Pith. sign in

REVIEW 4 major objections 5 minor 17 references

Synthetic Time Series Data Generation for Healthcare Applications: A PCG Case Study

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that WaveNet, DoppelGANger, and DiffWave can generate synthetic phonocardiogram signals that are statistically similar to real normal heart sounds, with low distribution-distance scores and a classifier that cannot…

desk verdict Routine application of three generative models to PCG signals, but the fidelity claim is unsupported because the metrics are computed on training data. read the letter →

arxiv 2412.16207 v1 pith:ZN37XA45 submitted 2024-12-17 cs.LG cs.CEeess.SP

Ainaz Jamshidi , Muhammad Arif , Sabir Ali Kalhoro , Alexander Gelbukh This is my paper · ORCID
classification cs.LGcs.CEeess.SP
keywords synthetictimeseriesgenerationphonocardiogramPCGgenerativemodelsWaveNetDoppelGANgerDiffmaximummeandiscrepancy
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

This paper tries to establish that three generative models from different families — an autoregressive network (WaveNet), a GAN (DoppelGANger), and a diffusion model (DiffWave) — can synthesize realistic phonocardiogram (PCG) recordings of normal heart sounds. Using heart-sound data from a public pediatric challenge dataset, the authors filter and segment the signals, train the models, and compare real versus synthetic segments with distribution-distance metrics and a trained discriminator. They report low maximum mean discrepancy and Jensen-Shannon divergence values and near-chance discriminative accuracy, which they take as evidence that the synthetic data closely resemble the original recordings. If true, this would make synthetic PCG data a practical resource for cardiac pre-screening and for augmenting scarce abnormal-heart-sound datasets.

What carries the argument

The argument is carried by three generative architectures paired with a fidelity-evaluation protocol. WaveNet uses stacked dilated causal convolutions with gated activations to predict future waveform samples; DoppelGANger uses two generators — one for static features and one for time series — plus a discriminator that sees both; DiffWave learns to reverse a forward process that adds Gaussian noise, reconstructing audio from noise. The evaluation machinery compares real and synthetic segments through maximum mean discrepancy in a reproducing kernel Hilbert space, Jensen-Shannon divergence between the two distributions, t-SNE overlap, and the accuracy of a small recurrent classifier trained to tell the sources apart. Low distances and near-chance classifier accuracy are the evidence that the synthetic segments sit close to the real data.

What would settle it

Retrain the three models on one group of patients and evaluate on a separate held-out group of patients; if the distance scores rise substantially or the classifier clearly separates real from synthetic, the claim that the generated PCG data closely resembles the original data would be falsified.

Watch

Extended reading notes

Core claim

The central claim is that generative models can produce PCG segments that are statistically indistinguishable from real normal heart sounds under the tested measures. For WaveNet, the paper reports forecasting errors of MAE 0.02, MSE 0.002, SMAPE 3.51, and ACD 0.006 on the last 20% of the time series. For DoppelGANger and DiffWave, it reports JSD values of 0.014 and 0.015 and MMD values of 0.0001 for both, with a 20-unit RNN discriminator reaching only 52% and 51% accuracy when trying to label real versus synthetic segments. The authors interpret these numbers as evidence that the generated data closely mirrors the distribution of the original PCG signals and that the models are effective at producing realistic synthetic heart-sound data.

Load-bearing premise

The results assume that the models were tested on recordings they had not seen during training, and that the simple 20-unit classifier would have been able to detect any real differences between real and synthetic data if they existed.

Editorial extensions

If this is right

  • If the fidelity result holds, synthetic normal PCG segments can be produced in arbitrary volume without new patient recordings.
  • The same evaluation metrics give future work a ready benchmark for comparing new generative models on heart-sound data.
  • Synthetic normal segments could serve as the base class in a data-augmentation pipeline, which is the paper's stated next step toward detecting abnormal heart sounds and murmurs.
  • Sharing synthetic recordings instead of raw patient data becomes more plausible if the synthetic distribution is close enough for downstream diagnostic use.

Reading between the lines

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

  • The near-chance discriminative score depends entirely on the capacity of the 20-unit classifier; a stronger discriminator might separate the distributions, so the fidelity claim should be tested with a larger or convolutional classifier before relying on it.
  • If the fidelity holds on properly held-out patients, the same approach could be extended to conditional generation of murmur and abnormal classes, though those classes are rarer and more variable and will likely require conditioning or class-balanced training.
  • A stricter test would compare raw waveform statistics, such as power spectra or S1/S2 timing, rather than only the distance scores, which may not capture clinically meaningful features.
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

4 major / 5 minor

Summary. The paper compares three generative modeling approaches—WaveNet, DoppelGANger, and DiffWave—for synthesizing normal phonocardiogram (PCG) segments from the normal-subject subset of the PhysioNet Challenge 2022 dataset. The authors preprocess the recordings into 110-unit segments, train each model, and evaluate quality with forecasting metrics (MAE, MSE, SMAPE, ACD) for WaveNet and distributional metrics (t-SNE, discriminative score, MMD, JSD) for DGAN and DiffWave. The central claim is that all three models produce synthetic PCG data closely resembling the original data, supported by low MMD/JSD values and near-chance discriminative accuracy.

Significance. If the central claim were established, the paper would provide a useful comparative benchmark for raw-waveform PCG synthesis and a step toward data augmentation for murmur detection. The work uses a relevant public dataset, applies standard time-series generation metrics, and documents the preprocessing pipeline. However, because the evaluation does not separate training and test data for the generative models, and because WaveNet is evaluated only as a forecaster rather than as a full-sequence generator, the current evidence does not support the claimed fidelity. The paper would be strengthened by a held-out evaluation, a stronger discriminative classifier, and uncertainty quantification; these are fixable within the manuscript's scope.

major comments (4)
  1. [VI-B, VI-C, V-B] No held-out split is described for DGAN or DiffWave. The MMD=0.0001 and JSD=0.014/0.015 values in Tables II and III are computed between generated samples and the real segments that were used to train the models. If the models memorized training segments, these metrics and the t-SNE overlap in Figures 2 and 3 would appear equally good even though the models have not learned the distribution of unseen PCG signals. The authors should split the real segments into training and held-out sets, train only on the former, and report MMD, JSD, and discriminative scores against the held-out real segments, ideally alongside a memorization baseline such as resampled training segments. Until this is done, the central claim that generated data closely resembles the original dataset is unsupported.
  2. [III-A, VI-A] WaveNet is evaluated as a forecaster rather than as a whole-signal generator: it predicts the last 20% of each sequence, and Table I reports MAE=0.02, MSE=0.002, SMAPE=3.51, and ACD=0.006. Forecasting accuracy on tail portions does not measure whether WaveNet can synthesize a complete PCG segment from scratch, so the abstract's claim that WaveNet generates PCG data is not supported by these numbers. The authors should either generate full sequences by ancestral sampling from WaveNet and evaluate them with the same generative metrics used for the other two models, or explicitly reframe the WaveNet part as a forecasting baseline and remove it from the generative comparison.
  3. [V-B2, VI-B2, VI-C2] The discriminative score uses a single-layer RNN with 20 hidden units, trained for 50 epochs in Section V but for 100 epochs in Section VI, and the reported near-chance accuracies (52% and 51%) are point estimates without error bars. A weak classifier can fail to separate real and synthetic data even when the distributions are clearly different, so chance-level accuracy is not evidence of high fidelity. The authors should use a higher-capacity discriminator, report train and validation accuracy, include standard deviations over several random seeds, and resolve the inconsistent epoch count between Sections V and VI.
  4. [Tables I-III] All reported metrics are single point estimates with no confidence intervals or standard deviations across runs or bootstrap resamples. With only about 110 real segments, MMD values of 0.0001 and JSD values around 0.014 need uncertainty quantification before the claim of "closely resembles" can be assessed; the discriminative accuracy of 51-52% is also within random fluctuation for a small test set. Please report multiple-seed or bootstrap intervals for every metric.
minor comments (5)
  1. [V-A4] The ACD metric is described only verbally as a coverage difference, and no formula is given in Section V-A4, while Table I reports ACD=0.006. Provide the exact computation or remove the metric.
  2. [I, II] The Introduction states that no existing studies focus on generating high-quality PCG signals using generative models, but Section II cites Takezaki et al. [6], which applies a GAN to heart sound spectrogram data. Qualify the novelty claim to raw waveform synthesis or to the specific comparison performed here.
  3. [IV-B] The thresholds RMSSD<0.1, zero-crossing ratio<0.3, and noise-free window ratio≥50% are introduced as "my experimental analysis" without a derivation or citation; describe how they were established and whether they were tuned on the same data.
  4. [III-A, Table I] The metric name "SAMPE" appears in Table I; this should be "SMAPE" as defined in Section V-A3.
  5. [Authors' affiliations] The affiliation for Alexander Gelbukh lists "New Mexico, USA" for Instituto Politécnico Nacional, which is located in Mexico; correct this error.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the generative models and metrics are external/standard, and the central fidelity claim is an in-sample report rather than a self-referential derivation.

full rationale

The paper's derivation chain uses three externally published generative models (WaveNet, DoppelGANger, DiffWave) and standard evaluation metrics (MAE, MSE, SMAPE, ACD, t-SNE, discriminative score, MMD, JSD). There are no author self-citations, no imported uniqueness theorems, and no ansatz smuggled through a citation. WaveNet is evaluated with a genuine forecast split on the last 20% of each series, so those numbers are not in-sample. For DGAN and DiffWave, the MMD/JSD/t-SNE comparisons are computed against the same real segments used for training because the paper describes no train/test split for the generative models; however, the low scores are not guaranteed by the training objectives, and the paper does not fit any parameter directly to the evaluation metric or define the target conclusion as the metric value. The fidelity claim is therefore an in-sample validation that would need a held-out real set to support generalization to new PCG recordings, but that limitation is an experimental validity gap rather than a circular derivation. No quoted equation reduces to its own input, so no circular step meets the required standard.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim depends on several hand-chosen preprocessing thresholds and model capacities, and the evaluation does not separate training from test data, so the reported similarity scores may reflect in-sample fitting.

free parameters (5)
  • RMSSD quality threshold = 0.1
    Adopted from the quality assessment protocol in [13]; signals with RMSSD at or above 0.1 are discarded as corrupted. This filtering changes the input distribution for all generative models.
  • Zero crossing ratio threshold = 0.3
    From [13]; signals with zero-crossing ratio above 0.3 are marked noisy and removed. The threshold affects which PCG recordings are retained.
  • Noise-free window ratio threshold = 50%
    At least 50 percent of a signal's windows must be noise-free for inclusion; this further filters the training set.
  • Segment length = 110 units
    After peak localization and normalization, each segment is cut to 110 samples. All models operate on this fixed length, and the measured distributions depend on this choice.
  • Discriminative classifier hidden units = 20
    The RNN classifier has a single 20-hidden-node layer, which is likely too weak to distinguish real from synthetic data, contributing to the near-chance accuracy.
assumptions (4)
  • domain assumption The PhysioNet 2022 dataset labels identify normal subjects correctly and the expert clinical outcome labels are accurate.
    The study selects normal subjects using these labels and does not independently verify them.
  • domain assumption The PCG quality assessment protocol from Akram et al. [13] is appropriate for this dataset and its thresholds produce clean signals.
    The authors apply the protocol without adapting thresholds to the Challenge 2022 data; different thresholds would change the training set.
  • standard math WaveNet, DoppelGANger, and DiffWave, as cited, are valid generative models whose training produces samples from the target distribution under the given hyperparameters.
    The authors rely on the published methods and do not justify their specific architectural choices (e.g., 89 hidden channels, 199 skip channels).
  • domain assumption The S1 peak detection algorithm correctly localizes systolic peaks in the filtered signals.
    Segments are centered on detected S1 peaks; detection errors would corrupt the generated data and the evaluation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Synthetic Time Series Data Generation for Healthcare Applications: A PCG Case Study." pith.science (2026). https://pith.science/paper/ZN37XA45

@misc{pith2026241216207,
  author       = {Pith},
  title        = {Pith review of: Synthetic Time Series Data Generation for Healthcare Applications: A PCG Case Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZN37XA45}},
  note         = {Machine review of arXiv:2412.16207}
}
read the original abstract

The generation of high-quality medical time series data is essential for advancing healthcare diagnostics and safeguarding patient privacy. Specifically, synthesizing realistic phonocardiogram (PCG) signals offers significant potential as a cost-effective and efficient tool for cardiac disease pre-screening. Despite its potential, the synthesis of PCG signals for this specific application received limited attention in research. In this study, we employ and compare three state-of-the-art generative models from different categories - WaveNet, DoppelGANger, and DiffWave - to generate high-quality PCG data. We use data from the George B. Moody PhysioNet Challenge 2022. Our methods are evaluated using various metrics widely used in the previous literature in the domain of time series data generation, such as mean absolute error and maximum mean discrepancy. Our results demonstrate that the generated PCG data closely resembles the original datasets, indicating the effectiveness of our generative models in producing realistic synthetic PCG data. In our future work, we plan to incorporate this method into a data augmentation pipeline to synthesize abnormal PCG signals with heart murmurs, in order to address the current scarcity of abnormal data. We hope to improve the robustness and accuracy of diagnostic tools in cardiology, enhancing their effectiveness in detecting heart murmurs.

Figures

Figures reproduced from arXiv: 2412.16207 by the authors.

Figure 1
Figure 1. The distribution of the class labels in our dataset. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. t-SNE analysis on the distributions of real and synthetic [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. t-SNE analysis on the distributions of real and syn [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 9 canonical work pages

  1. [6]

    Data augmentation and the improvement of the performance of convolutional neural networks for heart sound classification,

    S. Takezaki and K. Kishida, “Data augmentation and the improvement of the performance of convolutional neural networks for heart sound classification,” IAENG Int. J. Comput. Sci , vol. 49, no. 4, 2022

    work page 2022

  2. [1]

    Real-valued (medical) time series generation with recurrent conditional gans,

    C. Esteban, S. L. Hyland, and G. R ¨atsch, “Real-valued (medical) time series generation with recurrent conditional gans,” arXiv preprint arXiv:1706.02633, 2017

    arXiv 2017

  3. [2]

    Synsiggan: Generative adversarial networks for synthetic biomedical signal generation,

    D. Hazra and Y .-C. Byun, “Synsiggan: Generative adversarial networks for synthetic biomedical signal generation,” Biology, vol. 9, no. 12, p. 441, 2020

    work page 2020

  4. [3]

    Corgan: correlation-capturing convolutional generative adversarial networks for generating synthetic healthcare records,

    A. Torfi and E. A. Fox, “Corgan: correlation-capturing convolutional generative adversarial networks for generating synthetic healthcare records,” arXiv preprint arXiv:2001.09346 , 2020

    arXiv 2001

  5. [4]

    Differentially private synthetic medical data generation using convolutional gans,

    A. Torfi, E. A. Fox, and C. K. Reddy, “Differentially private synthetic medical data generation using convolutional gans,”Information Sciences, vol. 586, pp. 485–500, 2022

    work page 2022

  6. [5]

    Dp-ctgan: Differentially pri- vate medical data generation using ctgans,

    M. L. Fang, D. S. Dhami, and K. Kersting, “Dp-ctgan: Differentially pri- vate medical data generation using ctgans,” in International Conference on Artificial Intelligence in Medicine . Springer, 2022, pp. 178–188

    work page 2022

  7. [7]

    Wavenet: A generative model for raw audio,

    A. van den Oord, S. Dieleman, H. Zen, K. Simonyan, O. Vinyals, A. Graves, N. Kalchbrenner, A. Senior, and K. Kavukcuoglu, “Wavenet: A generative model for raw audio,” arXiv preprint arXiv:1609.03499 , 2016

    arXiv 2016

  8. [8]

    Generating high- fidelity, synthetic time series datasets with doppelganger,

    Z. Lin, A. Jain, C. Wang, G. Fanti, and V . Sekar, “Generating high- fidelity, synthetic time series datasets with doppelganger,” arXiv preprint arXiv:1909.13403, 2019

    arXiv 1909

Show all 17 references
  1. [9]

    Dif- fwave: A versatile diffusion model for audio synthesis,

    Z. Kong, W. Ping, J. Huang, K. Zhao, and B. Catanzaro, “Dif- fwave: A versatile diffusion model for audio synthesis,” arXiv preprint arXiv:2009.09761, 2020

    2009 arXiv

  2. [10]

    PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals,

    A. L. Goldberger, L. A. N. Amaral, L. Glass, J. M. Hausdorff, P. C. Ivanov, R. G. Mark, J. E. Mietus, G. B. Moody, C.-K. Peng, and H. E. Stanley, “PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals,”Circulation, vol....

    doi

  3. [11]

    The circor digiscope dataset: From murmur detection to murmur classification,

    J. Oliveira, F. Renna, P. D. Costa, M. Nogueira, C. Oliveira, C. Ferreira, A. Jorge, S. Mattos, T. Hatem, T. Tavares, A. Elola, A. B. Rad, R. Sameni, G. D. Clifford, and M. T. Coimbra, “The circor digiscope dataset: From murmur detection to murmur classification,” IEEE Journal...

    2022

  4. [12]

    Heart murmur detection from phonocardiogram recordings: The george b. moody physionet challenge 2022,

    M. A. Reyna, Y . Kiarashi, A. Elola, J. Oliveira, F. V . Renna, A. P. Gu, E. A. Perez-Alday, N. Sadr, A. Sharma, S. Mattos, M. T. Coimbra, R. Sameni, A. B. Rad, and G. D. Clifford, “Heart murmur detection from phonocardiogram recordings: The george b. moody physionet challenge...

    2022

  5. [13]

    Analysis of pcg signals using quality assessment and homomorphic filters for localization and classification of heart sounds,

    M. U. Akram, A. Shaukat, F. Hussain, S. G. Khawaja, W. H. Butt et al., “Analysis of pcg signals using quality assessment and homomorphic filters for localization and classification of heart sounds,” Computer methods and programs in biomedicine , vol. 164, pp. 143–157, 2018

    2018

  6. [14]

    A. V . Oppenheim, Discrete-time signal processing. Pearson Education India, 1999

    1999

  7. [15]

    SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python,

    P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey,˙I. Polat, Y . Feng...

    2020

  8. [16]

    Transfusion: Generating long, high fidelity time series using diffusion models with transformers,

    M. F. Sikder, R. Ramachandranpillai, and F. Heintz, “Transfusion: Generating long, high fidelity time series using diffusion models with transformers,” arXiv e-prints, pp. arXiv–2307, 2023

    2023

  9. [17]

    Time-series generative ad- versarial networks,

    J. Yoon, D. Jarrett, and M. Van der Schaar, “Time-series generative ad- versarial networks,” Advances in neural information processing systems, vol. 32, 2019

    2019

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.