Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 63 of the 63 references it cites were found in a registry or an index.

A wearable cardiac ultrasound imager

Open access

Abstract

Continuous imaging of cardiac functions is highly desirable for the assessment of long-term cardiovascular health, detection of acute cardiac dysfunction and clinical management of critically ill or surgical patients 1–4 . However, conventional non-invasive approaches to image the cardiac function cannot provide continuous measurements owing to device bulkiness 5–11 , and existing wearable cardiac devices can only capture signals on the skin 12–16 . Here we report a wearable ultrasonic device for continuous, real-time and direct cardiac function assessment. We introduce innovations in device design and material fabrication that improve the mechanical coupling between the device and human skin, allowing the left ventricle to be examined from different views during motion. We also develop a deep learning model that automatically extracts the left ventricular volume from the continuous image recording, yielding waveforms of key cardiac performance indices such as stroke volume, cardiac output and ejection fraction. This technology enables dynamic wearable monitoring of cardiac performance with substantially improved accuracy in various environments.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41586-022-05498-z

Cites

63 of 63 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Design and characterization of the wearable cardiac imager. a, Schematics showing the exploded view of the wearable imager, with key components labelled (left) and its working principle (right). b, Resistance of the liquid metal composite electrode as a function of uniaxial tensile strain. The electrode can be stretched up to about 750% without failure. The y axis is the relative resistance defined as R/R0, in which R0 and R are the measured resistances at 0% strain and a given strain, respectively. The inset is a scanning electron micrograph of the liquid metal composite electrodes with a width as small as about 30 µm. Scale bar, 50 μm. c, Cycling performance of the electrode between 0% and 100% uniaxial tensile strain, showing the robustness of the electrode. The inset shows the zoomed-in features of the graph during cyclic stretching and relaxing of the electrode. d, Lap shear strength of the bonding between transducer elements and SEBS or liquid metal composite electrode. Data are mean and s.d. from n = 3 tests. The inset is a schematic setup of the lap shear test. e, Finite element analysis of the entire device under 110% biaxial stretching. f, Optical images showing the mechanical compliance of the wearable imager when bent on a developable surface, wrapped around a non-developable surface, poked and twisted. Scale bars, 5 mm.
Fig. 1 Design and characterization of the wearable cardiac imager. a, Schematics showing the exploded view of the wearable imager, with key components labelled (left) and its working principle (right). b, Resistance of the liquid metal composite electrode as a function of uniaxial tensile strain. The electrode can be stretched up to about 750% without failure. The y axis is the relative resistance defined as R/R0, in which R0 and R are the measured resistances at 0% strain and a given strain, respectively. The inset is a scanning electron micrograph of the liquid metal composite electrodes with a width as small as about 30 µm. Scale bar, 50 μm. c, Cycling performance of the electrode between 0% and 100% uniaxial tensile strain, showing the robustness of the electrode. The inset shows the zoomed-in features of the graph during cyclic stretching and relaxing of the electrode. d, Lap shear strength of the bonding between transducer elements and SEBS or liquid metal composite electrode. Data are mean and s.d. from n = 3 tests. The inset is a schematic setup of the lap shear test. e, Finite element analysis of the entire device under 110% biaxial stretching. f, Optical images showing the mechanical compliance of the wearable imager when bent on a developable surface, wrapped around a non-developable surface, poked and twisted. Scale bars, 5 mm.CC BY · PubMed Central
Fig. 2 B-mode imaging strategies and characterizations. a, Imaging results on wire (100 µm in diameter) phantoms using different transmit beamforming strategies. The first three columns show the images through plane-wave, mono-focus and wide-beam compounding at different depths, respectively. The fourth column shows the imaging resolution of wide-beam compounding in the elevational direction. The bottom row shows images of laterally distributed wires by the wide-beam compounding, from which the lateral accuracy and spatial resolutions at different lateral distances from the central axis can be obtained. b, Signal-to-noise ratios as a function of the imaging depth under different transmission strategies. c, Simulated acoustic fields of the wide-beam compounding, with enhanced acoustic field across the entire insonation area. d, Elevational, lateral and axial resolutions of the device using wide-beam compounding at different depths. e, Lateral and axial resolutions of the device using wide-beam compounding with different lateral distances from the central axis. Data in d and e are mean and s.d. from five tests (n = 5). f, Imaging inclusions with different contrasts to the matrix. On the basis of these B-mode images, the dynamic range (g) and contrast-to-noise ratio (h) of the device can be quantified.
Fig. 2 B-mode imaging strategies and characterizations. a, Imaging results on wire (100 µm in diameter) phantoms using different transmit beamforming strategies. The first three columns show the images through plane-wave, mono-focus and wide-beam compounding at different depths, respectively. The fourth column shows the imaging resolution of wide-beam compounding in the elevational direction. The bottom row shows images of laterally distributed wires by the wide-beam compounding, from which the lateral accuracy and spatial resolutions at different lateral distances from the central axis can be obtained. b, Signal-to-noise ratios as a function of the imaging depth under different transmission strategies. c, Simulated acoustic fields of the wide-beam compounding, with enhanced acoustic field across the entire insonation area. d, Elevational, lateral and axial resolutions of the device using wide-beam compounding at different depths. e, Lateral and axial resolutions of the device using wide-beam compounding with different lateral distances from the central axis. Data in d and e are mean and s.d. from five tests (n = 5). f, Imaging inclusions with different contrasts to the matrix. On the basis of these B-mode images, the dynamic range (g) and contrast-to-noise ratio (h) of the device can be quantified.CC BY · PubMed Central
Fig. 3 Echocardiography in several standard views. a, Schematics and B-mode images of cardiac anatomies from the wearable and commercial imagers. The wearable imager was placed in the parasternal position for imaging in the parasternal long-axis and short-axis views and relocated at the apical position for imaging in the apical four-chamber and two-chamber views. b, 17-segment model representation of the left ventricular wall. Each of the concentrically nested rings that make up the circular plot represents the parasternal short-axis view of the myocardial wall from a different level of the left ventricle. c, B-mode images of the left ventricle in basal, mid-cavity and apical views (top row) and corresponding typical displacement for segments 3, 10 and 14, respectively (bottom row). The physical regions of the left ventricular wall represented by each segment of the 17-segment model have been labelled on the corresponding short-axis views. The peaks are marked with red dots. d, M-mode images (upper left) extracted from parasternal long-axis view and corresponding electrocardiogram signals (lower left). A zoomed-in plot shows the different phases of a representative cardiac cycle (right). Primary events include diastole and opening of the mitral valve during the P-wave of the electrocardiogram, opening of the aortic valve and systole during the QRS complex and closure of the aortic valve during the T-wave. AC, atrial contraction; AMVL, anterior mitral valve leaflet; C.I., commercial imager; ERF, early rapid filling; Ej., ejection; IVCT, isovolumetric contraction time; IVRT, isovolumetric relaxation time; IVS, interventricular septum; LA, left atrium; LV, left ventricle; LVIDd, left ventricular internal diameter end diastole; LVIDs, left ventricular internal diameter end systole; LVOT, left ventricular outflow tract; LVPW, left ventricular posterior wall; MV, mitral valve; RA, right atrium; RV, right ventricle; TV, tricuspid valve; W.I., wearable imager.
Fig. 3 Echocardiography in several standard views. a, Schematics and B-mode images of cardiac anatomies from the wearable and commercial imagers. The wearable imager was placed in the parasternal position for imaging in the parasternal long-axis and short-axis views and relocated at the apical position for imaging in the apical four-chamber and two-chamber views. b, 17-segment model representation of the left ventricular wall. Each of the concentrically nested rings that make up the circular plot represents the parasternal short-axis view of the myocardial wall from a different level of the left ventricle. c, B-mode images of the left ventricle in basal, mid-cavity and apical views (top row) and corresponding typical displacement for segments 3, 10 and 14, respectively (bottom row). The physical regions of the left ventricular wall represented by each segment of the 17-segment model have been labelled on the corresponding short-axis views. The peaks are marked with red dots. d, M-mode images (upper left) extracted from parasternal long-axis view and corresponding electrocardiogram signals (lower left). A zoomed-in plot shows the different phases of a representative cardiac cycle (right). Primary events include diastole and opening of the mitral valve during the P-wave of the electrocardiogram, opening of the aortic valve and systole during the QRS complex and closure of the aortic valve during the T-wave. AC, atrial contraction; AMVL, anterior mitral valve leaflet; C.I., commercial imager; ERF, early rapid filling; Ej., ejection; IVCT, isovolumetric contraction time; IVRT, isovolumetric relaxation time; IVS, interventricular septum; LA, left atrium; LV, left ventricle; LVIDd, left ventricular internal diameter end diastole; LVIDs, left ventricular internal diameter end systole; LVOT, left ventricular outflow tract; LVPW, left ventricular posterior wall; MV, mitral valve; RA, right atrium; RV, right ventricle; TV, tricuspid valve; W.I., wearable imager.CC BY · PubMed Central
Fig. 4 Monitoring during motion. a, Three stages of stress echocardiography. In the rest stage, the subject laid supine for around 4 min. In the exercise stage, the subject rode a stationary bike for about 15 min, with intervals for rest. In the recovery stage, the subject laid supine again for about 10 min. The wearable imager was attached to the chest of the subject throughout the entire test, even during the transitions between the stages. b, Continuous M-mode echocardiography extracted from the parasternal long-axis-view B-mode images of the entire process. Key features of the interventricular septum and left ventricular posterior wall are identified. The stages of rest, exercise (with intervals of rest) and recovery are labelled. c, Variations in the heart rate extracted from the M-mode echocardiography. d, Zoomed-in images of sections 1 (rest), 2 (exercise) and 3 (recovery) (dashed boxes) in b. e, Left ventricular internal diameter end diastole (LVIDd) and left ventricular internal diameter end systole (LVIDs) waveforms of the three different sections of the recording and corresponding average fractional shortenings. f, Zoomed-in images of section 4 (dashed box) during exercise with intervals of rest in b. In the first interval, the subject took a rhythmic deep breath six times, whereas during exercise, there seems to be no obvious signs of a deep breath, probably because the subject switched from diaphragmatic (rest) to thoracic (exercise) breathing, which is shallower and usually takes less time.
Fig. 4 Monitoring during motion. a, Three stages of stress echocardiography. In the rest stage, the subject laid supine for around 4 min. In the exercise stage, the subject rode a stationary bike for about 15 min, with intervals for rest. In the recovery stage, the subject laid supine again for about 10 min. The wearable imager was attached to the chest of the subject throughout the entire test, even during the transitions between the stages. b, Continuous M-mode echocardiography extracted from the parasternal long-axis-view B-mode images of the entire process. Key features of the interventricular septum and left ventricular posterior wall are identified. The stages of rest, exercise (with intervals of rest) and recovery are labelled. c, Variations in the heart rate extracted from the M-mode echocardiography. d, Zoomed-in images of sections 1 (rest), 2 (exercise) and 3 (recovery) (dashed boxes) in b. e, Left ventricular internal diameter end diastole (LVIDd) and left ventricular internal diameter end systole (LVIDs) waveforms of the three different sections of the recording and corresponding average fractional shortenings. f, Zoomed-in images of section 4 (dashed box) during exercise with intervals of rest in b. In the first interval, the subject took a rhythmic deep breath six times, whereas during exercise, there seems to be no obvious signs of a deep breath, probably because the subject switched from diaphragmatic (rest) to thoracic (exercise) breathing, which is shallower and usually takes less time.CC BY · PubMed Central
Fig. 5 Automatic image processing by deep learning. a, Schematic workflow. Pre-processed images are used to train the FCN-32 model. The trained model accepts the unprocessed images and predicts the left ventricular (LV) volume, based on which stroke volume, cardiac output and ejection fraction are derived. b, Left ventricular volume waveform generated by the FCN-32 model from both the wearable imager (W.I.) and the commercial imager (C.I.) (left). Critical features are labelled in one detailed cardiac cycle (right). c, Bland–Altman analysis of the average of (x axis) and the difference between (y axis) the model-generated and manually labelled left ventricular volumes for the wearable (black) and commercial (red) imagers. Dashed lines indicate the 95% confidence interval and about 95% of the data points are within the interval for both imagers. Solid lines indicate mean differences. d, Comparing the stroke volume, cardiac output and ejection fraction extracted from results by the wearable and commercial imagers. Data are mean and s.d. from twelve cardiac cycles (n = 12). e, The model-generated left ventricular volume waveform in the recovery stage. f, Three representative sections of the recording from the initial, middle and end stages of e. End-systolic volume (ESV), end-diastolic volume (EDV), stroke volume and ejection fraction (g) and cardiac output and heart rate waveforms (h) derived from the left ventricular volume waveform. The end-systolic volume and end-diastolic volume gradually recover to the normal range in the end section. The stroke volume increases from about 60 ml to about 70 ml. The ejection fraction decreases from about 80% to about 60%. The cardiac output decreases from about 11 l min−1 to about 9 l min−1, indicating that the decreasing heart rate from about 175 bpm to about 130 bpm overshadowed the increasing stroke volume. AS, atrial systole; IC, isovolumetric contraction; IR, isovolumetric relaxation; RI, rapid inflow.
Fig. 5 Automatic image processing by deep learning. a, Schematic workflow. Pre-processed images are used to train the FCN-32 model. The trained model accepts the unprocessed images and predicts the left ventricular (LV) volume, based on which stroke volume, cardiac output and ejection fraction are derived. b, Left ventricular volume waveform generated by the FCN-32 model from both the wearable imager (W.I.) and the commercial imager (C.I.) (left). Critical features are labelled in one detailed cardiac cycle (right). c, Bland–Altman analysis of the average of (x axis) and the difference between (y axis) the model-generated and manually labelled left ventricular volumes for the wearable (black) and commercial (red) imagers. Dashed lines indicate the 95% confidence interval and about 95% of the data points are within the interval for both imagers. Solid lines indicate mean differences. d, Comparing the stroke volume, cardiac output and ejection fraction extracted from results by the wearable and commercial imagers. Data are mean and s.d. from twelve cardiac cycles (n = 12). e, The model-generated left ventricular volume waveform in the recovery stage. f, Three representative sections of the recording from the initial, middle and end stages of e. End-systolic volume (ESV), end-diastolic volume (EDV), stroke volume and ejection fraction (g) and cardiac output and heart rate waveforms (h) derived from the left ventricular volume waveform. The end-systolic volume and end-diastolic volume gradually recover to the normal range in the end section. The stroke volume increases from about 60 ml to about 70 ml. The ejection fraction decreases from about 80% to about 60%. The cardiac output decreases from about 11 l min−1 to about 9 l min−1, indicating that the decreasing heart rate from about 175 bpm to about 130 bpm overshadowed the increasing stroke volume. AS, atrial systole; IC, isovolumetric contraction; IR, isovolumetric relaxation; RI, rapid inflow.CC BY · PubMed Central
Extended Data Fig. 1 Schematics and optical images of the orthogonal imager. a, The orthogonal imager consists of four arms, in which six small elements in one column are combined as one long element, and a central part that is shared by the four arms. The blue and red boxes label a long element integrated by six small pieces in each direction. The number of elements in one direction is 32. The pitch between the elements is 0.4 mm. Optical images in top view (b) and isometric view (c) showing the morphology of the orthogonal array. We used an automatic alignment strategy to fabricate the orthogonal array by bonding a large piece of backing layer with a large piece of 1-3 composite and then dicing them together into small elements with designed configurations. Inset in c shows the details of the elements. The 1-3 composite and backing layer have been labelled.
Extended Data Fig. 1 Schematics and optical images of the orthogonal imager. a, The orthogonal imager consists of four arms, in which six small elements in one column are combined as one long element, and a central part that is shared by the four arms. The blue and red boxes label a long element integrated by six small pieces in each direction. The number of elements in one direction is 32. The pitch between the elements is 0.4 mm. Optical images in top view (b) and isometric view (c) showing the morphology of the orthogonal array. We used an automatic alignment strategy to fabricate the orthogonal array by bonding a large piece of backing layer with a large piece of 1-3 composite and then dicing them together into small elements with designed configurations. Inset in c shows the details of the elements. The 1-3 composite and backing layer have been labelled.CC BY · PubMed Central
Extended Data Fig. 2 Characterization of the effects of phase correction on imaging quality. B-mode images of a line phantom obtained from different situations (a). Left, from a planar surface. Middle, from a curvilinear surface without phase correction. Right, from a curvilinear surface with phase correction. b, The axial and lateral resolutions at different depths under these three situations. No obvious difference in axial resolution was found because it is mainly dependent on the transducer frequency and bandwidth. The lateral resolution of the wearable imager was improved after phase correction. Images collected in (c), the parasternallong-axis (PLAX) view of the heart and (d), the apical four-chamber view when measured by a planar probe (left panel), a curved probe without phase correction (middle panel) and a curved probe with phase correction (right panel). The left ventricular boundaries are labelled by white dashed lines in the images. e, Comparison of measured cardiac indices showing the impact of phase correction. Each measurement is based on the mean of five consecutive cardiac cycles (n = 5). The standard deviations are indicated by the error bars.
Extended Data Fig. 2 Characterization of the effects of phase correction on imaging quality. B-mode images of a line phantom obtained from different situations (a). Left, from a planar surface. Middle, from a curvilinear surface without phase correction. Right, from a curvilinear surface with phase correction. b, The axial and lateral resolutions at different depths under these three situations. No obvious difference in axial resolution was found because it is mainly dependent on the transducer frequency and bandwidth. The lateral resolution of the wearable imager was improved after phase correction. Images collected in (c), the parasternallong-axis (PLAX) view of the heart and (d), the apical four-chamber view when measured by a planar probe (left panel), a curved probe without phase correction (middle panel) and a curved probe with phase correction (right panel). The left ventricular boundaries are labelled by white dashed lines in the images. e, Comparison of measured cardiac indices showing the impact of phase correction. Each measurement is based on the mean of five consecutive cardiac cycles (n = 5). The standard deviations are indicated by the error bars.CC BY · PubMed Central
Extended Data Fig. 3 Optical images showing positions and orientations for ultrasound heart imaging. a, Parasternal long-axis view. b, Parasternal short-axis view. c, Apical four-chamber view. d, Apical two-chamber view. The orthogonal wearable cardiac imager combines parasternal long-axis and short-axis views (e) and apical four-chamber and apical two-chamber views without rotation (f). The wearable imager can capture two parasternal views from a single position or two apical views from another single position. The sternum and ribs are labelled to indicate intercostal spaces.
Extended Data Fig. 3 Optical images showing positions and orientations for ultrasound heart imaging. a, Parasternal long-axis view. b, Parasternal short-axis view. c, Apical four-chamber view. d, Apical two-chamber view. The orthogonal wearable cardiac imager combines parasternal long-axis and short-axis views (e) and apical four-chamber and apical two-chamber views without rotation (f). The wearable imager can capture two parasternal views from a single position or two apical views from another single position. The sternum and ribs are labelled to indicate intercostal spaces.CC BY · PubMed Central
Extended Data Fig. 4 B-mode images collected from a subject with different postures. The four views collected when the subject is sitting (a), standing (b), bending over (c), lying flat (d) and lying on their side (e). The PLAX and PSAX views can keep their quality at different postures, whereas the quality of A4C and A2C views can only be achieved when lying on the side. A2C, apical two-chamber view; A4C, apical four-chamber view; PLAX, parasternal long-axis view; PSAX, parasternal short-axis view.
Extended Data Fig. 4 B-mode images collected from a subject with different postures. The four views collected when the subject is sitting (a), standing (b), bending over (c), lying flat (d) and lying on their side (e). The PLAX and PSAX views can keep their quality at different postures, whereas the quality of A4C and A2C views can only be achieved when lying on the side. A2C, apical two-chamber view; A4C, apical four-chamber view; PLAX, parasternal long-axis view; PSAX, parasternal short-axis view.CC BY · PubMed Central
Extended Data Fig. 5 Continuous cardiac imaging during rest, exercise and recovery. Representative B-mode and M-mode images during rest (a), exercise (b) and recovery (c). The red line highlights the M-mode section corresponding to the current B-mode frame. More details can be seen in Supplementary Video 3.
Extended Data Fig. 5 Continuous cardiac imaging during rest, exercise and recovery. Representative B-mode and M-mode images during rest (a), exercise (b) and recovery (c). The red line highlights the M-mode section corresponding to the current B-mode frame. More details can be seen in Supplementary Video 3.CC BY · PubMed Central
Extended Data Fig. 6 Segmentation results of the left ventricle with different deep learning models. By qualitatively evaluating the result, we found no ‘jitteriness’ in Supplementary Video 4. The segmented left ventricle contracts and relaxes as naturally as the B-mode video. The segmentation boundaries are smooth with the highest fidelity. Compared with the original B-mode image, the FCN-32 model has the best agreement among all models used in this study.
Extended Data Fig. 6 Segmentation results of the left ventricle with different deep learning models. By qualitatively evaluating the result, we found no ‘jitteriness’ in Supplementary Video 4. The segmented left ventricle contracts and relaxes as naturally as the B-mode video. The segmentation boundaries are smooth with the highest fidelity. Compared with the original B-mode image, the FCN-32 model has the best agreement among all models used in this study.CC BY · PubMed Central
Extended Data Fig. 7 Waveforms of the left ventricular volume obtained with different deep learning models. Those waveforms are from segmenting the same B-mode video. Qualitatively, the waveform generated by the FCN-32 model gains the best stability and the least noise, and the waveform morphology is more constant from cycle to cycle. Quantitatively, the comparison results of those models is in Supplementary Fig. 26, which shows that the FCN-32 model has the highest mean intersection over union, showing the best performance in this study.
Extended Data Fig. 7 Waveforms of the left ventricular volume obtained with different deep learning models. Those waveforms are from segmenting the same B-mode video. Qualitatively, the waveform generated by the FCN-32 model gains the best stability and the least noise, and the waveform morphology is more constant from cycle to cycle. Quantitatively, the comparison results of those models is in Supplementary Fig. 26, which shows that the FCN-32 model has the highest mean intersection over union, showing the best performance in this study.CC BY · PubMed Central
Extended Data Fig. 8 Different phases in a cardiac cycle obtained from B-mode imaging. The rows are B-mode images of A4C, A2C, PLAX and PSAX views in the same phase. The columns are B-mode images of the same view during ventricular filling, atrial contraction, isovolumetric contraction, end of ejection and isovolumetric relaxation. The dashed lines highlight the main features of the current phase. Bluish lines mean shrinking in the volume of the labelled chamber. Reddish lines mean expansion in the volume of the labelled chamber. Yellowish lines mean retention in the volume of the labelled chamber. A2C, apical two-chamber view; A4C, apical four-chamber view; LA, left atrium; LV, left ventricle; LVOT, left ventricular outflow tract; RA, right atrium; RV, right ventricle; PLAX, parasternal long-axis view; PSAX: parasternal short-axis view.
Extended Data Fig. 8 Different phases in a cardiac cycle obtained from B-mode imaging. The rows are B-mode images of A4C, A2C, PLAX and PSAX views in the same phase. The columns are B-mode images of the same view during ventricular filling, atrial contraction, isovolumetric contraction, end of ejection and isovolumetric relaxation. The dashed lines highlight the main features of the current phase. Bluish lines mean shrinking in the volume of the labelled chamber. Reddish lines mean expansion in the volume of the labelled chamber. Yellowish lines mean retention in the volume of the labelled chamber. A2C, apical two-chamber view; A4C, apical four-chamber view; LA, left atrium; LV, left ventricle; LVOT, left ventricular outflow tract; RA, right atrium; RV, right ventricle; PLAX, parasternal long-axis view; PSAX: parasternal short-axis view.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 63 references

Research works printed but not linked: 2

  1. García-González, M. A., Argelagós-Palau, A., Fernández-Chimeno, M. & Ramos-Castro, J. in Computing in Cardiology 2013 461–464 (IEEE, 2014).
  2. Sung, F. et al. in Proc. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition 1199–1208 (IEEE, 2018).

Cited by

See all 644

Metadata

Authors
Hu, H., Huang, H., Li, M., Gao, X., Yin, L., Qi, R., Wu, R. S., Chen, X., Ma, Y., Shi, K., Li, C., Maus, T. M., Huang, B., Lu, C., Lin, M., Zhou, S., Lou, Z., Gu, Y., Chen, Y., Lei, Y., Wang, X., Wang, R., Yue, W., Yang, X., Bian, Y., Mu, J., Park, G., Xiang, S., Cai, S., Corey, P. W., Wang, J., & Xu, S.
Affiliations
  • University of California San Diego (US)
  • University of California, Riverside (US)
  • Yale University (US)
  • Sharp Memorial Hospital (US)
Corresponding author
Xu, Sheng
Venue
Nature
Published
2023-01-25
Volume
613
Issue
7945
Pages
667-675
ISSN
  • 0028-0836 (ISSN-L)
  • 1476-4687
OpenAlex
W4317951947
Cited by
644
References
63
Citations by year
2026: 148 · 2025: 216 · 2024: 181 · 2023: 99
FWCI
50.87
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • eScholarship (California Digital Library) (submittedVersion, cc-by)
Funders
  • National Institutes of Health (3r21eb027303-02s1, 1r21eb027303-01a1, 1r21eb025521-01, 1r01eb033464-01)
  • University of California, San Diego
  • National Institute of Biomedical Imaging and Bioengineering (R21 EB027303, R21 EB025521, 1 R01 EB03346401, R01 EB033464, 1R21EB027303-01A1, 1R21EB025521-01, 3R21EB027303-02S1)
Topics
  • Physical Sciences › Engineering › Biomedical Engineering › Non-Invasive Vital Sign Monitoring
  • Health Sciences › Medicine › Critical Care and Intensive Care Medicine › Ultrasound in Clinical Applications
  • Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › Cardiovascular Function and Risk Factors
Keywords
cardiac function monitoring, deep learning segmentation, left ventricular volume, stroke volume, cardiac output, ejection fraction, wearable device design
MeSH terms
Wearable Electronic Devices, Cardiac Output, Echocardiography, Equipment Design, Heart, Heart Ventricles, Humans, Skin, Stroke Volume

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: A wearable cardiac ultrasound imager. https://app.cochat.ai/papers/10.1038/s41586-022-05498-z
The paper
Hu, H., Huang, H., Li, M., Gao, X., Yin, L., Qi, R., Wu, R. S., Chen, X., Ma, Y., Shi, K., Li, C., Maus, T. M., Huang, B., Lu, C., Lin, M., Zhou, S., Lou, Z., Gu, Y., Chen, Y., . . . Xu, S. (2023). A wearable cardiac ultrasound imager. Nature. https://doi.org/10.1038/s41586-022-05498-z

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.