arxiv: 2604.12934 · v1 · submitted 2026-04-14 · 📡 eess.IV

A Wearable ECG Device for Differentiating Hypertrophic Cardiomyopathy from Acquired Left Ventricular Hypertrophy

Jiachen Li , Hanyu Zhu , Edward Kim , Shihao Li , Katherine Cavanaugh , Arpan Patel , Sovik De Sirkar , Mauricio Hong

show 2 more authors

Wei Li Dongmei Chen

This is my paper

Pith reviewed 2026-05-10 14:05 UTC · model grok-4.3

classification 📡 eess.IV

keywords hypertrophic cardiomyopathywearable ECGleft ventricular hypertrophyECG classificationHCM screeningsignal indicesportable cardiac device

0 comments

The pith

A wearable ECG device extracts two indices from each heartbeat to separate genetic hypertrophic cardiomyopathy from acquired left ventricular hypertrophy.

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

The paper introduces a portable three-lead ECG system built from standard components that records heart signals and feeds them into an algorithm calculating two specific quantitative indices per heartbeat. These indices are compared against fixed statistical thresholds to label a recording as HCM or acquired LVH. The approach matters because HCM is a leading cause of sudden cardiac death in young athletes yet current confirmation requires costly imaging or genetic tests that are unavailable in many settings. Validation against hundreds of LVH records and dozens of HCM records produces high specificity with moderate sensitivity, and the authors show the separation survives leave-one-out testing and is not explained by how the records were obtained. The result is presented as a practical first-step screening tool that could be deployed where full diagnostic equipment is absent.

Core claim

The wearable device paired with the dual-index classification algorithm distinguishes HCM from acquired LVH using ECG signals alone, achieving 75.86 percent sensitivity and 99.17 percent specificity on 483 LVH cases from PhysioNet plus 29 HCM cases from digitized records, with leave-one-out cross-validation yielding 72.41 percent sensitivity and 98.96 percent specificity and a digitization-confound check confirming the separation is physiological rather than artifactual.

What carries the argument

Two quantitative indices (HCM Index 1 and HCM Index 2) extracted from each heartbeat and compared against dual statistical thresholds.

If this is right

The device can function as an affordable initial screen in clinics or regions lacking access to CMR or echocardiography.
High specificity means most patients without HCM would not be referred unnecessarily for further testing.
Leave-one-out validation indicates the thresholds are not overfitted to the particular patient split.
Hardware simulation confirms the Arduino-based acquisition chain produces signals compatible with the index calculations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the indices remain stable across larger digital HCM cohorts, they could be added as a software feature to existing consumer ECG wearables for population-level alerts.
The method might be tested on other ECG-detectable distinctions between genetic and acquired cardiac conditions to see whether the same two-index structure generalizes.
A prospective study collecting simultaneous wearable and clinical-grade ECGs on the same HCM patients would directly measure how much signal quality affects index values.

Load-bearing premise

The two indices and their thresholds reflect real physiological differences between HCM and acquired LVH rather than artifacts from digitizing clinical records or the small number of HCM cases.

What would settle it

Running the identical indices and thresholds on a fresh set of at least 50 HCM patients whose ECGs were recorded directly in digital form and checking whether sensitivity stays near or above 70 percent.

read the original abstract

Hypertrophic Cardiomyopathy (HCM) is a genetic heart disease affecting approximately 1 in 500 people and is the leading cause of sudden cardiac death in young athletes. Current diagnostic methods -- cardiovascular magnetic resonance (CMR), echocardiography, and genetic testing -- are limited by high costs, operator dependency, or insufficient accuracy, while standard electrocardiogram (ECG) analysis cannot reliably distinguish HCM from acquired left ventricular hypertrophy (LVH). This paper presents a wearable ECG device paired with a classification algorithm that differentiates HCM from acquired LVH using ECG signals alone. The portable device integrates a 3-lead electrode system, an AD8232 signal conditioning module, an Arduino Nano 33 BLE microcontroller, and a lithium polymer battery. The algorithm extracts two quantitative indices -- HCM Index~1 and HCM Index~2 -- from each heartbeat and classifies patients via dual statistical thresholds. Validation on 483 LVH patients (PhysioNet) and 29 HCM patients (digitized clinical records) yields 75.86\% sensitivity, 99.17\% specificity, and an F1-score of 80.00\%. Leave-one-out cross-validation confirms generalizability, with cross-validated sensitivity of 72.41\%, specificity of 98.96\%, and F1-score of 76.36\% (95\% confidence intervals reported). A digitization confound analysis demonstrates that the classification is driven by physiological cardiac features rather than data source artifacts. A simulated device acquisition chain analysis confirms that the wearable hardware's signal characteristics are compatible with the classification algorithm. The system offers a promising tool for affordable HCM screening in resource-limited settings.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper builds a practical 3-lead wearable ECG with two new indices for HCM versus acquired LVH, but the tiny HCM sample and thresholds tuned on the same data keep the results preliminary.

read the letter

The paper describes a 3-lead wearable ECG built from off-the-shelf parts and two new quantitative indices that separate HCM from acquired LVH on ECG alone. Standard ECG rules do not do this reliably, so the hardware-plus-indices idea is the main addition here. The authors also run a digitization check and a simulated device chain to argue the separation comes from cardiac features rather than recording differences. Those steps are useful and show some care with the data sources. The reported performance on 483 LVH cases from PhysioNet and 29 HCM cases from digitized records reaches 75.86% sensitivity and 99.17% specificity, with LOOCV numbers close behind. That is the concrete output. The soft spots sit right where the stress-test note flags them. The HCM group is only 29 patients, all from old clinical records rather than fresh device recordings. The two indices and their dual thresholds are derived from the same dataset used to measure performance, so the numbers depend on parameters fitted to these particular cases. The class imbalance is extreme, and no independent prospective test on new patients wearing the actual device is shown. The confound analysis addresses some distribution shift but does not remove the core risk that the separation may not hold outside this collection. This work is for engineers and clinicians who build low-cost ECG screening tools, especially for settings without easy access to imaging. Readers who need a ready-to-deploy clinical device will see the validation gaps immediately. The paper shows clear thinking about the clinical problem and the hardware constraints, with honest engagement on the ECG limitations in the literature. It deserves a serious referee to press on cohort size and external testing. I would send it for peer review rather than desk reject, with the expectation that reviewers will require larger independent HCM data and prospective device recordings before any stronger claims.

Referee Report

3 major / 2 minor

Summary. The manuscript presents a wearable 3-lead ECG device (integrating AD8232, Arduino Nano 33 BLE, and battery) paired with an algorithm that extracts two custom indices (HCM Index 1 and HCM Index 2) from heartbeat signals and applies dual statistical thresholds to classify hypertrophic cardiomyopathy (HCM) versus acquired left ventricular hypertrophy (LVH). It reports validation results of 75.86% sensitivity, 99.17% specificity, and 80.00% F1-score on 483 LVH cases from PhysioNet and 29 HCM cases from digitized clinical records, supported by leave-one-out cross-validation (72.41% sensitivity, 98.96% specificity, 76.36% F1), a digitization confound analysis, and a simulated device acquisition chain analysis.

Significance. If the indices and thresholds prove to capture stable physiological differences that generalize beyond the current cohorts, the work could enable low-cost, portable HCM screening in resource-limited settings where CMR, echocardiography, and genetic testing are inaccessible. The LOOCV, digitization confound analysis, and hardware simulation are explicit strengths that partially address reproducibility and real-world applicability concerns.

major comments (3)

[Results] Results section (performance metrics and cohort description): The HCM cohort contains only 29 patients, so the reported sensitivity of 75.86% rests on roughly 22 true positives; combined with thresholds fitted to this same small set, the metrics are fragile and the central claim of reliable differentiation requires stronger evidence of stability.
[Methods] Methods section (index definitions): The explicit mathematical formulas for HCM Index 1 and HCM Index 2 are not supplied, so it is impossible to verify how the indices are computed from the ECG waveform or to reproduce the classification pipeline independently.
[Validation] Validation and Results sections (threshold derivation): The two classification thresholds are derived statistically from the validation dataset itself (the same 29 HCM + 483 LVH cases used for performance reporting), creating circularity that LOOCV only partially mitigates; this directly undermines the claim that the indices capture true physiological differences rather than dataset-specific separation.

minor comments (2)

[Abstract] Abstract and Results: The abstract states that 95% confidence intervals are reported, but the numerical intervals are not shown in the provided text; include them explicitly alongside the cross-validated metrics.
[Methods] Figure captions and Methods: Ensure all ECG preprocessing steps (filtering, R-peak detection) are described with sufficient detail and parameter values so that the index extraction can be replicated from raw signals.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the detailed and constructive comments. We address each of the major comments below and have revised the manuscript to incorporate clarifications and additional details where possible.

read point-by-point responses

Referee: [Results] Results section (performance metrics and cohort description): The HCM cohort contains only 29 patients, so the reported sensitivity of 75.86% rests on roughly 22 true positives; combined with thresholds fitted to this same small set, the metrics are fragile and the central claim of reliable differentiation requires stronger evidence of stability.

Authors: We acknowledge the limitation posed by the small HCM cohort size of 29 patients. This is a recognized challenge in studying rare conditions such as HCM. The sensitivity of 75.86% is based on 22 true positives, and we have now included 95% confidence intervals (as mentioned in the abstract) to better reflect the uncertainty in the estimates. The LOOCV results (72.41% sensitivity) offer an internal validation of stability. We have added a dedicated Limitations subsection in the Discussion to emphasize the need for larger, prospective validation studies. However, expanding the cohort is not feasible within the current work due to data access constraints. revision: partial
Referee: [Methods] Methods section (index definitions): The explicit mathematical formulas for HCM Index 1 and HCM Index 2 are not supplied, so it is impossible to verify how the indices are computed from the ECG waveform or to reproduce the classification pipeline independently.

Authors: We apologize for this oversight in the original manuscript. The explicit mathematical formulas for HCM Index 1 and HCM Index 2 have been added to the Methods section. These formulas describe the computation based on specific ECG features, such as normalized amplitudes and time intervals between R-peaks and other fiducial points, allowing for full reproducibility of the pipeline. revision: yes
Referee: [Validation] Validation and Results sections (threshold derivation): The two classification thresholds are derived statistically from the validation dataset itself (the same 29 HCM + 483 LVH cases used for performance reporting), creating circularity that LOOCV only partially mitigates; this directly undermines the claim that the indices capture true physiological differences rather than dataset-specific separation.

Authors: The referee raises a valid point regarding potential circularity in threshold selection. The thresholds were determined using statistical criteria (e.g., based on means and standard deviations from the LVH and HCM groups) on the full dataset to achieve clear separation. To address this, the LOOCV was implemented such that for each iteration, the thresholds are re-estimated excluding the left-out sample, and performance is evaluated accordingly. We have revised the Validation section to provide a more detailed description of this process and to explicitly discuss the limitations of using internal validation. While this does not fully eliminate the issue, the additional analyses (digitization confound and hardware simulation) support that the separation is physiologically driven. We agree that external validation would strengthen the claims but is outside the current scope. revision: partial

standing simulated objections not resolved

The small size of the HCM cohort (29 patients) limits the strength of evidence for stability, and we are unable to acquire additional cases for this revision.

Circularity Check

1 steps flagged

HCM indices and dual statistical thresholds are derived from the same 29+483 patient cohort used for validation and LOOCV reporting

specific steps

fitted input called prediction [Abstract (algorithm + validation paragraph); implied in methods for index extraction and threshold setting]
"The algorithm extracts two quantitative indices -- HCM Index~1 and HCM Index~2 -- from each heartbeat and classifies patients via dual statistical thresholds. Validation on 483 LVH patients (PhysioNet) and 29 HCM patients (digitized clinical records) yields 75.86% sensitivity, 99.17% specificity, and an F1-score of 80.00%. Leave-one-out cross-validation confirms generalizability, with cross-validated sensitivity of 72.41%, specificity of 98.96%, and F1-score of 76.36%."

Statistical thresholds are necessarily computed from the same labeled cohort (mean/variance or separation-optimized cutoffs on the 29 HCM + 483 LVH cases). The quoted performance numbers and LOOCV therefore measure how well the fitted rule separates the data it was tuned on, not generalization to new patients or device-acquired signals whose distribution was not used to set the thresholds.

full rationale

The paper defines two indices and applies dual statistical thresholds whose values are computed from the validation set itself (small HCM n=29 from digitized records vs. PhysioNet LVH). Reported sensitivity/specificity and LOOCV figures therefore reflect performance after fitting the decision rule to the evaluated cases rather than on fully independent held-out data or prospective recordings. This matches the fitted-input-called-prediction pattern; the digitization-confound and hardware-simulation checks do not remove the dependence of the thresholds on the test distribution. No external benchmark or pre-specified thresholds independent of this cohort are shown.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 2 invented entities

The claim rests on two newly invented indices whose thresholds are fitted parameters plus the assumption that 3-lead ECG signals contain separable physiological features across data sources of different origins.

free parameters (2)

HCM Index 1 threshold
Statistical cut-off chosen to separate the two patient groups in the collected data
HCM Index 2 threshold
Statistical cut-off chosen to separate the two patient groups in the collected data

axioms (2)

domain assumption 3-lead ECG signals contain distinguishable features between HCM and acquired LVH
Invoked when defining the two indices and applying dual thresholds
domain assumption Digitized clinical HCM records are comparable to PhysioNet LVH recordings for feature extraction
Required for pooling the two datasets; addressed by confound analysis but not independently verified

invented entities (2)

HCM Index 1 no independent evidence
purpose: Quantitative measure extracted from each ECG heartbeat to indicate HCM
Newly defined index without prior independent validation or external benchmark
HCM Index 2 no independent evidence
purpose: Second quantitative measure extracted from each ECG heartbeat for classification
Newly defined index without prior independent validation or external benchmark

pith-pipeline@v0.9.0 · 5629 in / 1828 out tokens · 67461 ms · 2026-05-10T14:05:04.547271+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

27 extracted references

[1]

New insights and recommendations for athletes with hypertrophic cardiomyopathy,

McKinney, J., Isserow, M., Wong, J., Isserow, S., and Moulson, N., 2024, “New insights and recommendations for athletes with hypertrophic cardiomyopathy,” Canadian Journal of Cardiology,40(5), pp. 921–933

2024
[2]

HypertrophicCardiomyopathyinAthletes,

Malhotra,A.andSharma,S.,2017,“HypertrophicCardiomyopathyinAthletes,” European Cardiology Review,12(2), pp. 80–82

2017
[3]

Genetics of hypertrophic cardiomyopathy: established and emerging implications for clinical practice,

Lopes, L. R., Ho, C. Y., and Elliott, P. M., 2024, “Genetics of hypertrophic cardiomyopathy: established and emerging implications for clinical practice,” European Heart Journal,45(30), pp. 2727–2734

2024
[4]

Hypertension, leftventricularhypertrophy, and sudden cardiac death,

Shenasa, M.andShenasa, H., 2017, “Hypertension, leftventricularhypertrophy, and sudden cardiac death,” International journal of cardiology,237, pp. 60–63

2017
[5]

Hypertrophic Cardiomyopathy: A Review,

Hensley, N., Dietrich, J., Nyhan, D., Mitter, N., Yee, M.-S., and Brady, M., 2015, “Hypertrophic Cardiomyopathy: A Review,” Anesthesia & Analgesia, 120(3), pp. 554–569

2015
[6]

Clinical indications for cardiovascular magnetic resonance (CMR): Consensus Panel report,

Pennell, D. J., Sechtem, U. P., Higgins, C. B., Manning, W. J., Pohost, G. M., Rademakers, F. E., van Rossum, A. C., Shaw, L. J., and Yucel, E. K., 2004, “Clinical indications for cardiovascular magnetic resonance (CMR): Consensus Panel report,” European heart journal,25(21), pp. 1940–1965

2004
[7]

Bringing MRI to low- and middle-income countries: directions, challenges and potential solutions,

Murali, S., Ding, H., Adedeji, F., Qin, C., Obungoloch, J., Asllani, I., Anazodo, U., Ntusi, N. A., Mammen, R., Niendorf, T., et al., 2024, “Bringing MRI to low- and middle-income countries: directions, challenges and potential solutions,” NMR in Biomedicine,37(7), p. e4992

2024
[8]

Ryding, A., 2008,Essential echocardiography, Elsevier Health Sciences

2008
[9]

European Association of Echocardiography recommendations for training, competence, and quality improvement in echocardiography,

Popescu, B. A., Andrade, M. J., Badano, L. P., Fox, K. F., Flachskampf, F. A., Lancellotti, P., Varga, A., Sicari, R., Evangelista, A., Nihoyannopoulos, P., et al., 2009, “European Association of Echocardiography recommendations for training, competence, and quality improvement in echocardiography,” European Journal of Echocardiography,10(8), pp. 893–905

2009
[10]

Genetic Alliance, 2009,Understanding Genetics: A New York, Mid-Atlantic Guide for Patients and Health Professionals, Genetic Alliance, Washington, DC

2009
[11]

Genetic Testing in Patients With Hypertrophic Cardiomyopathy,

Bonaventura, J., Polakova, E., Vejtasova, V., and Veselka, J., 2021, “Genetic Testing in Patients With Hypertrophic Cardiomyopathy,” International Journal of Molecular Sciences,22(19), p. 10401

2021
[12]

Computer-aided diagnosis of atrial fibrillation based on ECG Signals: A review,

Hagiwara, Y., Fujita, H., Oh, S. L., Tan, J. H., San Tan, R., Ciaccio, E. J., and Acharya, U. R., 2018, “Computer-aided diagnosis of atrial fibrillation based on ECG Signals: A review,” Information Sciences,467, pp. 99–114

2018
[13]

State-of-the-Art Diagnosis of Myocardial Infarction,

Vafaie, M., 2016, “State-of-the-Art Diagnosis of Myocardial Infarction,” Diag- nosis,3(4), pp. 137–142

2016
[14]

Artificial intelligence–enabled classification of hypertrophic heart diseases us- ingelectrocardiograms,

Haimovich, J. S., Diamant, N., Khurshid, S., Di Achille, P., Reeder, C., Fried- man, S., Singh, P., Spurlock, W., Ellinor, P. T., Philippakis, A., et al., 2023, “Artificial intelligence–enabled classification of hypertrophic heart diseases us- ingelectrocardiograms,” CardiovascularDigitalHealthJournal,4(2), pp.48–59

2023
[15]

Real-time bradycardia pre- diction in preterm infants using a dynamic system identification approach,

Bakshi, S., Feng, T., Chen, D., and Li, W., 2020, “Real-time bradycardia pre- diction in preterm infants using a dynamic system identification approach,” Journal of Engineering and Science in Medical Diagnostics and Therapy,3(1), p. 011006

2020
[16]

A new system of multiple-lead exercise electrocardiography,

Mason, R. E. and Likar, I., 1966, “A new system of multiple-lead exercise electrocardiography,” American heart journal,71(2), pp. 196–205

1966
[17]

Review of Noise Removal Techniques in ECG Signals,

Chatterjee, S., Thakur, R. S., Yadav, R. N., Gupta, L., and Raghuvanshi, D. K., 2020, “Review of Noise Removal Techniques in ECG Signals,” IET Signal Processing,14(9), pp. 569–590

2020
[18]

Generalization of spectral flatness measure for non-gaussian linear processes,

Dubnov, S., 2004, “Generalization of spectral flatness measure for non-gaussian linear processes,” IEEE Signal Processing Letters,11(8), pp. 698–701

2004
[19]

Hastie, T., Tibshirani, R., and Friedman, J., 2009,The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd ed., Springer, New York

2009
[20]

The Use of Confidence or Fiducial Limits Illustrated in the Case of the Binomial,

Clopper, C. J. and Pearson, E. S., 1934, “The Use of Confidence or Fiducial Limits Illustrated in the Case of the Binomial,” Biometrika,26(4), pp. 404–413

1934
[21]

Efron,B.andTibshirani,R.J.,1994,AnIntroductiontotheBootstrap,Chapman and Hall/CRC, New York

1994
[22]

A large scale 12-lead electrocardiogram database for arrhythmia study (version 1.0. 0),

Zheng, J., Guo, H., and Chu, H., 2022, “A large scale 12-lead electrocardiogram database for arrhythmia study (version 1.0. 0),” PhysioNet 2022Available online httpphysionet orgcontentecg arrhythmia10 0accessed on,23, p. 7

2022
[23]

Optimalmulti- stage arrhythmia classification approach,

Zheng, J., Chu, H., Struppa, D., Zhang, J., Yacoub, S.M., El-Askary, H., Chang, A.,Ehwerhemuepha,L.,Abudayyeh,I.,Barrett,A.,etal.,2020,“Optimalmulti- stage arrhythmia classification approach,” Scientific reports,10(1), p. 2898

2020
[24]

PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals,

Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P.C., Mark, R., and Stanley, H. E., 2000, “PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals,” Circulation, 101(23), pp. e215–e220

2000
[25]

Left ventricular hypertrophy is associated with obe- sity, male gender, and symptoms in healthy adolescents,

Movahed, M. R., Martinez, A., Greaves, J., Greaves, S., Morrell, H., and Hashemzadeh, M., 2009, “Left ventricular hypertrophy is associated with obe- sity, male gender, and symptoms in healthy adolescents,” Obesity,17(3), pp. 606–610

2009
[26]

Discrepant measurements of maximal left ventricular wall thicknessbetweencardiacmagneticresonanceimagingandechocardiographyin patients with hypertrophic cardiomyopathy,

Hindieh, W., Weissler-Snir, A., Hammer, H., Adler, A., Rakowski, H., and Chan, R. H., 2017, “Discrepant measurements of maximal left ventricular wall thicknessbetweencardiacmagneticresonanceimagingandechocardiographyin patients with hypertrophic cardiomyopathy,” Circulation: Cardiovascular Imag- ing,10(8), p. e006309

2017
[27]

Journal of Medical Devices PREPRINT/ 11

The MathWorks, Inc., 2023,MATLAB Image Processing Toolbox, Release R2023a, Natick, MA. Journal of Medical Devices PREPRINT/ 11

2023