Pith. sign in

REVIEW 2 major objections 5 minor 45 references

MD-ViSCo: A Unified Model for Multi-Directional Vital Sign Waveform Conversion

T0 review · 2 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that one unified model can generate any target waveform among ECG, PPG, and ABP from any single source waveform, with lower average error than six task-specific baselines and AAMI/BHS-compliant ABP output.

desk verdict The unified multi-directional conversion claim holds up, but the AAMI/BHS clinical compliance headline is not established for calibration-free use—it rests on a calibration-based split and reduces to the BP regressor's accuracy. read the letter →

arxiv 2506.08357 v1 pith:MHKBEP73 submitted 2025-06-10 cs.SD cs.AIeess.AS

Franck Meyer , Kyunghoon Hur , Edward Choi This is my paper · ORCID
classification cs.SDcs.AIeess.AS
keywords vitalsignwaveformconversionunifiedmodelECGPPGABPAdaINSwinTransformerAAMI/BHSstandards
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a single model can replace the usual collection of six separate source-to-target waveform converters: from ECG, PPG, or ABP as input, it can generate any of the other waveform types. The payoff is practical: one architecture, one training run, and one deployed model instead of several task-specific pipelines, which matters in hospital settings where the available sensor can change. On two public ICU datasets, the model reports an 8.8% lower mean absolute error and a 4.9% higher Pearson correlation averaged over all conversion directions and single-direction baselines. For generated ABP waveforms, the paper reports compliance with the AAMI standard and Grade B on the BHS standard, which are clinical thresholds used for blood-pressure measurement devices.

What carries the argument

The workhorse is Adaptive Instance Normalization (AdaIN) injected through a shallow 1-D U-Net with a Swin Transformer bottleneck. AdaIN re-centers and re-scales the feature map's mean and standard deviation using a style vector $\gamma, \beta$ learned from the one-hot target choice, so the same weights generate every target modality while the Swin Transformer supplies global temporal context. The refinement stage uses a PatchTSMixer encoder for the source waveform, a DistilBERT text encoder for patient demographics, a weighted contrastive loss that aligns clinically similar samples, and a linear transformation $\hat{y}_{ABP} = \hat{y}_{apx} \cdot (\hat{y}_{SBP} - \hat{y}_{DBP}) + \hat{y}_{DBP}$ that rescales the normalized waveform into mmHg.

What would settle it

On a strictly patient-disjoint test set with no calibration finetuning, compute BHS grades from the systolic and diastolic values detected directly on the generated ABP waveform rather than from the regressor's predictions, and check whether the pass rates still meet the AAMI/BHS thresholds; if they drop, the compliance claim is confined to the calibrated setting.

Watch

Extended reading notes

Core claim

The central discovery is that multi-directional waveform conversion can be treated as a style-transfer problem. All waveforms are locally min-max normalized so that amplitude scales match, all modalities are pooled into one training set, and a one-hot target code is mapped to a style embedding that Adaptive Instance Normalization uses to shift and scale internal feature statistics. The resulting single model generates normalized waveforms in any target modality. A refinement stage then converts the normalized ABP waveform into millimeters of mercury by multiplying by the predicted systolic-to-diastolic range and adding the predicted diastolic pressure, using patient demographics encoded as text. On the tested datasets, this one-model design matches or beats per-direction baselines, and the authors report that the generated ABP waveforms satisfy AAMI and BHS Grade B.

Load-bearing premise

The load-bearing premise is that the AAMI/BHS compliance scores, which come from a calibration finetuning split where the same patient can appear in both tuning and test sets and where the waveform's systolic and diastolic values are set by the regression model, should be taken as evidence about the generated waveform itself.

Editorial extensions

If this is right

  • Only one trained model needs to be deployed for conversions among ECG, PPG, and ABP, so adding a conversion direction costs no new architecture or separate maintenance pipeline.
  • Multi-directional training improves even per-direction error relative to uni-directional training, with reported average MAE reductions of 5.4% for ECG generation, 3.1% for PPG generation, and 4.0% for ABP generation on PulseDB.
  • If the AAMI/BHS results hold, the model supports non-invasive-to-invasive ABP monitoring in wards or pre-hospital settings where only PPG or ECG is available.
  • Encoding patient information as text lets the refinement stage accept different demographic schemas without manual feature engineering, easing portability across datasets with different patient tables.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves implicit that its AAMI/BHS compliance scores are, in effect, scores for the blood-pressure regressor followed by a linear rescale, not for the generated waveform's own morphology; a stricter reading would grade systolic and diastolic values detected directly on the raw generated ABP waveform.
  • An untested extension is that the same style-transfer design could absorb additional waveform types, such as respiration, as new target domains by enlarging the one-hot code and reusing the normalization and AdaIN machinery.
  • Because inference requires only one input waveform, the model is a natural fit for wearable or ward monitoring where sensor availability varies; the paper does not quantify performance when the single input waveform is noisy or missing, which a deployment study could add.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper proposes MD-ViSCo, a two-stage framework for multi-directional vital sign waveform conversion among ECG, PPG, and ABP. The approximation stage uses a 1D U-Net with a Swin Transformer and AdaIN to generate a locally normalized target waveform from any single source waveform and a target domain indicator. The refinement stage predicts SBP and DBP from the source waveform and optional patient demographics, then rescales the normalized ABP waveform to mmHg via a linear transformation. The authors evaluate on PulseDB and UCI against NabNet, PPG2ABP, PatchTST, and P2E-WGAN, reporting MAE/PC, physiological feature fidelity, and AAMI/BHS compliance, and claim that one unified model matches or outperforms six separate uni-directional baselines while satisfying AAMI and BHS Grade B for ABP generation.

Significance. If the waveform-conversion results hold, the unified framework is a practically useful contribution: it removes the need to train and maintain six separate source-to-target models and is backed by five-seed experiments, a patient-level calibration-free split for the approximation stage, multiple evaluation metrics, ablations, and public code. The physiological feature fidelity results and multi-directional-versus-uni-directional ablation are informative. The clinical-compliance claim, however, is currently over-stated because the AAMI/BHS evaluation is performed under a calibration-based split and, through the linear transformation, reduces to BP-regressor accuracy rather than waveform morphology. The core waveform-generation contribution is sound enough to warrant a major revision rather than rejection, provided the clinical claims are re-scoped or re-evaluated.

major comments (2)
  1. [Section 3.5.1, Table 2] The AAMI/BHS evaluations are reported on a calibration-based refinement split: the original patient-disjoint test set is further split into finetune-train, finetune-val, and finetune-test, and the text explicitly states that "the same patient may appear in both the training and test sets." Section 4.3 and the abstract nevertheless present the AAMI pass and BHS Grade B without this caveat. Because AAMI/BHS are device-approval standards that assume evaluation on subjects not used for calibration, the headline clinical-compliance claim is not established for the calibration-free setting that the paper otherwise emphasizes. The conclusion even concedes that the refinement step "may involve task- or dataset-specific finetuning." Please either report AAMI/BHS on the original patient-disjoint test set with the refinement model frozen after training on the train split, or clearly qualify all clinical-compliance claims as calibration-based in the abstract and Section 4.3.
  2. [Section 3.4, Eq. (5)] The linear transformation y_ABP = y_ABP_apx · (SBP_pred − DBP_pred) + DBP_pred, combined with the local min-max normalization of y_ABP_apx, makes the generated waveform's systolic and diastolic values identically equal to the refinement model's outputs SBP_pred and DBP_pred. Consequently, the AAMI/BHS metrics in Table 3 and Figure 5 evaluate the accuracy of the BP regressor, not the morphology generated by the approximation model. Morphology is assessed separately by MAE/PC and feature fidelity, but the text in Section 4.3 and the abstract overstates what the clinical-compliance experiments test. Please rephrase these claims to state that the AAMI/BHS evaluation validates the amplitude (BP) estimates, and clarify that the waveform shape itself is not evaluated by these standards.
minor comments (5)
  1. [Section 3.3, Eq. (1)] The notation for the Swin Transformer encoder is inconsistent: z = SwinT_E(h) appears after B(h, s) = \tilde{h} is defined, so the relationship between the bottleneck B and SwinT_E is unclear; please align the names and make the flow x → h → \tilde{h} → \hat{y} unambiguous.
  2. [Table 3] Please report the number of unique subjects used for the AAMI evaluation, since the AAMI standard requires at least 85 subjects and the table currently lists only ME and SD.
  3. [Figure 3] The y-axis scales differ across panels because ECG/PPG are in normalized units while ABP is in mmHg; please add explicit annotations or a supplementary common-scale figure to prevent misleading cross-panel comparisons.
  4. [Appendix 6.1.1] The global normalization bounds for PulseDB (2.34–286.58 mmHg) appear extreme; please clarify whether these are raw min/max values and whether outlier removal was applied before computing the bounds.
  5. [Appendix 6.3] The WCL similarity threshold symbols (\tau_s, T_s) are used in Eq. (6) and Eq. (7) before being defined; please define all hyperparameters in the main text or refer explicitly to Table 8 when the symbols first appear.

Circularity Check

2 steps flagged · score 6.0 of 10

AAMI/BHS compliance reduces by construction to the fitted BP-regressor, and it is evaluated under an admitted same-patient calibration split.

  1. self definitional [Section 3.4, Linear transformation; Section 4.3, AAMI/BHS Standard Evaluation]
    "Given the predicted ˆySBP and ˆyDBP values from the multi-modal encoder, we perform a linear transformation to convert the normalized ABP waveform ˆyABP_apx, generated by the approximation model, into a real-valued waveform in mmHg ˆyABP. ... ˆyABP = ˆyABP_apx ·(ˆySBP −ˆyDBP) + ˆyDBP. ... Specifically, we compare SBP, DBP, and mean arterial pressure (MAP) from the generated ABP waveforms to ground-truth values."

    The final ABP waveform is defined by a linear transformation that maps the normalized waveform to the predicted range of DBP and SBP. Since the approximation output is locally min-max normalized, the generated waveform's minimum is exactly ˆyDBP and its maximum is exactly ˆySBP by construction. AAMI and BHS scores are then computed from the SBP, DBP, and MAP values of this generated waveform, so those scores evaluate only the refinement model's fitted BP regression outputs. The approximation model's morphology does not affect any of the AAMI/BHS numbers; a completely wrong shape with the same predicted SBP/DBP would yield identical clinical-compliance results.

  2. fitted input called prediction [Section 3.5.1, Data split; Table 2 note; Section 4.3 results]
    "the original test set is further split into new train, validation, and test subsets to support finetuning under a calibration-based setting where the same patient may appear in both training and test sets."

    The AAMI and BHS results in Table 3 and Figure 5 are reported on the finetune-test subset created by further splitting the original calibration-free test set. That finetune-test subset shares patients with the finetune-train subset, as the paper explicitly states. The refinement model is therefore fitted on data from the same patients on which the clinical-compliance 'prediction' is evaluated. The abstract presents the AAMI/BHS compliance as a general capability, but the supporting numbers are calibration-based, same-patient evaluations; the paper itself concedes in the conclusion that the refinement step 'may involve task- or dataset-specific finetuning.' This makes the clinical-compliance prediction a fitted, calibration-dependent result rather than an unseen-patient prediction.

full rationale

The paper's core contributions—the unified multi-directional approximation model, AdaIN style injection, and the MAE/PC waveform-conversion comparisons against NabNet, PPG2ABP, PatchTST, and P2E-WGAN on PulseDB and UCI—are empirical and self-contained, and they are evaluated against independent baselines on public datasets. Those parts are not circular. The circularity is confined to the headline clinical-compliance claim. In Section 3.4, the final ABP waveform is defined by the linear transformation ŷABP = ŷABP_apx·(ŷSBP − ŷDBP) + ŷDBP, so the generated waveform's systolic and diastolic values coincide with the refinement model's predicted SBP and DBP. AAMI/BHS evaluation then compares SBP, DBP, and MAP extracted from this waveform against ground truth, which by construction measures only the fitted BP regressor, not the waveform morphology produced by the approximation stage. Section 3.5.1 additionally admits that the finetune-test set shares patients with finetune-train, so the reported AAMI/BHS pass is a calibration-based, same-patient result, not a calibration-free prediction; the conclusion similarly concedes that the refinement step 'may involve task- or dataset-specific finetuning.' These are two separate reductions of the same central claim: the clinical numbers are determined by fitted BP values and are obtained after calibrating on the evaluation cohort. The cited text-embedding works [32,33] involving the same authors are implementation choices for patient-information encoding and are not load-bearing justifications for the main result. Overall score 6: one central 'prediction' (clinical compliance) reduces by construction to a fitted regression and a same-patient calibration split, while the rest of the paper's empirical waveform-conversion claims retain independent content.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new physical entities, particles, forces, or conserved quantities are introduced. The paper is an empirical machine learning contribution; its non-derived inputs are normalization bounds, contrastive-learning thresholds, and the domain assumptions listed above about synchronized paired waveforms and the applicability of device standards to generated signals.

free parameters (3)
  • WCL similarity loss weights and thresholds = lambda_MAE=0.001, lambda1=0.01, lambda2=0.01, tau_s=4, T_s=0.0235, tau_w=4
    Table 8 lists hand-set hyperparameters for the weighted contrastive learning objective in the refinement model; no derivation or sensitivity analysis is given.
  • Global ABP normalization bounds = UCI: 50-189.98 mmHg; PulseDB: 2.34-286.58 mmHg
    Section 6.1.1 fixes these bounds from the training set to rescale normalized ABP predictions back to mmHg; they are fitted to the training data and affect all ABP metrics.
  • AdaIN style vector dimension = 64
    Section 6.2.1 sets the style-conditioned embedding size to 64; it is chosen by hand and not ablated.
assumptions (4)
  • domain assumption Source and target waveforms in each training pair are synchronized, same-patient, same-time recordings (Equation 3).
    The conversion supervision assumes y(j) is the exact target paired with x(i); Section 3.1 states all three waveforms come from the same records at 125 Hz.
  • domain assumption AAMI and BHS device-validation standards are applicable to model-generated waveforms.
    Section 4.3 applies regulatory device standards designed for physical blood pressure measurement devices to neural network outputs without discussing whether the simultaneous-recording setting satisfies the intent of those standards.
  • ad hoc to paper DistilBERT tokenization of demographic strings preserves the patient information needed for BP refinement.
    Section 3.4 converts age/gender/height/weight/BMI to text and embeds with DistilBERT; no evidence is given that textual encoding is at least as informative as structured features.
  • standard math Neural network training and backpropagation converge to a model that generalizes to unseen patients.
    Standard deep learning assumption; the paper relies on five-seed averages and patient-level splits rather than formal guarantees.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MD-ViSCo: A Unified Model for Multi-Directional Vital Sign Waveform Conversion." pith.science (2026). https://pith.science/paper/MHKBEP73

@misc{pith2026250608357,
  author       = {Pith},
  title        = {Pith review of: MD-ViSCo: A Unified Model for Multi-Directional Vital Sign Waveform Conversion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MHKBEP73}},
  note         = {Machine review of arXiv:2506.08357}
}
read the original abstract

Despite the remarkable progress of deep-learning methods generating a target vital sign waveform from a source vital sign waveform, most existing models are designed exclusively for a specific source-to-target pair. This requires distinct model architectures, optimization procedures, and pre-processing pipelines, resulting in multiple models that hinder usability in clinical settings. To address this limitation, we propose the Multi-Directional Vital-Sign Converter (MD-ViSCo), a unified framework capable of generating any target waveform such as electrocardiogram (ECG), photoplethysmogram (PPG), or arterial blood pressure (ABP) from any single input waveform with a single model. MD-ViSCo employs a shallow 1-Dimensional U-Net integrated with a Swin Transformer that leverages Adaptive Instance Normalization (AdaIN) to capture distinct waveform styles. To evaluate the efficacy of MD-ViSCo, we conduct multi-directional waveform generation on two publicly available datasets. Our framework surpasses state-of-the-art baselines (NabNet & PPG2ABP) on average across all waveform types, lowering Mean absolute error (MAE) by 8.8% and improving Pearson correlation (PC) by 4.9% over two datasets. In addition, the generated ABP waveforms satisfy the Association for the Advancement of Medical Instrumentation (AAMI) criterion and achieve Grade B on the British Hypertension Society (BHS) standard, outperforming all baselines. By eliminating the need for developing a distinct model for each task, we believe that this work offers a unified framework that can deal with any kind of vital sign waveforms with a single model in healthcare monitoring.

Figures

Figures reproduced from arXiv: 2506.08357 by the authors.

Figure 1
Figure 1. For the conventional waveform conver￾sion framework, a distinct model must be trained for each source-to-target direction, each with its own architecture, objective function, and subsets of selected demographic features. photoplethysmograms (PPG), and arterial blood pressure (ABP) waveforms is critical for track￾ing patients’ status, which in turn facilitates timely and effective medical intervention [1, 2]. However… view at source ↗
Figure 2
Figure 2. Overview of MD-ViSCo. The approximation model [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparison of MAE (↓) and PC (↑) from waveform conversion experiments on the PulseDB and UCI datasets across six waveform conversion directions (source waveform → target waveform). Bar colors correspond to each method. Error caps show ±1 SD over five random seeds. MD-ViSCo (ours) method (dark red) consistently has shown better or comparable performance with baseline models (orange gradient). Statistical significance… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Comparison of relative mean error (%) in physiological feature estimation across different [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: BHS-standard evaluation of generated ABP waveforms on the PulseDB and UCI test [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 23 canonical work pages

  1. [1]

    A. J. E. Seely and K. Newman and R. Ram- chandani and C. Herry and N. Scales and 16 N. Hudek and J. Brehaut and D. Jones and T. Ramsay and D. Barnaby and S. Fer- nando and J. Perry and S. Dhanani and K. E. A. Burns, “Roadmap for the evolution of monitoring: developing and evaluating waveform-based variability-derived artificial intelligence-powered predic...

    work page doi:10.1186/s13054-024-05140- 2024

  2. [2]

    A novel artificial in- telligence based intensive care unit monitor- ing system: using physiological waveforms to identify sepsis,

    M. Mollura and L.-W. H. Lehman and R. G. Mark and R. Barbieri, “A novel artificial in- telligence based intensive care unit monitor- ing system: using physiological waveforms to identify sepsis,”Philosophical Transac- tions of the Royal Society A: Mathemati- cal, Physical and Engineering Sciences, vol. 379, no. 2212, pp. 20200252, 2021, doi: 10.1098/rsta....

    work page arXiv 2021

  3. [3]

    Impact on Pa- tient Outcomes of Continuous Vital Sign Monitoring on Medical Wards: Propensity- Matched Analysis,

    B. Rowland, A. Saha, V. Motamedi, R. Bundy, S. Winsor, D. McNavish, W. Lip- pert, and A. K. Khanna, “Impact on Pa- tient Outcomes of Continuous Vital Sign Monitoring on Medical Wards: Propensity- Matched Analysis,”J. Med. Internet Res., vol. 27, p. e66347, Mar. 11, 2025, doi: 10.2196/66347

    work page doi:10.2196/66347 2025

  4. [4]

    Cross-Domain Joint Dictionary Learning for ECG Inference From PPG,

    X. Tian and Q. Zhu and Y. Li and M. Wu, “Cross-Domain Joint Dictionary Learning for ECG Inference From PPG,” IEEE Internet of Things Journal, vol. 10, no. 9, pp. 8140–8154, 2023, doi: 10.1109/JIOT.2022.3231862

    arXiv 2023

  5. [5]

    Estimation of Arte- rial Blood Pressure Waveform from Photo- plethysmogram Signal using Linear Trans- fer Function Approach,

    A. Dash and N. Ghosh and A. Patra and A. D. Choudhury, “Estimation of Arte- rial Blood Pressure Waveform from Photo- plethysmogram Signal using Linear Trans- fer Function Approach,” inProceedings of the 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Mon- treal, QC, Canada, 2020, pp. 2691–2694, doi: ...

    arXiv 2020

  6. [6]

    A review of noninvasive methodologies to estimate the blood pressure waveform,

    T. Athaya and S. Choi, “A review of noninvasive methodologies to estimate the blood pressure waveform,”Sensors, vol. 22, no. 10, pp. 3953, 2022,MDPIdoi: 10.3390/s22103953

    work page doi:10.3390/s22103953 2022

  7. [7]

    Clinical review: Complications and risk factors of peripheral arterial catheters used for haemodynamic monitoring in anaesthe- sia and intensive care medicine,

    B. V. Scheer and A. Perel and U. J. Pfeiffer, “Clinical review: Complications and risk factors of peripheral arterial catheters used for haemodynamic monitoring in anaesthe- sia and intensive care medicine,”Critical Care, vol. 6, no. 3, pp. 199, 2002,Springer Naturedoi: 10.1186/cc1489

    work page doi:10.1186/cc1489 2002

  8. [8]

    An estimation method of continuous non-invasive arte- rial blood pressure waveform Using pho- toplethysmography: a U-Net architecture- based approach,

    T. Athaya and S. Choi, “An estimation method of continuous non-invasive arte- rial blood pressure waveform Using pho- toplethysmography: a U-Net architecture- based approach,”Sensors, vol. 21, no. 5, pp. 1867, 2021,MDPIdoi: 10.3390/s21051867

    work page doi:10.3390/s21051867 2021

Show all 45 references
  1. [9]

    A re- view of deep learning methods for pho- toplethysmography data,

    G. Nie, J. Zhu, G. Tang, D. Zhang, S. Geng, Q. Zhao,et al., “A re- view of deep learning methods for pho- toplethysmography data,”arXiv preprint arXiv:2401.12783, 2024. [Online]. Avail- able: 10.48550/arXiv.2401.12783

    2024 doi arXiv

  2. [10]

    Reconstruction of missing phys- iological signals using artificial neural net- works,

    A. Sullivan and H. Xia and J. McBride and X. Zhao, “Reconstruction of missing phys- iological signals using artificial neural net- works,” inProceedings of the 2010 Comput- ing in Cardiology, Belfast, UK, 2010, pp. 317–320

    2010

  3. [11]

    Foundation models for physiolog- ical data, how to develop them, and what 17 to expect of them,

    X. Hu, “Foundation models for physiolog- ical data, how to develop them, and what 17 to expect of them,”Physiological Measure- ment, vol. 45, no. 2, pp. 020301, 2024,IOP Publishingdoi: 10.1088/1361-6579/ad252f

    2024 doi

  4. [12]

    Zhang and D

    G. Zhang and D. Choi and D. E. Kim and J. Jung, Reconstruction of ABP wave- form from the ECG or PPG signals us- ing enhanced 1-D UNetwork. Presented at 2023 IEEE/ACIS 23rd International Conference on Computer and Informa- tion Science (ICIS). [Online]. Available: 10.1109/ICIS...

    2023

  5. [13]

    Ro- bust modelling of arterial blood pressure re- construction from photoplethysmography,

    J. Pan and L. Liang and Y. Liang and Q. Tang and Z. Chen and J. Zhu, “Ro- bust modelling of arterial blood pressure re- construction from photoplethysmography,” Scientific Reports, vol. 14, no. 1, pp. 30333, 2024,MDPIdoi: 10.1038/s41598- 024-82026-1

    2024 doi

  6. [14]

    Cuff-less high-accuracy calibration-free blood pres- sure estimation using pulse transit time,

    M. Kachuee, M. M. Kiani, H. Moham- madzade, and M. Shabany, “Cuff-less high-accuracy calibration-free blood pres- sure estimation using pulse transit time,” inProc. 2015 IEEE Int. Symp. Circuits Syst. (ISCAS), Long Beach, CA, USA, 2015, pp. 1006–1009, doi: 10.1109/IS- CAS.2015.7168806

    2015

  7. [15]

    PulseDB: A large, cleaned dataset based on MIMIC- III and VitalDB for benchmarking cuff-less blood pressure estimation methods,

    W. Wang and P. Mohseni and K. L. Kil- gore and L. Najafizadeh, “PulseDB: A large, cleaned dataset based on MIMIC- III and VitalDB for benchmarking cuff-less blood pressure estimation methods,”Fron- tiers in Digital Health, vol. 4, 2023, doi: 10.3389/fdgth.2022.1090854

    2023

  8. [16]

    PPG2ABP: translating photoplethysmo- gram (PPG) signals to arterial blood pres- sure (ABP) waveforms,

    N. Ibtehaz and S. Mahmud and M. E. H. Chowdhury and A. Khandakar and M. Salman Khan and M. A. Ayari and A. M. Tahir and M. S. Rahman, “PPG2ABP: translating photoplethysmo- gram (PPG) signals to arterial blood pres- sure (ABP) waveforms,”Bioengineering, vol. 11, no. 9, pp. 692,...

    2022 doi

  9. [17]

    Mahmud, N

    S. Mahmud, N. Ibtehaz, A. Khandakar, M. S. Rahman, A. J. R. Gonzales, T. Rah- man,et al., “NABNet: A nested attention- guided BiConvLSTM network for a ro- bust prediction of blood pressure com- ponents from reconstructed arterial blood pressure waveforms using PPG and ECG sign...

    2023

  10. [18]

    ECG Reconstruction via PPG: A Pilot Study,

    Q. Zhu and X. Tian and C.-W. Wong and M. Wu, “ECG Reconstruction via PPG: A Pilot Study,” in2019 IEEE EMBS Interna- tional Conference on Biomedical & Health Informatics (BHI),Chicago, IL, USA, 2019, pp. 1–4, doi: 10.1109/BHI.2019.8834612

    2019

  11. [19]

    Learning your heart actions from pulse: ECG waveform reconstruction from PPG,

    Q. Zhu, X. Tian, C.-W. Wong, and M. Wu, “Learning your heart actions from pulse: ECG waveform reconstruction from PPG,” IEEE Internet of Things Journal, vol. 8, no. 23, pp. 16734–16748, Dec. 2021, doi: 10.1109/JIOT.2021.3097946

    2021

  12. [20]

    Rishi Vardhan and S

    K. Rishi Vardhan and S. Vedanth and G. Poojah and K. Abhishek and M. Ni- tish Kumar and V. Vijayaraghavan, BP- Net: efficient deep Learning for con- tinuous arterial blood pressure estima- tion using photoplethysmogram. Presented at 2021 20th IEEE International Con- ference on...

    2021

  13. [21]

    A continuous blood pres- sure estimation method using photoplethys- mography by GRNN-based model,

    Z. Li and W. He, “A continuous blood pres- sure estimation method using photoplethys- mography by GRNN-based model,”Sen- sors, vol. 21, no. 21, pp. 7207, 2021,MDPI doi: 10.3390/s21217207

    2021 doi

  14. [22]

    CardioGAN: An Attention- based Generative Adversarial Net- work for Generation of Electrocardio- grams,

    S. Dasgupta and S. Das and U. Bhat- tacharya, “CardioGAN: An Attention- based Generative Adversarial Net- work for Generation of Electrocardio- grams,” in2020 25th International Conference on Pattern Recognition (ICPR),Milan, Italy, 2021, pp. 3193–3200, doi: 10.1109/ICPR48806....

    2021

  15. [23]

    P2E-WGAN: ECG waveform synthesis from PPG with conditional Wasserstein generative adver- sarial networks,

    K. Vo, E. K. Naeini, A. Naderi, D. Jilani, A. M. Rahmani, N. Dutt,et al., “P2E-WGAN: ECG waveform synthesis from PPG with conditional Wasserstein generative adver- sarial networks,” inProc. 36th Annu. ACM Symp. Appl. Comput., Virtual Event, Re- public of Korea, 2021, pp. 1030–...

    2021

  16. [24]

    Tang and Z

    Q. Tang and Z. Chen and R. Ward and C. Menon and M. Elgendi, “PPG2ECGps: An End-to-End Subject-Specific Deep Neural Network Model for Electrocardiogram Re- construction from Photoplethysmography Signals without Pulse Arrival Time Ad- justments,”Bioengineering., vol. 10, no. 6,...

    2023 doi

  17. [25]

    Robust Reconstruction of Electrocardiogram Using Photoplethys- mography: A Subject-Based Model,

    Q. Tang and Z. Chen and Y. Guo and Y. Liang and R. Ward and C. Menon and M. Elgendi, “Robust Reconstruction of Electrocardiogram Using Photoplethys- mography: A Subject-Based Model,”Fron- tiers in Physiology, vol. 13, 2022, doi: 10.3389/fphys.2022.859763

    2022

  18. [26]

    Blood pressure morphology assessment from pho- toplethysmogram and demographic infor- mation using deep learning with attention mechanism,

    N. Aguirre and E. Grall-Ma¨ es and L. J. Cymberknop and R. L. Armentano, “Blood pressure morphology assessment from pho- toplethysmogram and demographic infor- mation using deep learning with attention mechanism,”Sensors, vol. 21, no. 6, pp. 2167, 2021,MDPIdoi: 10.3390/s21062167

    2021 doi

  19. [27]

    Imputa- tion of the continuous arterial line blood pressure waveform from non-invasive mea- surements using deep learning,

    B. L. Hill and N. Rakocz and ´A. Rudas and J. N. Chiang and S. Wang and I. Hofer and M. Cannesson and E. Halperin, “Imputa- tion of the continuous arterial line blood pressure waveform from non-invasive mea- surements using deep learning,”Scientific Reports, vol. 11, no. 1, pp...

    2021 doi

  20. [28]

    Vi- talDB, a high-fidelity multi-parameter vital signs database in surgical patients,

    H.-C. Lee and Y. Park and S. B. Yoon and S. M. Yang and D. Park and C.-W. Jung, “Vi- talDB, a high-fidelity multi-parameter vital signs database in surgical patients,”Scien- tific Data, vol. 9, no. 1, pp. 279, 2022,Na- turedoi: 10.1038/s41597-022-01411-5

    2022 doi

  21. [29]

    MIMIC-III Waveform Database Matched Subset (ver- sion 1.0),

    B. Moody and G. Moody and M. Villarroel and G. D. Clifford and I. Silva, “MIMIC-III Waveform Database Matched Subset (ver- sion 1.0),”PhysioNet, 2020. [Online]. Avail- able: https://doi.org/10.13026/c2294b

    2020 doi

  22. [30]

    StarGAN v2: Diverse Image Synthesis for Multiple Domains,

    Y. Choi and Y. Uh and J. Yoo and J. Ha, “StarGAN v2: Diverse Image Synthesis for Multiple Domains,” in2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),Seattle, W A, USA, 2020, pp. 8185–8194, doi: 10.1109/CVPR42600.2020.00821

    2020

  23. [31]

    Beyond BatchNorm: Towards a Uni- fied Understanding of Normalization in Deep Learning,

    E. S. Lubana, R. P. Dick, and H. Tanaka, “Beyond BatchNorm: Towards a Uni- fied Understanding of Normalization in Deep Learning,” inProc. 35th Int. Conf. 19 Neural Inf. Process. Syst. (NeurIPS), Red Hook, NY, USA: Curran Asso- ciates Inc., 2021, Art. no. 365, pp. 1–14. [Online...

    2021

  24. [32]

    Unifying het- erogeneous electronic health records systems via text-based code embed- ding,

    K. Hur, J. Lee, J. Oh, W. Price, Y. Kim, and E. Choi, “Unifying het- erogeneous electronic health records systems via text-based code embed- ding,” inProc. Conf. Health, Inference, and Learn. (CHIL), vol. 174, PMLR, pp. 183–203, Apr. 2022. [Online]. Available: https://proceedi...

    2022

  25. [33]

    GenHPF: General healthcare predictive framework for multi-task multi- source learning,

    K. Hur, J. Oh, J. Kim, J. Kim, M. J. Lee, E. Cho,et al., “GenHPF: General healthcare predictive framework for multi-task multi- source learning,”IEEE J. Biomed. Health Inform., vol. 28, no. 1, pp. 502–513, Jan. 2024, doi: 10.1109/JBHI.2023.3327951

    2024

  26. [34]

    End- to-end personalized cuff-less blood pres- sure monitoring using ECG and PPG sig- nals,

    S. B. N., R. S. Srinivasa, Y. M. Saidutta, J. Cho, C.-H. Lee, C. Yang,et al., “End- to-end personalized cuff-less blood pres- sure monitoring using ECG and PPG sig- nals,” inProc. IEEE Int. Conf. Acoust., Speech Signal Process. (ICASSP), Seoul, South Korea, 2024, pp. 2101–2105...

    2024

  27. [35]

    TSMixer: Lightweight MLP-Mixer Model for Multi- variate Time Series Forecasting,

    V. Ekambaram, A. Jati, N. Nguyen, P. Sinthong, and J. Kalagnanam, “TSMixer: Lightweight MLP-Mixer Model for Multi- variate Time Series Forecasting,” inPro- ceedings of the 29th ACM SIGKDD Con- ference on Knowledge Discovery and Data Mining,Long Beach, CA, USA, 2023, pp. 459-–4...

    2023

  28. [36]

    A Shallow U-Net Architec- ture for Reliably Predicting Blood Pressure (BP) from Photoplethysmogram (PPG) and Electrocardiogram (ECG) Signals,

    S. Mahmud and N. Ibtehaz and A. Khan- dakar and A. M. Tahir and T. Rahman and K. R. Islam and M. S. Hossain and M. S. Rahman and F. Musharavati and M. A. Ayari and M. T. Islam and M. E. H. Chowdhury, “A Shallow U-Net Architec- ture for Reliably Predicting Blood Pressure (BP) f...

    2022 doi

  29. [37]

    A Time Series is Worth 64 Words: Long-term Forecast- ing with Transformers,

    Y. Nie and N. H. Nguyen and P. Sinthong and J. Kalagnanam, “A Time Series is Worth 64 Words: Long-term Forecast- ing with Transformers,” inProceedings of the Eleventh International Conference on Learning Representations (ICLR), 2023. [Online]. Available:https://openreview. net...

    2023

  30. [38]

    SAPTSTA-AnoECG: A PatchTST-based ECG anomaly detection method with subtractive attention and data augmentation,

    Y. Li, M. Wang, M. Guan, C. Lu, Z. Li, and T. Chen, “SAPTSTA-AnoECG: A PatchTST-based ECG anomaly detection method with subtractive attention and data augmentation,”Appl. Intell., vol. 55, no. 3, Art. no. 184, 2025, doi: 10.1007/s10489- 024-05881-5

    2025 doi

  31. [39]

    The British Hypertension Society protocol for the evaluation of automated and semi-automated blood pressure measur- ing devices with special reference to ambu- latory systems,

    E. O’Brien, J. Petrie, W. Littler, M. de Swiet, P. L. Padfield, K. O’Malley, M. Jamieson, D. Altman, M. Bland, and N. Atkins, “The British Hypertension Society protocol for the evaluation of automated and semi-automated blood pressure measur- ing devices with special reference...

    1990 doi

  32. [40]

    Na- tional standard for measurement of resting and ambulatory blood pressures with au- tomated sphygmomanometers,

    W. B. White and A. S. Berson and C. Rob- bins and M. J. Jamieson and L. M. Prisant 20 and E. Roccella and S. G. Sheps, “Na- tional standard for measurement of resting and ambulatory blood pressures with au- tomated sphygmomanometers,”Hyperten- sion, vol. 21, no. 4, pp. 504–509...

    1993 doi

  33. [41]

    NeuroKit2: A Python toolbox for neurophysiological sig- nal processing,

    D. Makowski, T. Pham, Z. J. Lau, J. C. Brammer, F. Lespinasse, H. Pham, C. Sch¨ olzel, and S. A. Chen, “NeuroKit2: A Python toolbox for neurophysiological sig- nal processing,”Behav. Res. Methods, vol. 53, no. 4, pp. 1689–1696, Aug. 2021, doi: 10.3758/s13428-020-01516-y

    2021 doi

  34. [42]

    pyPPG: a Python toolbox for comprehensive photoplethysmography sig- nal analysis,

    M. ´A. Goda and P. H. Charlton and J. A. Behar, “pyPPG: a Python toolbox for comprehensive photoplethysmography sig- nal analysis,”Physiological Measurement, vol. 45, no. 4, pp. 045001, Apr. 2024,IOP Publishingdoi: 10.1088/1361-6579/ad33a2

    2024 doi

  35. [43]

    Swin- Unet: Unet-Like Pure Transformer for Med- ical Image Segmentation,

    H. Cao, Y. Wang, J. Chen, D. Jiang, X. Zhang, Q. Tian, and M. Wang, “Swin- Unet: Unet-Like Pure Transformer for Med- ical Image Segmentation,” inComputer Vi- sion – ECCV 2022 Workshops, Tel Aviv, Is- rael, Oct. 2022, Lecture Notes in Computer Science, vol. 13841, Springer, 202...

    2022 doi

  36. [44]

    DistilBERT, a distilled ver- sion of BERT: smaller, faster, cheaper and lighter,

    V. Sanh and L. Debut and J. Chaumond and T. Wolf, “DistilBERT, a distilled ver- sion of BERT: smaller, faster, cheaper and lighter,” arXiv, 2020

    2020

  37. [45]

    U-Net: Convolutional Networks for Biomedical Image Segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” in Medical Image Computing and Computer- Assisted Intervention – MICCAI 2015, Lec- ture Notes in Computer Science, vol. 9351, Cham: Springer, 2015, pp. 234–241, doi: 10....

    2015 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.