please be aware that some cleared products may instead be combination products. The 510(k) Premarket Notification Database available at https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm identifies combination product submissions. The general controls provisions of the Act include requirements for annual registration, listing of devices, good manufacturing practice, labeling, and prohibitions against misbranding and adulteration. Please note: CDRH does not evaluate information related to contract liability warranties. We remind you, however, that device labeling must be truthful and not misleading.

If your device is classified (see above) into either class II (Special Controls) or class III (PMA), it may be subject to additional controls. Existing major regulations affecting your device can be found in the Code of Federal Regulations, Title 21, Parts 800 to 898. In addition, FDA may publish further announcements concerning your device in the Federal Register.

FDA's substantial equivalence determination also included the review and clearance of your Predetermined Change Control Plan (PCCP). Under section 515C(b)(1) of the Act, a new premarket notification is not required for a change to a device cleared under section 510(k) of the Act, if such change is consistent with an

{1}------------------------------------------------

established PCCP granted pursuant to section 515C(b)(2) of the Act. Under 21 CFR 807.81(a)(3), a new premarket notification is required if there is a major change or modification in the intended use of a device, or if there is a change or modification in a device that could significantly affect the safety or effectiveness of the device, e.g., a significant change or modification in design, material, chemical composition, energy source, or manufacturing process. Accordingly, if deviations from the established PCCP result in a major change or modification in the intended use of the device, or result in a change or modification in the device that could significantly affect the safety or effectiveness of the a new premarket notification would be required consistent with section 515C(b)(1) of the Act and 21 CFR 807.81(a)(3). Failure to submit such a premarket submission would constitute adulteration and misbranding under sections 501(f)(1)(B) and 502(o) of the Act, respectively.

Additional information about changes that may require a new premarket notification are provided in the FDA guidance documents entitled "Deciding When to Submit a 510(k) for a Change to an Existing Device" (https://www.fda.gov/media/99812/download) and "Deciding When to Submit a 510(k) for a Software Change to an Existing Device" (https://www.fda.gov/media/99785/download).

Your device is also subject to, among other requirements, the Quality System (OS) regulation (21 CFR Part 820), which includes, but is not limited to, 21 CFR 820.30, Design controls; 21 CFR 820.90, Nonconforming product; and 21 CFR 820.100, Corrective and preventive action. Please note that regardless of whether a change requires premarket review, the QS regulation requires device manufacturers to review and approve changes to device design and production (21 CFR 820.30 and 21 CFR 820.70) and document changes and approvals in the device master record (21 CFR 820.181).

Please be advised that FDA's issuance of a substantial equivalence determination does not mean that FDA has made a determination that your device complies with other requirements of the Act or any Federal statutes and regulations administered by other Federal agencies. You must comply with all the Act's requirements, including, but not limited to: registration and listing (21 CFR Part 807); labeling (21 CFR Part 801); medical device reporting of medical device-related adverse events) (21 CFR Part 803) for devices or postmarketing safety reporting (21 CFR Part 4, Subpart B) for combination products (see https://www.fda.gov/combination-products/guidance-regulatory-information/postmarketing-safety-reportingcombination-products); good manufacturing practice requirements as set forth in the quality systems (QS) regulation (21 CFR Part 820) for devices or current good manufacturing practices (21 CFR Part 4, Subpart A) for combination products; and, if applicable, the electronic product radiation control provisions (Sections 531-542 of the Act); 21 CFR Parts 1000-1050.

Also, please note the regulation entitled, "Misbranding by reference to premarket notification" (21 CFR 807.97). For questions regarding the reporting of adverse events under the MDR regulation (21 CFR Part 803), please go to https://www.fda.gov/medical-device-safety/medical-device-reportingmdr-how-report-medical-device-problems.

For comprehensive regulatory information about medical devices and radiation-emitting products, including information about labeling regulations, please see Device Advice (https://www.fda.gov/medicaldevices/device-advice-comprehensive-regulatory-assistance) and CDRH Learn (https://www.fda.gov/training-and-continuing-education/cdrh-learn). Additionally, you may contact the Division of Industry and Consumer Education (DICE) to ask a question about a specific regulatory topic. See the DICE website (https://www.fda.gov/medical-device-advice-comprehensive-regulatory

{2}------------------------------------------------

assistance/contact-us-division-industry-and-consumer-education-dice) for more information or contact DICE by email (DICE@fda.hhs.gov) or phone (1-800-638-2041 or 301-796-7100).

Sincerely,

Image /page/2/Picture/3 description: The image contains the text "Patrick Antkowiak -S" in black font. To the left of the text is a faded, light blue graphic that appears to be the letters "FDA". The text is arranged in two lines, with "Patrick" on the first line and "Antkowiak -S" on the second line.

for Jay Gupta Assistant Director DHT5A: Division of Neurosurgical, Neurointerventional and Neurodiagnostic Devices OHT5: Office of Neurological and Physical Medicine Devices Office of Product Evaluation and Quality Center for Devices and Radiological Health

Enclosure

{3}------------------------------------------------

Indications for Use

Submission Number (if known)

K233438

Device Name

SleepStageML

Indications for Use (Describe)

SleepStageML is a software-only medical device to be used to analyze physiological signals and automatically score sleep stages. All outputs are subject to review by a qualified clinician.

Type of Use (Select one or both, as applicable)

Prescription Use (Part 21 CFR 801 Subpart D)

Over-The-Counter Use (21 CFR 801 Subpart C)

CONTINUE ON A SEPARATE PAGE IF NEEDED.

This section applies only to requirements of the Paperwork Reduction Act of 1995.

DO NOT SEND YOUR COMPLETED FORM TO THE PRA STAFF EMAIL ADDRESS BELOW.

The burden time for this collection of information is estimated to average 79 hours per response, including the time to review instructions, search existing data sources, gather and maintain the data needed and complete and review the collection of information. Send comments regarding this burden estimate or any other aspect of this information collection, including suggestions for reducing this burden, to:

Department of Health and Human Services Food and Drug Administration Office of Chief Information Officer Paperwork Reduction Act (PRA) Staff PRAStaff(@fda.hhs.gov

"An agency may not conduct or sponsor, and a person is not required to respond to, a collection of information unless it displays a currently valid OMB number."

{4}------------------------------------------------

Image /page/4/Picture/0 description: The image shows the logo for Beacon Biosignals. The logo has two parts: the text on the left and a graphic on the right. The text reads "BEACON" in large, bold, dark gray letters above "BIOSIGNALS" in smaller, lighter gray letters. To the right of the text is a graphic of a brain made of a continuous, squiggly line in a light green color.

Submitter

Name: Beacon Biosignals, Inc.

Contact Person: Delphine Lemoine

Address: 22 Boston Wharf Road, 7th floor, Unit 41, Boston MA, 02210

Telephone: +33 159068311

Date Submitted: October 13, 2023

Subject Device

Trade Name: SleepStageML Common Name: SleepStageML Product Code: OLZ Regulatory Class: II (21 C.F.R. 882.1400) Review Panel: Neurology Predicate Device: Sleep Profiler (K153412, Advanced Brain Monitoring, Inc.) Reference Device: SomnoMetry (K221179, Neumetry Medical Inc.)

Device description

The intended patients for PSG studies include those who are suspected of having, or have been diagnosed with, a sleep or sleep-related disorder. This could include sleep disordered breathing, insomnia, periodic limb movement disorder, narcolepsy, hypersomnia, or other parasomnias.

SleepStageML uses a deep learning algorithm based on convolutional neural networks, which was trained on a large and diverse set of PSG recordings with sleep staging labels.

{5}------------------------------------------------

Image /page/5/Picture/0 description: The image shows the logo for Beacon Biosignals. The logo consists of the word "BEACON" in bold, dark gray letters stacked on top of the word "BIOSIGNALS" in a smaller font and the same color. To the right of the text is a stylized graphic of a brain, depicted as a series of connected, rounded lines in a light teal color. The lines form a continuous, wavy pattern that resembles brainwaves.

The recordings used for development were collected from a variety of sources containing a wide range of patient demographics and clinical sites for robust performance.

Indication for Use

SleepStageML is a software-only medical device to be used to analyze physiological signals and automatically score sleep stages. All outputs are subject to review by a qualified clinician.

Indication for Use Comparison and Technological Characteristics Comparison

SleepStageML uses an automated algorithm that performs signal preprocessing and categorizes each epoch as one of the defined sleep stages using a machine learning classifier. It has similar intended use and indications for use as the predicate device Sleep Profiler (K153412, Advanced Brain Monitoring, Inc.). The predicate device does not explicitly mention the use of AI/ML, therefore, SomnoMetry (K221179, Neumetry Medical Inc.), which does use a similar Al/ML-based algorithm for staging sleep, was added as the reference device. Verification and validation methods utilized to evaluate safety and efficacy of SleepStageML were consistent with methods used for evaluating the predicate and reference devices. Software documentation was provided as recommended in FDA "Guidance for the Content of Premarket Submissions for Software Contained in Medical Devices" dated June 14th, 2023, and the pivotal clinical validation study confirmed that the SleepStageML AI/ML software performance was substantially equivalent to that of the predicate device.

Element	Subject deviceSleepStageML	Predicate DeviceSleep Profiler(K153412)	Reference DeviceSomnoMetry(K221179)	Comparison ofTechnologicalCharacteristics
Classification	Class II, OLZ,Automated EventDetection Software forPolysomnograph withElectroencephalograph	Class II, OLZ,Automated EventDetection Software forPolysomnograph withElectroencephalograph	Class II, OLZ,Automated EventDetection Software forPolysomnograph withElectroencephalograph	Same.
Indication for use	SleepStageML isintended for assistingthe diagnosticevaluation by aqualified clinician toassess sleep qualityfrom level 1polysomnography(PSG) recordings in aclinical environment inpatients aged 18 andolder. SleepStageML isa software-onlymedical device to beused to analyzephysiological signalsand automaticallyscore sleep stages. Alloutputs are subject toreview by a qualifiedclinician.	Sleep Profiler isintended for use forthe diagnosticevaluation by aphysician to assesssleep quality and scoresleep disorderedbreathing events inadults only. The SleepProfiler is a software-only device to be usedunder the supervisionof a clinician to analyzephysiological signalsand automaticallyscore sleep studyresults; including thestaging of sleep,detection of arousals,snoring and sleepdisordered breathing	SomnoMetry isintended for use forthe diagnosticevaluation by aphysician to assesssleep quality and as anaid for the diagnosis ofsleep and respiratory-related sleep disordersin adults only.SomnoMetry is asoftware-only medicaldevice to be used toanalyze physiologicalsignals andautomatically scoresleep study results,including the stagingof sleep, AHI, anddetection of sleep-disordered breathing	Similar.SleepStageML onlyassists a qualifiedclinician in assessingsleep staging and doesnot include outputsrelated to sleepdisordered breathing.

		events (obstructiveapneas, hypopneasand respiratory eventrelatedarousals). Central andmixed apneas can bemanually markedwithin the records	events includingobstructive apneas. Itis intended to be usedunder the supervisionof a clinician in aclinical environment.All automaticallyscored events aresubject to verificationby a qualified clinician.
	Adults only (≥ 18 yearsold)	Adults only	Adults only	Similar.SleepStageML iscleared for individuals18 years and older,while the predicate andreference devices arecleared for individualsadults 22 years andolder.
Patientpopulation
Sleep stagingguidelines	American Academy ofSleep Medicine scoringmanual and guidelines.	American Academy ofSleep Medicine scoringmanual and guidelines.	American Academy ofSleep Medicine scoringmanual and guidelines.	Same.
Environment ofuse	Professional HealthcareFacility	Professional HealthcareFacility	Professional HealthcareFacility	Same.
Score sleepstaging	Yes	Yes	Yes	Same.
Algorithmdescription	Automated algorithmthat performs signalpreprocessing andcategorizes eachepoch as one of thedefined sleep stagesusing a machinelearning classifier.	Automated algorithmwhich spectrallydecomposes the EEGsignal, computesdescriptors of sleepmacro- andmicrostructure, andcategorizes eachepoch as one of thedefined sleep stages	A broad array of signalprocessing, dataindexing, conventionalmachine learning anddeep learningalgorithms/approachesare applied to the rawsignals to deriveactionable clinicalinsights.	Same.SleepStageML utilizesmachine learning anddeep learning in asimilar fashion aspredicate andreference devices.
PhysicalCharacteristics	User requires acomputer with accessto the internet toprovide original PSGrecordings in EDF filesto and receive returnedEDF+ files with addedsoftware-generatedsleep stageannotations fromBeacon Biosignals viathe provided securefile transfer system.	Sleep Profiler willoperate on anypersonal computerwith Windows XPoperating system andat least 2 GB of RAM.The speed by which astudy will process isdependent on thecomputer processorand the amount ofRAM	Web-based softwareoperates in the cloudwith Windows, MacOS, or Linux	The subject devicerequires that the userhas a computer tosend/receive EDFsto/from BeaconBiosignals. Thepredicate andreference devices aresoftware that aredirectly operated bythe user.
Software operateswithin infrastructurethat requires userauthentication,encryption of data at-rest and in-transit,checksum verifications,and access controls.		The person installingthe Java applet musthave administratorprivileges.User must have rightsto the account andgroup to access datavia the portalThe firewall is notconfigured to prevent	User authenticationwith strong password,authorization, end toend SSL encryption,access controls,checksum, networkand database controls,intrusion preventionsystem, andanonymization.	Similar.SleepStageMLoperates withininfrastructure thatimplementscybersecuritymeasures similar to themeasures implementedfor the predicate andreference devices.
Cybersecurity

{6}------------------------------------------------

Image /page/6/Picture/0 description: The image shows the logo for Beacon Biosignals. The logo has the word "BEACON" stacked on top of the word "BIOSIGNALS" on the left side of the image. To the right of the words is a graphic that looks like a brainwave in a light green color.

{7}------------------------------------------------

Image /page/7/Picture/0 description: The image shows the logo for Beacon Biosignals. The logo consists of two parts: the text "BEACON BIOSIGNALS" on the left and a stylized brainwave graphic on the right. The text is in a sans-serif font, with "BEACON" in a larger, bolder font than "BIOSIGNALS." The brainwave graphic is a series of connected, rounded lines in a light green color, resembling a sine wave or brainwave pattern.

		communication withaccess to the server viaports 22 and 30-29
PredeterminedChange ControlPlan (PCCP)	Includes PCCP thatallows for update ofthe signalpreprocessing,machine learningmodel, probabilitypost-processing, andsignal quality check.	No PCCP	No PCCP	Predicate andreference devices donot include a PCCP.

Summary of Tests

Beacon Biosignals conducted the necessary non-clinical testing and clinical validation testing of SleepStageML with results supporting the determination of substantial equivalence. Activities that were conducted included the following:

. Software verification testing which included software code reviews, automated testing, acceptance testing, labeling review, cybersecurity and data protection, which confirmed that all software requirements are developed as expected.
. Design validation testing which simulated the intended use to confirm that the endto-end functionality of SleepStageML meets the design requirements.
. Study that utilized clinical data to demonstrate automatic scoring sleep study performance.
. Design traceability confirming that all requirement tracing is complete from design inputs to verification/validation and that all risk controls are implemented and effective.

Non-Clinical Tests

Support for the substantial equivalence of SleepStageML was provided by requirements verification, risk management and software testing, at the unit, integration, and system levels. Unit and integration tests verify the functionality for individual software parts; and system-level integration tests cover each specified requirement. System level tests were defined with detailed protocols and objective pass/fail criteria and were clearly documented with test execution results for review. SleepStageML passed all unit, integration, and system level testing demonstrating that all requirements were verified and validated.

Clinical Validation Test

The efficacy of SleepStageML was established through a retrospective pivotal validation study using previously collected clinical polysomnography (PSG) recordings from a representative set of 100 patients. Recordings for inclusion in the validation data set were randomly selected from three level 1 clinical PSG data sources and were required to (1) have all EEG channels recommended by the American Academy of Sleep Medicine (AASM), (2) have EEG sampling rates between 128 and 512Hz, (3) have a recording duration between 4 and 24 hours, and (4) be from patients of least 18 years of age, The included patients spanned a variety of ages (between 19 and 83), were balanced across sex (46 female and 54 male) and were stratified across clinical apnea categories (25 patients in each category, normal (AHI < 5), mild (5 ≤ AHI < 15), moderate (15 ≤ AHI < 30), and severe (AHI > 30)).

Each PSG recording was manually and independently sleep staged by three (3) reqistered PSG technologists (RPSGTs) each with at least 5 years of experience in clinical scoring of

{8}------------------------------------------------

Image /page/8/Picture/0 description: The image shows the text "510(K) SUMMARY" in bold, black font. The text is centered and appears to be the title or heading of a document. The letters are capitalized and evenly spaced.

Image /page/8/Picture/2 description: The image shows the logo for Beacon Biosignals. The logo has the word "BEACON" in bold, dark gray letters above the word "BIOSIGNALS" in smaller, lighter gray letters. To the right of the words is a stylized image of a brain in a light teal color, represented by a series of connected curved lines.

sleep studies. To ensure that the sleep staging outputs obtained from the experts were of sufficient quality, several data-quality and consistency checks were performed. SleepStageML software performance was evaluated against the expert consensus sleep stages that were constructed using 2/3 majority scoring (i.e., the stage per epoch where at least 2 of the 3 experts agree).

Objective performance targets and acceptance criteria of positive agreement (PA), negative agreement (NA), and overall agreement (OA) were predefined to demonstrate non-inferiority to the identified predicate device, Sleep Profiler (K153412). These metrics were computed for all 30-second epochs that had an expert consensus sleep stage (excluding epochs where all 3 RPSGTs disagreed) after pooling across all recordings in the study.

Table 1 shows the SleepStageML clinical performance results as compared to the predicate reported results. Comparing the 95% bootstrapped confidence intervals and mean values to the reference values indicates that SleepStageML is non-inferior to the predicate device. Therefore, the results confirm that the positive, negative, and overall agreement performances of SleepStageML are substantially equivalent to those of Sleep Profiler (K153412), respectively.

Sleep StagingComparisons		N=100subjects	Percent Agreement (%) with 95%percentile bootstrap confidenceinterval (N=5000 resamples)			N=43subjects	Point-estimate of PercentAgreement (%)
		TotalEpochs	OverallAgreement(OA)	PositiveAgreement(PA)	NegativeAgreement(NA)	TotalEpochs	OverallAgreement(OA)	PositiveAgreement(PA)	NegativeAgreement(NA)
Overall epochs using 2/3 consensus scoring	W	21,668	96.1%(95.4%,96.8%)	88.9%(86.5%,91.2%)	98.5%(98.2%,98.8%)	7,424	89%	73%	94%
	N1	3,877	94.5%(93.7%,95.2%)	58.4%(54.2%,62.4%)	96.2%(95.4%,96.9%)	1,752	89%	25%	93%
	N2	42,587	87.1%(85.9%,88.3%)	79.8%(77.7%,81.8%)	94.2%(93.2%,95.0%)	12,582	81%	77%	84%
	N3	6,210	92.9%(91.8%,93.8%)	93.0%(89.8%,95.7%)	92.9%(91.7%,93.9%)	4,704	91%	76%	94%
	R	12,641	97.3%(96.7%,97.9%)	93.1%(91.5%,94.5%)	98.0%(97.3%,98.6%)	3,749	95%	74%	97%
	Total	86,983	Multi-stageAgreement:84.02%	--	--	31,361	--	--	--
	NoConsensus	2,289	--	--	--	1,150	--	--	--

Table 1: SleepStageML Clinical Performance Results

{9}------------------------------------------------

Image /page/9/Picture/2 description: The image shows the logo for Beacon Biosignals. The logo consists of the word "BEACON" in bold, dark gray letters stacked on top of the word "BIOSIGNALS" in a smaller, lighter gray font. To the right of the text is a stylized graphic of a brain, represented by a series of connected, rounded lines in a light teal color. The brain graphic is positioned so that it appears to be emanating from the text.

Conclusion

Non-Clinical verification, validation, and clinical validation testing were conducted in accordance with FDA guidance recommendations to confirm the device design meets all specifications and user needs. SleepStageML has passed all software verification and validation tests and its clinical validation testing results demonstrated effective performance and non-inferiority to predicate device's performance. It is therefore concluded that SleepStageML is substantially equivalent to the predicate device.

Predetermined Change Control Plan

SleepStageML includes an authorized Predetermined Change Control Plan (PCCP) that allows for planned updates of the machine learning software device function (ML-DSF) and non-ML algorithmic components to improve sleep staging performance within the existing intended use and indications for use. This PCCP allows for the modification of the algorithmic components of SleepStageML including the signal preprocessing, machine learning model, postprocessing, or signal quality check to achieve increased sleep staging performance. The four modifications are summarized in Table 2 below.

#	Modification	Description
1	Update of MachineLearning Model	SleepStageML's sleep staging neural network may be modified for thepurposes of improving sleep staging performance within the intendeduse population by:Retraining with an updated training/tuning dataset Retraining with updated hyper-parameters, loss function,optimizer Retraining with updated model selection criteria Retraining with an updated neural network architecture withlimitations on model size and type
2	Update of SignalPreprocessingSteps	SleepStageML's EEG signal preprocessing may be modified for thepurposes of improving sleep staging performance within the intendeduse population by:Updating the parameters of the digital signal processing steps(e.g., filtering) applied to the EEG signals before being input tothe machine learning model
3	Update ofProbabilityPostprocessing	SleepStageML's probability postprocessing may be modified for thepurposes of improving sleep staging performance within the intendeduse population by:Updating the methods by which sleep stages are generatedfrom the model output sleep stage probabilities
4	Update of SignalQuality Check	SleepStageML's signal quality check component may be modified forthe purposes of improving sleep staging performance within theintended use population by:Updating the criteria/thresholds used to check that the inputEEG signals are analyzable

Table 2: Summary of PCCP Modifications

Modifications 1 and 2 above would trigger re-training of the machine learning model, while modifications 3 and 4 would not trigger re-training of the machine learning model.

The testing of any modification to SleepStageML within the scope of the PCCP will include comprehensive software verification and validation testing, including repeating all unit, integration, and system level tests performed in the development for the original

{10}------------------------------------------------

Image /page/10/Picture/2 description: The image shows the logo for Beacon Biosignals. The logo has the word "BEACON" in bold, dark gray letters on top of the word "BIOSIGNALS" in smaller, light gray letters. To the right of the words is a light green image that looks like a brainwave.

SleepStageML software. All software verification tests linked to requirements and specifications must pass for a modification to be considered valid. In addition, clinical performance validation will also be repeated and will require that the performance of any modification to SleepStageML to be non-inferior in per-stage performance metrics to both the original SleepStageML device and its predicate device. In addition, the performance of any modification to SleepStageML must also be non-inferior to the best performance among released versions of SleepStageML with respect to the multi-stage agreement. The acceptance criteria for the clinical performance validation are summarized in the Table 3 below, as well as the modifications for which each criterion is required to pass.

Acceptance Criteria	Modifications thatrequire passingthis criterion
The per-stage overall, positive, and negative agreements of the modifiedSleepStageML, as compared to human expert consensus, is non-inferior toboth the Original SleepStageML and the predicate device for full overnightrecordings.	1, 2, 3, 4
The percentage of human scorable recordings that are unanalyzable by theupdated SleepStageML is less than or equal to a predefined threshold.	4
The per-stage overall, positive, and negative agreements of the modifiedSleepStageML, as compared to human expert consensus, is non-inferior toboth the Original SleepStageML and the predicate device for 2-hourrecording segments.	1, 2, 3, 4
The per-stage overall, positive, and negative agreements of the modifiedSleepStageML, as compared to human expert consensus, is non-inferior toboth the Original SleepStageML and the predicate device for recordings withthe minimum number of AASM recommended EEG channels.	1, 2, 3, 4
The overall multi-stage agreement (across all 5 sleep stages) of the modifiedSleepStageML is non-inferior to the highest overall multi-stage agreementacross all released versions of SleepStageML.	1, 2, 3, 4

TABLE 3: SUMMARY OF PCCP CLINICAL VALIDATION ACCEPTANCE CRITERIA

Upon a release of an updated version of SleepStageML based on this PCCP, communication will be sent to all external clinical users of SleepStageML, informing them that a new version of SleepStageML is available, with a description of the release and its updated performance.

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).