Learn · Sleep

Cardiopulmonary coupling (CPC): how heart and breath reveal sleep stability

Sleep stages say which kind of sleep you were in. Cardiopulmonary coupling asks a different question: was that sleep stable? It reads the answer from how heart rate and breathing move together, using one ECG or PPG signal.

What is cardiopulmonary coupling?

Cardiopulmonary coupling (CPC) is a sleep measure built from how closely your heart rate follows your breathing. In deep, stable NREM sleep, the heart speeds up a little on each breath in and slows on each breath out, and the two rhythms lock together at the breathing rate. When sleep turns unstable, breathing and heart rate start rising and falling together over much slower cycles, often 20 to 60 seconds long. CPC measures which of those patterns dominates, minute by minute, from a single ECG or PPG recording (Thomas et al., 2005; Al Ashry, Ni & Thomas, 2021).

That gives a different picture from the stage labels on a sleep report. Standard sleep staging calls each 30 seconds wake, N1, N2, N3 or REM. CPC splits NREM sleep into two states, stable and unstable, that cut across N1, N2 and N3, and it lumps REM and wake into a third state. Its authors describe the result as a bimodal view of NREM sleep rather than a graded one (Wood et al., 2020).

CPC is not new. The method was published in 2005, has FDA-cleared software behind it, and has been used in well over a hundred papers. It is still little known outside sleep research, partly because it does not fit the stage-and-event vocabulary that sleep reports are written in. This guide covers the method, the history, the evidence, every vendor we could verify, and where it falls short.

How CPC sleep analysis works

CPC needs two time series from one sensor. The first is the gap between heartbeats. From an ECG, that is the R-to-R interval; after removing ectopic beats and outliers you are left with normal-to-normal (NN) intervals, the same series behind heart rate variability and RMSSD. From a PPG, it is the pulse-to-pulse interval, close kin to the interbeat interval (and not quite the same thing, as pulse rate variability vs HRV explains).

The second series is breathing, taken from the same recording. On an ECG, each breath shifts the heart's electrical axis relative to the electrodes, so QRS amplitude swings with respiration. That swing is ECG-derived respiration (EDR) (Wood et al., 2020). SleepImage's instructions describe the same idea using R-S amplitude and QRS area changes, and its PPG version reads breathing and tidal-volume changes from the pleth waveform (SleepImage IFU).

The two series are resampled to 2 Hz and cut into 8.5-minute windows of 1,024 samples. Within each window the method computes the cross-spectral power, which says how much the two signals vary together at each frequency, and the coherence, which says how consistently their phases line up. Multiplying the two gives coherent cross-power. Stepping the window across the night produces the sleep spectrogram (Thomas et al., 2005).

  1. Record one signal

    A single-lead ECG, or a PPG (pleth) from a finger or wrist sensor. CPC was built on the ECG; later systems added PPG.
  2. Find every beat

    Detect each R-peak (ECG) or pulse (PPG), drop ectopic beats and outliers, keep normal-to-normal (NN) intervals.
  3. Pull out breathing

    Breathing shifts the heart's electrical axis, so QRS amplitude rises and falls with each breath: ECG-derived respiration (EDR). PPG gives a pulse-derived equivalent.
  4. Resample and window

    Put both series on a 2 Hz grid and cut them into 1,024-sample (8.5-minute) windows that step forward through the night.
  5. Cross-spectrum × coherence

    For each window, compute how much power the two series share at each frequency, and how phase-locked they are. Their product is coherent cross-power.
  6. Label the window

    Power at the breathing rate (0.1 to 0.4 Hz) is HFC, stable NREM. Power at 0.01 to 0.1 Hz is LFC, unstable NREM; strong, narrow LFC is e-LFC. Below 0.01 Hz is VLFC, REM or wake.
The CPC pipeline, step by step. Adapted from Thomas et al. (2005) and the band limits in Schramm et al. (2013). Commercial systems add their own artifact handling and thresholds.

Try it: score one CPC window

The widget runs that maths in your browser on simulated signals. Pick a state, then move the breathing rate: in stable sleep the shared power follows your breathing frequency. Add noise and watch how much the call can tolerate.

Illustrative simulation

1. Beat-to-beat (NN) intervals, ms

2. Respiration surrogate (EDR or PDR)

First 3 minutes of an 8.5-minute window, resampled at 2 Hz.

3. Coherent cross-power (coherence × cross-spectrum)

VLFLFHF00.10.20.30.4Hz
High-frequency coupling (stable NREM)
HF100%
LF0%
VLF0%
How one CPC window is scored. The signals are simulated, not recorded. The maths is the published outline: two series at 2 Hz, three overlapping 512-sample segments, cross-spectral power multiplied by coherence. Stable sleep puts the shared power at the breathing rate (HF, 0.1 to 0.4 Hz); cycling breathing and heart rate every ~35 s pushes it into LF (0.01 to 0.1 Hz); REM and wake drift into VLF. Real systems add their own thresholds and artifact rules, so this simple "largest band wins" call is for teaching only.

HFC, LFC, VLFC and e-LFC: the CPC sleep states

VLFCREM or wakeLFCUnstable NREM (e-LFC when strong)HFCStable NREM0.0039 Hzcycle 4.3 min0.01 Hzcycle 1.7 min0.1 Hzcycle 10 s0.4 Hzcycle 3 sslower cyclesfaster (breath-by-breath)
The three CPC bands on one axis. HFC sits at normal breathing rates (about 6 to 24 breaths a minute). LFC covers the 10-second to 100-second cycling of unstable sleep and repeated apneas. VLFC covers drifts of roughly 1.7 to 4.3 minutes. Band edges from Thomas et al. (2005) and Schramm et al. (2013); log scale.

High-frequency coupling (HFC), 0.1 to 0.4 Hz. Coupled power at the breathing rate. It marks stable NREM sleep: regular breath-to-breath breathing, strong respiratory sinus arrhythmia, few EEG arousals. HFC is the CPC biomarker of stable sleep (Schramm et al., 2013; Wood et al., 2020). In 3,150 Sleep Heart Health Study participants without sleep apnea, HFC power tracked EEG delta power across the whole night, inside and outside periods scored as N3 (r = 0.40) (Thomas et al., 2014).

Low-frequency coupling (LFC), 0.01 to 0.1 Hz. Coupled power in slow cycles. It marks unstable NREM sleep: tidal volume waxing and waning, cyclic variation in heart rate, frequent arousals and fragmentation (Schramm et al., 2013; Wood et al., 2020). LFC lines up better with the cyclic alternating pattern (CAP), the EEG signature of unstable NREM, than with stages (Thomas et al., 2005; Terzano et al., 2001).

Elevated LFC (e-LFC). The part of LFC with unusually high coherent power. It goes with fragmented sleep from many causes, including sleep apnea (Schramm et al., 2013). When that power sits in a narrow frequency band, it is called narrow-band e-LFC, and it points toward central apnea and periodic breathing rather than purely obstructive events (Thomas et al., 2007).

Very-low-frequency coupling (VLFC), 0.0039 to 0.01 Hz. Seen in REM sleep and wake, when breathing becomes irregular and heart rate drifts. CPC alone does not cleanly separate REM from wake; commercial systems use signal quality and motion to tell them apart (Al Ashry, Ni & Thomas, 2021).

Two summary numbers come out of these bands. The HFC/LFC ratio was used as a sleep quality ratio in early studies; a 2012 case report quoted expected adult values of 1.67 to 4.0 (Schramm & Thomas, 2012). SleepImage reports a proprietary sleep quality index (SQI) instead. Neither has an AASM-endorsed normal range.

The CPC sleep spectrogram vs the hypnogram

A hypnogram is a staircase: one stage per 30-second epoch. A CPC sleep spectrogram is a heat map: time across, frequency up, color for coherent heart-breathing power. Read side by side, two nights can have nearly the same hypnogram while one spends its N2 in stable coupling and the other in unstable, apnea-like cycling.

Illustrative, simulated data · not patient data

Night A: consolidated NREM

HFC 62%LFC 6%(e-LFC 0%)VLFC 32%
WRN1N2N300.10.20.30.4Frequency (Hz)HFLFVLF0h1h2h3h4h5h

Night B: same stages, unstable breathing (sleep-apnea-like)

HFC 23%LFC 46%(e-LFC 34%)VLFC 32%
WRN1N2N300.10.20.30.4Frequency (Hz)HFLFVLF0h1h2h3h4h5h
HFC, stable NREM (0.1 to 0.4 Hz)LFC, unstable NREM (0.01 to 0.1 Hz); narrow, low bands mark e-LFCVLFC, REM or wake (below 0.01 Hz)
Hypnogram vs CPC spectrogram. Top line in each panel: sleep stages scored from brain waves in 30-second epochs. Heat map: where heart rate and breathing share power, window by window. Both nights were drawn by us to show the idea; the stage plots are almost the same, yet night B spends most of its N2 in unstable coupling. That gap is what CPC adds to staging (Thomas et al., 2005; Wood et al., 2020).

In real data, healthy adults switch spontaneously between HFC and LFC within a single NREM period, and the switches happen across the whole night (Thomas et al., 2005). Wood and colleagues tied the two NREM states to wider physiology: stable NREM brings a steady slow oscillation, high delta power, blood pressure dipping and strong vagal tone; unstable NREM brings a fragmented slow oscillation, non-dipping blood pressure and cyclic heart rate (Wood et al., 2020). For how stage percentages usually look across a night, see the sleep stages chart.

CPC vs sleep staging: PSG, home tests and wearables

Sleep staging and CPC answer different questions, so grading CPC by how well it reproduces N1, N2 and N3 misses its purpose. In the 2005 paper, CPC states agreed poorly with standard stages (62.7% in the training set, 43.9% in the test set) and better with CAP scoring (74% and 77.3%) (Thomas et al., 2005). Staging itself is not perfectly repeatable: in the AASM's inter-scorer program, scorers agreed with the reference on 82.6% of epochs (Rosenberg & Van Hout, 2013).

How CPC compares with polysomnography (PSG) staging and consumer wearable staging.

How CPC compares with polysomnography (PSG) staging and consumer wearable staging.
DimensionPSG staging (AASM)Cardiopulmonary couplingConsumer wearable staging
SignalsEEG, EOG, chin EMG, plus breathing, oxygen, ECG and legsOne ECG or PPG (beat timing plus a derived breathing signal); oximetry optionalWrist or finger PPG and accelerometer, sometimes temperature
OutputW, N1, N2, N3, REM per 30-s epochStable NREM (HFC), unstable NREM (LFC, e-LFC), REM/wake (VLFC); some systems add an AHILight, deep, REM and awake estimates
Time scale30-second epochs8.5-minute windows stepped across the night in the original method; newer methods are finerUsually 30-second epochs, vendor-specific
What it captures bestBrain state; the reference for stagesSleep stability and breathing control; how sleep and autonomic state interactSleep timing and rough stage proportions
ScoringHuman scorers or validated software against AASM rulesFully automatedProprietary algorithms
WhereSleep lab (Type I) or home PSGHome or lab, one sensor, many nightsHome, every night
RegulatoryStandard clinical testFDA-cleared software exists (e.g. K182618, K233501)Mostly wellness features

Consumer staging is a different kind of estimate. A ring or watch that labels "deep sleep" is trying to imitate EEG stages from pulse and motion, and has to be judged against PSG; the PPG sleep staging algorithms post and the sleep staging algorithm page cover how well that works. CPC makes no claim about brain waves; it reports a property that heart and breathing signals carry directly. For the rules behind the stages themselves, see sleep staging explained.

Where CPC came from: a short history

CPC came out of the Beth Israel Deaconess group that also built PhysioNet's complex-signal tools. It grew from two older ideas: that NREM sleep alternates between stable and unstable phases visible in the EEG as the cyclic alternating pattern, and that a breathing signal can be recovered from an ECG.

  1. 2001Terzano and colleagues publish the scoring atlas for the cyclic alternating pattern (CAP), the EEG rhythm of sleep instability that CPC would later track (Terzano et al., 2001).
  2. 2005Robert Thomas, Joseph Mietus, Chung-Kang Peng and Ary Goldberger at Beth Israel Deaconess Medical Center in Boston describe the ECG sleep spectrogram in Sleep, tested on 70 polysomnograms (Thomas et al., 2005).
  3. 2007Narrow-band e-LFC is used to separate obstructive from central and complex apnea, starting from PhysioNet's Apnea-ECG recordings (Thomas et al., 2007).
  4. 2008DynaDx's Sleep Quality software, which estimates AHI from RR intervals and ECG-derived respiration, is cleared (K070855 (FDA)).
  5. 2009In 5,247 Sleep Heart Health Study recordings, narrow-band e-LFC is linked to prevalent hypertension and stroke (Thomas et al., 2009). MyCardio and Embla clear the M1 single-lead ECG recorder with CPC software (K092003 (FDA)).
  6. 2010 to 2013CPC is applied to family heritability of sleep apnea (Ibrahim et al., 2010), depression (Yang et al., 2011) and insomnia (Schramm et al., 2013).
  7. 2014HFC tracks EEG delta power in 3,150 SHHS participants without sleep apnea (Thomas et al., 2014).
  8. 2017 to 2019The SleepImage System is cleared for ECG (K163696 (FDA)), then for ECG or pleth in children and adults with an AHI (K182618 (FDA)). Huawei starts using CPC in TruSleep.
  9. 2020 to 2021PPG-based CPC AHI is tested in children from the CHAT trial (Hilmisson et al., 2020) and CPC plus oximetry in 833 adults from APPLES (Al Ashry et al., 2021).
  10. 2023 to 2026An independent Charité study compares the SleepImage ring with PSG (Lu et al., 2023); Largan's ECG system is cleared (K233501 (FDA)) and tested in 479 patients (Sun et al., 2026); a one-million-night CPC analysis is published (Hilmisson, Magnusdottir & Thomas, 2026).
CPC milestones. FDA numbers link to our FDA record pages and to FDA's database.

The US patent "System and method for assessing sleep quality" (US 8,401,626) names Mietus, Peng, Thomas and Goldberger as inventors (Google Patents). By 2009, CPC was a module in Embla's polysomnography software, and MyCardio and Embla had built a dedicated single-lead recorder for it (K092003 (FDA)). Papers by SleepImage staff state that MyCardio, SleepImage's parent company, licenses the CPC algorithms from Beth Israel Deaconess, and Thomas discloses royalties from that license (Hilmisson, Magnusdottir & Thomas, 2026; Hilmisson, Thomas & Magnusdottir, 2024). That matters when reading the evidence below.

Why CPC is useful

It measures stability as well as stage. A night can have normal stage percentages and still be fragmented. CPC gives a direct number for how much NREM sleep was stable, which tracks EEG slow-wave activity (Thomas et al., 2014) and maps onto known features of good and poor sleep (Wood et al., 2020).

It works from one sensor, for many nights. Because it needs only an ECG or PPG, CPC can be recorded at home night after night. In a multi-night study with the M1 recorder (10 healthy people, 18 with treated sleep apnea, 20 with insomnia; 314 nights in all), HFC was consistent from night to night in healthy people and in sleep apnea (intraclass correlation 0.83 and 0.89) but much less so in insomnia (0.66). Healthy participants got about 33 more minutes of HFC on weekends than on weekdays (275.8 vs 242.8 minutes) (Thomas et al., 2018).

Insomnia and mental health. People with primary insomnia had less HFC, more LFC and e-LFC, and a lower HFC/LFC ratio than good sleepers across two lab nights (Schramm et al., 2013). Unmedicated people with major depression had less HFC and more LFC and VLFC than controls, and medicated patients looked closer to controls (Yang et al., 2011).

Treatment response. During CPAP titration, HFC rose and LFC and e-LFC fell as the AHI came down (Cho & Kim, 2017). In a single-patient, ten-night case study, CPC distinguished an oral appliance from positional therapy and oxygen for mild sleep apnea (Schramm & Thomas, 2012). In the HeartBEAT trial, people whose CPC sleep quality index improved also had larger blood pressure improvements, mostly with CPAP (Magnusdottir et al., 2020).

Cardiovascular and developmental signals. Narrow-band e-LFC was associated with prevalent hypertension and stroke in 5,247 SHHS participants after adjustment (Thomas et al., 2009). About 30% of the variation in e-LFC was explained by family factors in 622 people from 137 families (Ibrahim et al., 2010). In 363 children and young adults aged 4 to 22, stable sleep fell from 70% to 48% of estimated sleep time with age while unstable sleep rose (Cysarz et al., 2018). A one-million-night analysis of SleepImage home recordings found the CPC sleep quality index declined much less with age than EEG-defined N3 usually does, and was higher in women in every age group (Hilmisson, Magnusdottir & Thomas, 2026).

These are associations, many from groups with a commercial interest in CPC. They show what CPC tracks; they do not show that improving a CPC number improves health.

CPC sleep studies for sleep apnea

Sleep apnea is where CPC has moved furthest into clinics. Repeated breathing events drive exactly the slow, coupled cycling CPC sees as LFC, and the shape of that cycling carries extra information. Broad-band e-LFC fits obstructive events; narrow-band e-LFC fits the regular periodic breathing of central or complex apnea, and predicted which patients developed central apneas when put on CPAP (Thomas et al., 2007). That is a phenotype a plain AHI cannot give. See central sleep apnea for the clinical picture.

Combined with oximetry, CPC can also produce an AHI. Published comparisons against manually scored polysomnography include:

  • 833 adults from the APPLES trial, ECG plus oximetry: correlation with PSG AHI of 0.972 (Al Ashry et al., 2021).
  • Children from the CHAT trial, PPG plus oximetry only: correlation 0.954 with PSG AHI, sensitivity 95.4% for mild OSA (Hilmisson et al., 2020).
  • 68 adult ECGs, ECG only: sensitivity 100% and specificity 81% for moderate to severe sleep apnea (Hilmisson et al., 2019).
  • 39 adults at Charité Berlin with the SleepImage ring worn during PSG: correlation 0.89 and areas under the curve of 1.00, 0.90 and 0.98 at AHI cut-offs of 5, 15 and 30 (Lu et al., 2023).
  • 479 adults at three Taiwanese centers with Largan's chest ECG device: correlation 0.889 and AUCs of 0.92 to 0.97, higher with added oximetry (Sun et al., 2026).

The first three were written with SleepImage staff and the fifth was funded by Largan; the Charité study had no vendor authors. CPC AHIs count hypopneas using oxygen dips, so the scoring rule matters (see hypopnea). For where home tests fit next to lab PSG, see home sleep apnea test vs polysomnography. For broader work on wearable detection, see PPG sleep apnea detection and PPG for sleep apnea. An expert review lists CPC among candidate severity metrics beyond the AHI, alongside hypoxic burden and arousal intensity (Malhotra et al., 2021).

CPC vendors and devices

We searched FDA's 510(k) database, vendor documentation, SEC filings and the CPC literature for every product that uses cardiopulmonary coupling. Five came up. Each FDA number below was checked in FDA's database on 2026-11-16; where a vendor claim could not be checked, the row says so. Filter by signal or regulatory status; the filters live in the address bar, so Copy link shares the view.

SignalUse5 of 5
Vendors and devices that use cardiopulmonary coupling: signal, form factor, outputs, FDA status and evidence
ProductSignalForm factorWhat it reportsFDAPeer-reviewed evidence
SleepImage SystemMyCardio LLC (SleepImage) · Clinical (prescription)SleepImage clinical IFU (PDF) · K182618 summary · Vivos 8-K, Feb 2021ECG, PPG (pleth)Cloud software. SleepImage sells finger-worn ring and fingertip PPG/oximetry recorders; the software also accepts ECG.Stable NREM, unstable NREM and REM, a sleep quality index (SQI), sleep timing, and with oximetry an apnea-hypopnea index (sAHI), central and obstructive sAHI, RDI, ODI and hypoxic burden (vendor instructions for use). Of the vendors we found, the only one whose FDA clearance names a pleth (PPG) signal as an input. Vivos Therapeutics sells it to dentists as VivoScore, "powered by SleepImage", under a 2021 agreement (SEC 8-K). The company states that its CPC algorithms are licensed from Beth Israel Deaconess Medical Center.FDA clearedK182618 (2019, FDA): SleepImage System: ECG or pleth input, children, adolescents and adults; sAHI when oximetry is present. Prescription use.K163696 (2017, FDA): SleepImage System: ECG input, software only.K092003 (2009, FDA): M1 Sleep Data Recorder and CPC Application Software, developed with Embla; single-lead ECG recorder.Hilmisson et al., 2019; Hilmisson et al., 2020; Al Ashry et al., 2021; Lu et al., 2023; Thomas et al., 2018
DOMII (LARGAN Health AI-Tech Sleep Apnea and Sleep Quality Examination System, ST-SLP03A)Largan Health AI-Tech / Largan Medical (Taiwan) · Clinical (prescription)Largan Health sleep system page · Largan Health launch noteECGChest-worn wireless ECG recorder; the vendor's "Ultimate Edition" adds a finger oximeter (cot or clip).Vendor lists AHI and apnea type, sleep structure and quality, ODI, heart rate and HRV. In the 2026 multicenter study the device reported a CPC-derived respiratory event index. Largan says DOMII builds on the CPC approach from Harvard-affiliated research. The 479-patient validation study was funded by Largan Health AI-Tech. Also registered with Taiwan FDA (TFDA 007003).FDA clearedK233501 (2024, FDA): Product code MNR (ventilatory effort recorder), cleared 26 July 2024.FDA's summary PDF for K233501 was not available when we checked, so we have not verified its indications or Rx/OTC status.Sun et al., 2026
Sleep Quality (apnea examination software)DynaDx Corporation · Clinical (prescription)K070855 summaryECGSoftware that reads single-lead ECG exported from a Holter recorder. A companion chest recorder, the EZsleep Sleep Quality Recorder (Platinum Team, K112573), was cleared for use with it.Detects episodes of sleep-disordered breathing and estimates AHI from RR intervals and ECG-derived respiration (510(k) summary). Adults only, as a screening device. A 2011 disclosure lists financial interests in DynaDx for two of the CPC co-inventors (Mietus and Peng).FDA clearedK070855 (2008, FDA): Sleep Quality software: adults, screening for the need for polysomnography.K112573 (2012, FDA): EZsleep Sleep Quality Recorder (Platinum Team Co.): chest ECG recorder for use with K070855.The 510(k) summary describes heart rate and respiratory interaction analysis but does not use the term CPC. We found no current product page.Bianchi et al., 2011
RemLogic CPC moduleEmbla Systems (now Natus) · Clinical (prescription)K092003 summaryECGAnalysis module inside polysomnography software; uses the ECG channel of a lab or home study.Sleep spectrogram with high, low and very-low-frequency coupling from the PSG's ECG. The 2009 M1 submission (K092003) says CPC was then a module in Embla's PSG software. Researchers at Beth Israel Deaconess used RemLogic 1.1 for CPC analysis in a 2020 paper.CPC not named in FDA recordK162140 (2016, FDA): RemLogic is cleared as PSG software; the summary does not mention CPC.We could not confirm whether the CPC module ships in current RemLogic versions.Wood et al., 2020
HUAWEI TruSleep (versions 1.0 to 4.0)Huawei · Consumer wellnessHuawei: sleep monitoring principlesPPG (pleth)Watch and band sleep tracking (wrist optical heart rate plus motion).Light, deep and REM sleep and a sleep score. Huawei says TruSleep 1.0 to 4.0 (2017 to 2025) applied the cyclic alternating pattern standard and CPC to judge sleep stability from heart rate and respiratory coherence. TruSleep 5.0 (2026) moves to AASM-style stages and adds a "stable sleep" indicator. The only consumer wearable maker we found that documents CPC in its own sleep algorithm.No FDA clearance foundA consumer wellness feature. We found no FDA clearance for it.No independent peer-reviewed validation found
  • SleepImage System

    MyCardio LLC (SleepImage) · ECG, PPG (pleth)

    Form: Cloud software. SleepImage sells finger-worn ring and fingertip PPG/oximetry recorders; the software also accepts ECG.

    Reports: Stable NREM, unstable NREM and REM, a sleep quality index (SQI), sleep timing, and with oximetry an apnea-hypopnea index (sAHI), central and obstructive sAHI, RDI, ODI and hypoxic burden (vendor instructions for use).

    FDA clearedK182618 (2019, FDA): SleepImage System: ECG or pleth input, children, adolescents and adults; sAHI when oximetry is present. Prescription use.K163696 (2017, FDA): SleepImage System: ECG input, software only.K092003 (2009, FDA): M1 Sleep Data Recorder and CPC Application Software, developed with Embla; single-lead ECG recorder.

    Evidence: Hilmisson et al., 2019; Hilmisson et al., 2020; Al Ashry et al., 2021; Lu et al., 2023; Thomas et al., 2018

  • DOMII (LARGAN Health AI-Tech Sleep Apnea and Sleep Quality Examination System, ST-SLP03A)

    Largan Health AI-Tech / Largan Medical (Taiwan) · ECG

    Form: Chest-worn wireless ECG recorder; the vendor's "Ultimate Edition" adds a finger oximeter (cot or clip).

    Reports: Vendor lists AHI and apnea type, sleep structure and quality, ODI, heart rate and HRV. In the 2026 multicenter study the device reported a CPC-derived respiratory event index.

    FDA clearedK233501 (2024, FDA): Product code MNR (ventilatory effort recorder), cleared 26 July 2024.FDA's summary PDF for K233501 was not available when we checked, so we have not verified its indications or Rx/OTC status.

    Evidence: Sun et al., 2026

  • Sleep Quality (apnea examination software)

    DynaDx Corporation · ECG

    Form: Software that reads single-lead ECG exported from a Holter recorder. A companion chest recorder, the EZsleep Sleep Quality Recorder (Platinum Team, K112573), was cleared for use with it.

    Reports: Detects episodes of sleep-disordered breathing and estimates AHI from RR intervals and ECG-derived respiration (510(k) summary). Adults only, as a screening device.

    FDA clearedK070855 (2008, FDA): Sleep Quality software: adults, screening for the need for polysomnography.K112573 (2012, FDA): EZsleep Sleep Quality Recorder (Platinum Team Co.): chest ECG recorder for use with K070855.The 510(k) summary describes heart rate and respiratory interaction analysis but does not use the term CPC. We found no current product page.

    Evidence: Bianchi et al., 2011

  • RemLogic CPC module

    Embla Systems (now Natus) · ECG

    Form: Analysis module inside polysomnography software; uses the ECG channel of a lab or home study.

    Reports: Sleep spectrogram with high, low and very-low-frequency coupling from the PSG's ECG.

    CPC not named in FDA recordK162140 (2016, FDA): RemLogic is cleared as PSG software; the summary does not mention CPC.We could not confirm whether the CPC module ships in current RemLogic versions.

    Evidence: Wood et al., 2020

  • HUAWEI TruSleep (versions 1.0 to 4.0)

    Huawei · PPG (pleth)

    Form: Watch and band sleep tracking (wrist optical heart rate plus motion).

    Reports: Light, deep and REM sleep and a sleep score. Huawei says TruSleep 1.0 to 4.0 (2017 to 2025) applied the cyclic alternating pattern standard and CPC to judge sleep stability from heart rate and respiratory coherence. TruSleep 5.0 (2026) moves to AASM-style stages and adds a "stable sleep" indicator.

    No FDA clearance foundA consumer wellness feature. We found no FDA clearance for it.

    Evidence: No independent peer-reviewed validation found

Some clinics and dental practices offer CPC testing through these vendors rather than selling their own device. Vivos Therapeutics markets SleepImage to dentists as VivoScore (SEC filing). We did not list products that only mention heart-rate variability or "sleep stability" without naming CPC.

CPC from PPG: can a ring or watch do it?

CPC started on the ECG, but nothing in the maths requires electrodes. It needs accurate beat-to-beat timing and a breathing surrogate. A finger PPG gives both: pulse intervals, and breathing-driven changes in pulse amplitude, baseline and timing (see how PPG estimates respiratory rate and the respiratory rate algorithm page). SleepImage's clearance names pleth input, and its pediatric validation used PPG and oximetry alone (Hilmisson et al., 2020). The CPC review authors expect wide use in wearable and non-contact devices for that reason (Al Ashry, Ni & Thomas, 2021).

The wrist is harder than the finger. Pulse timing from wrist PPG is noisier, motion corrupts it, and pulse intervals are not identical to ECG RR intervals. Most consumer devices also hide the raw waveform and expose only nightly summaries, which are useless for CPC. Researchers building their own pipeline need devices that export unfiltered PPG; some research wearables do, for example Sensor Bio's V2 wristband, which offers raw multi-wavelength PPG and accelerometer data (Sensor Bio raw data). The PPG sleep fragmentation post covers other pulse-based markers of broken sleep.

Newer signal processing also helps. A Hilbert-Huang version of CPC cut the time resolution from 8.5 minutes to 8 seconds (Liu et al., 2012), and a synchrosqueezing version raised per-minute apnea detection on PhysioNet data from 72% to 83% when combined with standard heart and breathing features (Wang, Shi & Yeh, 2023).

Limitations and evidence gaps

  • It needs sinus rhythm. CPC is built on normal-to-normal intervals. FDA labeling for the DynaDx software asks for recordings that are predominantly sinus rhythm (K070855 (FDA)), so atrial fibrillation, frequent ectopy and paced rhythms are a problem.
  • Coarse time steps. The original windows are 8.5 minutes long, far coarser than 30-second staging; finer methods exist but are less validated (Liu et al., 2012).
  • REM and wake look alike. Both land in VLFC; telling them apart depends on motion and signal-quality rules (Al Ashry, Ni & Thomas, 2021).
  • Who did the studies. Much of the evidence comes from the inventors or from SleepImage and Largan staff or funding. Independent validations exist (Lu et al., 2023; Cysarz et al., 2018), but they are smaller.
  • No standard normal range. CPC is not part of the AASM scoring rules, and vendors use their own indices, so numbers from different systems are not interchangeable.
  • Associations, not outcomes. Links with hypertension, depression or insomnia are cross-sectional or small. We found no trial showing that targeting a CPC metric improves health outcomes.

This guide is educational. A CPC report, like any sleep test, should be read by a clinician, and nothing here is a diagnosis or a treatment recommendation.

Frequently asked questions

What is cardiopulmonary coupling?

Cardiopulmonary coupling (CPC) is a measure of how tightly heart rate and breathing move together during sleep. It is computed from one signal, usually a single-lead ECG or a PPG: beat-to-beat intervals are compared with a breathing signal derived from the same recording, and the shared, phase-locked power at each frequency is mapped across the night (Thomas et al., 2005). Power at the breathing rate marks stable NREM sleep; slower cycling marks unstable sleep.

What is a CPC sleep study?

A CPC sleep study is a home or lab recording analyzed with cardiopulmonary coupling. In practice it is usually one sensor worn overnight, such as an ECG patch or a finger PPG ring with oximetry, whose data is uploaded to software that reports sleep stability (stable vs unstable NREM), a sleep quality index and, when oxygen saturation is recorded, an apnea-hypopnea index. FDA-cleared examples include the SleepImage System (K182618 (FDA)) and Largan's DOMII system (K233501 (FDA)). A CPC study does not record brain waves, so it does not score AASM sleep stages.

What do HFC, LFC and VLFC mean in a sleep report?

They are the three CPC bands. High-frequency coupling (HFC, 0.1 to 0.4 Hz) is coupling at the breathing rate and marks stable NREM sleep. Low-frequency coupling (LFC, 0.01 to 0.1 Hz) marks unstable NREM sleep, where breathing and heart rate cycle together over tens of seconds. Very-low-frequency coupling (VLFC, 0.0039 to 0.01 Hz) appears in REM sleep and wake (Schramm et al., 2013).

What is elevated low-frequency coupling (e-LFC)?

e-LFC is a subset of LFC with unusually strong, coherent low-frequency power. It goes with fragmented sleep and sleep-disordered breathing (Schramm et al., 2013). When that power sits in a narrow frequency band (narrow-band e-LFC), it has been linked to central apnea and periodic breathing, and in 77 split-night studies it predicted which patients developed central apneas on CPAP (Thomas et al., 2007).

Is CPC the same as sleep staging? How does CPC compare with PSG?

No. Polysomnography (PSG) scores 30-second epochs as wake, N1, N2, N3 or REM from EEG, eye movement and chin muscle signals (Silber et al., 2007). CPC scores longer windows by how heart rate and breathing couple, and splits NREM into stable and unstable sleep regardless of stage. In the original study, agreement between CPC and standard stages was poor (62.7% in training and 43.9% in test data), while agreement with cyclic alternating pattern scoring was higher (74% and 77.3%) (Thomas et al., 2005). CPC complements PSG; it does not replace it.

Is the SleepImage ring FDA cleared?

The SleepImage System software is FDA cleared under K182618 (FDA) (August 2019) for children, adolescents and adults. It analyzes ECG or pleth (PPG) data with cardiopulmonary coupling and, when oximetry is available, reports an apnea-hypopnea index (sAHI) to aid diagnosis of sleep-disordered breathing. It is prescription only. The ring is one of the recorders that supplies that data; our SleepImage ring explainer covers how it works.

Can a smartwatch or Oura-style ring measure cardiopulmonary coupling?

In principle, yes, if it records clean beat-to-beat pulse timing and a usable breathing signal through the night (Al Ashry, Ni & Thomas, 2021). Huawei says its TruSleep versions 1.0 to 4.0 used CPC to judge sleep stability (Huawei). We found no FDA clearance for CPC on any consumer smartwatch, and no independent validation of consumer CPC scores against polysomnography.

What is a sleep spectrogram?

A CPC sleep spectrogram is a time-frequency map of a night: time runs left to right, frequency runs bottom to top, and color shows how much coherent heart-breathing power sits at each frequency in each window (Thomas et al., 2005). Blocks at 0.1 to 0.4 Hz are stable NREM; blocks lower down are unstable sleep, REM or wake. The illustration above shows two made-up nights.

What is ECG-derived respiration (EDR)?

EDR is a breathing signal pulled out of an ECG. As you breathe, the heart's electrical axis shifts relative to the electrodes, so the height and area of each QRS complex rise and fall with each breath (Wood et al., 2020). Tracking those changes beat by beat gives a respiration surrogate without a chest belt or nasal sensor. CPC pairs EDR with the NN-interval series. A PPG can supply a similar pulse-derived signal; see PPG respiratory rate estimation.

Is a low CPC sleep quality score a diagnosis?

No. CPC outputs are meant to be read by a clinician alongside symptoms and other tests. The FDA-cleared SleepImage labeling says its output is not intended to be interpreted, or clinical action taken, without a qualified healthcare professional (K182618 (FDA)). Consumer sleep scores built on CPC are wellness features, not diagnoses.

References

  1. Al Ashry et al. (2021). Automated apnea-hypopnea index from oximetry and spectral analysis of cardiopulmonary coupling. Annals of the American Thoracic Society. doi:10.1513/AnnalsATS.202005-510OC
  2. Al Ashry, Ni & Thomas (2021). Cardiopulmonary sleep spectrograms open a novel window into sleep biology: implications for health and disease. Frontiers in Neuroscience. doi:10.3389/fnins.2021.755464
  3. Bianchi et al. (2011). Probabilistic sleep architecture models in patients with and without sleep apnea. Journal of Sleep Research. doi:10.1111/j.1365-2869.2011.00937.x
  4. Cho & Kim (2017). The effect of continuous positive airway pressure on cardiopulmonary coupling. Sleep and Breathing. doi:10.1007/s11325-016-1419-8
  5. Cysarz et al. (2018). Sleep instabilities assessed by cardiopulmonary coupling analysis increase during childhood and adolescence. Frontiers in Physiology. doi:10.3389/fphys.2018.00468
  6. Hilmisson et al. (2019). Sleep apnea detection: accuracy of using automated ECG analysis compared to manually scored polysomnography (apnea hypopnea index). Sleep and Breathing. doi:10.1007/s11325-018-1672-0
  7. Hilmisson et al. (2020). Sleep apnea diagnosis in children using software-generated apnea-hypopnea index (AHI) derived from data recorded with a single photoplethysmogram sensor (PPG). Sleep and Breathing. doi:10.1007/s11325-020-02049-6
  8. Hilmisson, Magnusdottir & Thomas (2026). Sex-specific changes and age-specific trends in sleep quality and sleep apnea: analysis of one-million nights of sleep. Sleep Medicine. doi:10.1016/j.sleep.2025.108728
  9. Hilmisson, Thomas & Magnusdottir (2024). Cardiopulmonary coupling-calculated sleep stability and nocturnal heart rate kinetics as a potential indicator for cardiovascular health: a relationship with blood pressure dipping. Frontiers in Sleep. doi:10.3389/frsle.2024.1230958
  10. Ibrahim et al. (2010). Heritability of abnormalities in cardiopulmonary coupling in sleep apnea: use of an electrocardiogram-based technique. Sleep. doi:10.1093/sleep/33.5.643
  11. Liu et al. (2012). HHT based cardiopulmonary coupling analysis for sleep apnea detection. Sleep Medicine. doi:10.1016/j.sleep.2011.10.035
  12. Lu et al. (2023). Comparative study of the SleepImage ring device and polysomnography for diagnosing obstructive sleep apnea. Biomedical Engineering Letters. doi:10.1007/s13534-023-00304-9
  13. Magnusdottir et al. (2020). Cardiopulmonary coupling-derived sleep quality is associated with improvements in blood pressure in patients with obstructive sleep apnea at high-cardiovascular risk. Journal of Hypertension. doi:10.1097/HJH.0000000000002553
  14. Malhotra et al. (2021). Metrics of sleep apnea severity: beyond the apnea-hypopnea index. Sleep. doi:10.1093/sleep/zsab030
  15. Rosenberg & Van Hout (2013). The American Academy of Sleep Medicine inter-scorer reliability program: sleep stage scoring. Journal of Clinical Sleep Medicine. doi:10.5664/jcsm.2350
  16. Schramm & Thomas (2012). Assessment of therapeutic options for mild obstructive sleep apnea using cardiopulmonary coupling measures. Journal of Clinical Sleep Medicine. doi:10.5664/jcsm.1924
  17. Schramm et al. (2013). Quantitative measurement of sleep quality using cardiopulmonary coupling analysis: a retrospective comparison of individuals with and without primary insomnia. Sleep and Breathing. doi:10.1007/s11325-012-0747-6
  18. Silber et al. (2007). The visual scoring of sleep in adults. Journal of Clinical Sleep Medicine. doi:10.5664/jcsm.26814
  19. Sun et al. (2026). The effectiveness of a novel sleep test based on cardiopulmonary coupling to detect obstructive sleep apnea: a multicenter prospective study. Nature and Science of Sleep. doi:10.2147/NSS.S581103
  20. Terzano et al. (2001). Atlas, rules, and recording techniques for the scoring of cyclic alternating pattern (CAP) in human sleep. Sleep Medicine. doi:10.1016/S1389-9457(01)00149-6
  21. Thomas et al. (2005). An electrocardiogram-based technique to assess cardiopulmonary coupling during sleep. Sleep. doi:10.1093/sleep/28.9.1151
  22. Thomas et al. (2007). Differentiating obstructive from central and complex sleep apnea using an automated electrocardiogram-based method. Sleep. doi:10.1093/sleep/30.12.1756
  23. Thomas et al. (2009). Prevalent hypertension and stroke in the Sleep Heart Health Study: association with an ECG-derived spectrographic marker of cardiopulmonary coupling. Sleep. doi:10.1093/sleep/32.7.897
  24. Thomas et al. (2014). Relationship between delta power and the electrocardiogram-derived cardiopulmonary spectrogram: possible implications for assessing the effectiveness of sleep. Sleep Medicine. doi:10.1016/j.sleep.2013.10.002
  25. Thomas et al. (2018). Cardiopulmonary coupling spectrogram as an ambulatory clinical biomarker of sleep stability and quality in health, sleep apnea, and insomnia. Sleep. doi:10.1093/sleep/zsx196
  26. Wang, Shi & Yeh (2023). A novel measure of cardiopulmonary coupling during sleep based on the synchrosqueezing transform algorithm. IEEE Journal of Biomedical and Health Informatics. doi:10.1109/JBHI.2023.3237690
  27. Wood et al. (2020). Multicomponent analysis of sleep using electrocortical, respiratory, autonomic and hemodynamic signals reveals distinct features of stable and unstable NREM and REM sleep. Frontiers in Physiology. doi:10.3389/fphys.2020.592978
  28. Yang et al. (2011). Sleep state instabilities in major depressive disorder: detection and quantification with electrocardiogram-based cardiopulmonary coupling analysis. Psychophysiology. doi:10.1111/j.1469-8986.2010.01060.x
  29. MyCardio LLC (2025). Introduction to SleepImage, clinical instructions for use, D-4.00050 Rev. 31. sleepimage.com
  30. Huawei. Introduction to sleep monitoring principles of HUAWEI smart wearables. consumer.huawei.com
  31. Largan Health AI-Tech. Sleep quality monitoring and management system. largan-health.com
  32. Vivos Therapeutics (2021). Form 8-K, licensing agreement with MyCardio LLC. sec.gov
  33. Mietus JE, Peng CK, Thomas RJ, Goldberger AL. US Patent 8,401,626: System and method for assessing sleep quality. Google Patents
  34. FDA 510(k) records K092003, K163696, K182618, K070855, K112573, K162140 and K233501 were checked in FDA's database on 16 Nov 2026.

Educational content, not medical advice. All signals and nights in the figures are simulated for teaching. Updated 2026-11-16.