cs.CVOct 21, 2026
SaveClinical radar heart sounds with echocardiographic phenotypes of ventricular dysfunction and valvular disease
Organizations: National Institute of Health Data Science, the Institute of Medical Technology, Peking University · Beijing University of Posts and Telecommunications · National Institute of Health Data Science, the Institute of Medical Technology, and the State Key Laboratory of General Artificial Intelligence, Peking University
Abstract
Radar heart sounds capture cardiac mechanical activity without contact, while echocardiography characterizes ventricular and valvular function. Paired measurements connect these signals to structural heart disease, yet such data remain sparsely represented in public resources. Here we present 537 radar heart-sound recordings paired with one echocardiographic report for each of 162 patients. The cohort covers three overlapping phenotypes: left ventricular ejection fraction (LVEF) below 50%, regional wall-motion abnormality and at least moderate valvular regurgitation, present in 38, 43 and 39 patients, respectively. Original complex signals, processed radar heart sounds and temporal quality segmentation support signal-to-phenotype analysis. Technical validation retains 95.93% of signal duration and confirms reproducible waveform reconstruction. Patient-level five-fold evaluation achieves a mean outer-test AUROC of for reduced LVEF, identifying mechanical information associated with ventricular systolic dysfunction. This resource supports research linking non-contact cardiac mechanical signals to echocardiographic structural and functional phenotypes.
Figures & tables
| Characteristic | All | LVEF | RWMA | Valve disease |
|---|---|---|---|---|
| Patients | 162 | 38 | 43 | 39 |
| Male | 109 (67.3%) | 30 (78.9%) | 29 (67.4%) | 23 (59.0%) |
| Female | 53 (32.7%) | 8 (21.1%) | 14 (32.6%) | 16 (41.0%) |
| Age, mean (SD), years | 63.22 (13.00) | 62.21 (16.43) | 64.19 (11.11) | 68.08 (11.69) |
| LVEF, mean (SD), % | 59.75 (14.58) | 37.65 (8.68) | 50.64 (8.80) | 54.87 (17.96) |
| Age, median (IQR), years | 66.5 (56.2–72.0) | 68.0 (50.2–73.0) | 67.0 (56.5–73.0) | 69.0 (65.0–75.5) |
Table 1: Demographic and LVEF distribution in the clinical cohort and phenotype-positive subgroups. Percentages use the column-specific denominator. Phenotype groups are nonexclusive. Each patient contributes one echocardiographic assessment. SD is the sample standard deviation; IQR denotes the first and third quartiles.
| Parameter | Setting | Interpretation |
|---|---|---|
| RF board / acquisition board | TI IWR1443 / DCA1000EVM | Non-contact FMCW front end and raw-data capture |
| Transmit / receive configuration | 1TX4RX | TX0 enabled; TX1/TX2 disabled; four complex receive channels |
| Chirp start frequency | 77.000 GHz | FMCW profile start |
| Frequency slope | 63.343 MHz/ s | Linear frequency modulation |
| Idle / TX start time | 10.00 / 1.00 s | Chirp timing settings |
| ADC start / ramp-end time | 6.00 / 63.14 s | Sampled aperture within the ramp |
Table 2: TI IWR1443/DCA1000EVM acquisition configuration and offline analysis settings. Full-ramp bandwidth and sampled bandwidth are distinct; the 1-ms frame period defines the exported 1,000-Hz vibration time series.
Figure 1: Clinical acquisition environment, FMCW configuration and radar heart-sound workflow. a, Photograph of the clinical acquisition setup, showing a supine patient and the arm-mounted IWR1443BOOST radar and DCA1000EVM capture board positioned above the chest. b, FMCW ramp and sampled aperture: 77-GHz start, 63.343-MHz/ s slope, 63.14- s ramp, 512 ADC samples at 9.121 MS/s, and one chirp per 1-ms frame. c, Complex-signal input, range FFT, maximum-magnitude range-cell selection, final-cell phase extraction, filtering and smoothing, followed by outlier removal and manual quality review. Offline reconstruction uses the final selected cell, preserving historical selection decisions. d, One retained report per selected patient supplies clinical context for three overlapping targets: LVEF below 50%, regional wall-motion abnormality and at least moderate regurgitation of any valve.
Figure 2: Demographic distribution and overlapping clinical phenotypes in 162 patients, each with one retained report. a, Age-band counts stacked by recorded sex; axis labels give the lower age bound. b, Reported clinical LVEF in 5-percentage-point bins, with the orange line denoting 50% and orange bars denoting values strictly below it. c, Positive counts and percentages for the three nonexclusive targets. Valvular disease means at least moderate regurgitation of any valve. d, Joint label counts; filled-dot rows EF, WM and V denote reduced LVEF, regional wall-motion abnormality and valvular disease. The empty intersection denotes the group with none of the three target labels. e, LVEF by nonexclusive group; boxes span the first to third quartile, horizontal lines show medians, and whiskers span the observed minimum to maximum. f, Sex composition of the whole cohort and phenotype-positive groups, with counts inside the bars. Clinical distributions use one observation per patient.
| Representation | Extent | Analytical role |
|---|---|---|
| Original complex radar signal | 60,000 by 512 per recording | Complex I/Q representation for the evaluated RX3 channel; supports range selection and phase-based signal extraction |
| Processed RHS | 60,000 samples at 1,000 Hz per recording | Signed cardiac-related vibration waveform; normalized amplitude is used for waveform illustrations and reconstruction comparisons |
| Temporal quality segmentation | Reviewed intervals on each 60-s time axis | Retained and excluded intervals following outlier screening and manual review; supports quality-aware exposure and spectral analysis |
Table 3: Three complementary representations of each radar measurement. MATLAB format is used for the complex signal and processed RHS; quality segments retain their positions on the original time axis.
Figure 3: Representative radar heart-sound waveforms and their spectra. a, c and e, Six-second excerpts from three recordings selected at approximately the first quartile, median and third quartile of the quality-masked spectral-centroid distribution. Blue denotes the signed RHS waveform and orange its smoothed Hilbert envelope; both use the RMS of the displayed signed excerpt as the amplitude scale. Excerpts lie within retained intervals. b, d and f, Quality-masked spectra from the respective full recordings, normalized to their maximum power; shading indicates 16–80 Hz. R-identifiers denote anonymous analytical recordings. The examples illustrate variation across the spectral-centroid distribution.
Figure 4: Quality-aware signal validation and analysis sensitivity in 537 selected recordings. a, Empirical cumulative distribution of retained duration, calculated from the union of effective deleted intervals before record-edge exclusion. b, Centroid before and after quality masking, with the same one-second edge exclusions. c, Quality-masked centroids in the first and last 30 s. d, Quality-masked centroids using 2.048- and 4.096-s windows, each with 50% overlap. Centroids are restricted to 16–80 Hz. Each point is one recording; gray dashed lines indicate equality. Comparisons use the final reviewed quality masks.
Figure 5: Reconstruction of RHS from the original complex radar signal. a, Pearson correlation of RHS and reconstructed waveforms in 12 deterministically selected recordings. b, Corresponding RMSE after independent mean centring and RMS normalization; the RMSE axis is logarithmic. Both comparisons omit the first and last second. Numbers 01–12 denote the sampled recordings used for normalized waveform-shape reconstruction.
| Experiment / metric | Unit / | Observed result |
| RHS numerical integrity | Recordings / 537 | All 60,000-point vectors finite and nonconstant |
| Quality interval accounting | Recordings / 537 | No kept-plus-deleted duration discrepancy above 2 ms |
| Retained signal exposure | Recordings / 537 | 30,908.45 s total; 95.93% of nominal duration |
| Retained duration, s | Recordings / 537 | 58.77 (56.25–60.00); minimum 40.04 |
| Quality-masked centroid, Hz | Recordings / 537 | 20.89 (20.19–21.66) |
| Processed band-power fraction | Recordings / 537 | 0.873 (0.826–0.905) |
Table 4: Signal and functional validation. Medians of signal metrics are followed by interquartile ranges. Signal-level observations use 537 recordings, including 12 sampled records for reconstruction; LVEF discrimination uses the 162-patient study cohort. AUROC summarizes the five outer test folds.
Appendix figures & tables2 assets
Supplementary material from the paper’s appendix.
Appendix
Figure S1: RF timing, sampled aperture and raw I/Q representation. a, The 77-GHz-start FMCW ramp, 63.343-MHz/ s slope, 6- s ADC delay and 512 samples at 9.121 MS/s distinguish the sampled aperture from the full ramp. b, The confirmed 1TX4RX configuration enables TX0 and records four receive channels. c, One chirp per 1-ms frame yields 60,000 temporal observations over 60 s; the evaluated receive-channel representation contains 60,000 by 512 complex I/Q samples.
Figure S2: Phase-derived radar heart sounds and quality-aware spectral sampling. a, Complex RX3 input, fast-time centring and range FFT, maximum-mean-magnitude target-cell selection, final-cell phase extraction with the 79-GHz conversion constant, forward–backward 16–80-Hz filtering and Savitzky–Golay smoothing, followed by outlier removal and manual quality inspection. Reconstruction uses the final selected cell. The waveform window is 31 points; the illustrated Hilbert envelope uses 101 points. b, Final reviewed intervals in the selected recording with the shortest retained exposure. Blue denotes retained intervals and gray exclusions. The first three complete 2.048-s analysis windows appear in green on separate tracks at their original positions; spectral eligibility additionally excludes the first and last second. No fragments are concatenated. c, Periodic Hann windows of 2,048 and 4,096 samples at 1,000 Hz, evaluated with 50% overlap. The windows specify the spectral-analysis weighting functions.
More in this topic
Left Ventricular Ejection Fraction Estimation
Left ventricular ejection fraction estimation uses cardiac imaging and physiological signals to quantify how much blood the heart pumps per beat.
See all 16 papers