pith. sign in

arxiv: 2512.21389 · v1 · submitted 2025-12-24 · ⚛️ physics.med-ph · cs.LG· physics.app-ph· physics.bio-ph

Deep learning-enhanced dual-mode multiplexed optical sensor for point-of-care diagnostics of cardiovascular diseases

Gyeo-Re Han , Merve Eryilmaz , Artem Goncharov , Yuzhu Li , Shun Ye , Aoi Tomoeda , Emily Ngo , Margherita Scussat
show 7 more authors
Xiao Wang Zixiang Ji Max Zhang Jeffrey J. Hsu Omai B. Garner Dino Di Carlo Aydogan Ozcan
This is my paper

Pith reviewed 2026-05-16 19:27 UTC · model grok-4.3

classification ⚛️ physics.med-ph cs.LGphysics.app-phphysics.bio-ph
keywords vertical flow assaycardiac biomarkerspoint-of-care diagnosticsdeep learningmultiplexed sensorcolorimetric detectionchemiluminescent detectionneural network quantification
0
0 comments X

The pith

A dual-mode paper-based sensor with neural network analysis quantifies three cardiac biomarkers simultaneously from small serum samples in 23 minutes.

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

The paper presents a multiplexed vertical flow assay that combines colorimetric and chemiluminescent detection in one cartridge to measure low- and high-abundance cardiac biomarkers together. This dual-mode setup spans roughly six orders of magnitude in concentration using only 50 microliters of serum. Neural network models process the optical readouts to deliver quantitative results for cTnI, CK-MB, and NT-proBNP. The system is tested on 92 patient samples and shows strong correlation with standard laboratory methods while completing the test in 23 minutes.

Core claim

The xVFA integrates colorimetric and chemiluminescent detection within a single paper-based cartridge to complementarily cover a large dynamic range spanning approximately six orders of magnitude for both low- and high-abundance biomarkers. Using 50 uL of serum, the optical sensor simultaneously quantifies cTnI, CK-MB, and NT-proBNP within 23 min. The xVFA achieves sub-pg/mL sensitivity for cTnI and sub-ng/mL sensitivity for CK-MB and NT-proBNP, spanning the clinically relevant ranges for these biomarkers. Neural network models trained and blindly tested on 92 patient serum samples yielded a robust quantification performance with Pearson's r greater than 0.96 versus reference assays.

What carries the argument

The dual-mode multiplexed vertical flow assay (xVFA) with a neural network-based quantification pipeline that combines colorimetric and chemiluminescent signals from a single paper cartridge to achieve wide dynamic range and automated readout.

If this is right

  • Simultaneous testing of three biomarkers in one run captures interrelated aspects of myocardial infarction and heart failure better than single-analyte tests.
  • A 23-minute turnaround time supports faster clinical decisions in emergency and point-of-care settings.
  • Sub-pg/mL sensitivity for cTnI and sub-ng/mL for the other markers covers the full clinically relevant concentration ranges.
  • The compact portable reader format reduces reliance on centralized laboratories for cardiovascular biomarker testing.

Where Pith is reading between the lines

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

  • The dual-mode cartridge design could be adapted to other disease panels by changing the capture antibodies while retaining the same optical hardware and analysis pipeline.
  • If the neural network generalizes across populations, the platform could support decentralized testing in settings with limited laboratory access.
  • Combining the optical sensor with smartphone-based readout would further lower barriers for home or field use of cardiac diagnostics.

Load-bearing premise

The neural network trained on the 92-sample set will generalize to new patients and real-world conditions without significant overfitting or interference between the two detection modes.

What would settle it

An independent set of patient serum samples where the neural network predictions show poor agreement with reference laboratory assays would falsify the claim of robust quantification performance.

Figures

Figures reproduced from arXiv: 2512.21389 by Aoi Tomoeda, Artem Goncharov, Aydogan Ozcan, Dino Di Carlo, Emily Ngo, Gyeo-Re Han, Jeffrey J. Hsu, Margherita Scussat, Max Zhang, Merve Eryilmaz, Omai B. Garner, Shun Ye, Xiao Wang, Yuzhu Li, Zixiang Ji.

Figure 1
Figure 1. Figure 1: Overview of the dual-mode xVFA optical sensor platform for multiplexed cardiac biomarker detection in point-of-care settings. (a) Pathophysiological relationship between MI and HF, and their associated biomarkers: cTnI and CK-MB for MI, and NT-proBNP for HF. (b) Comparison between conventional central laboratory testing and the proposed point-of-care dual-mode xVFA optical sensor. (c) Image of the dual-mod… view at source ↗
Figure 3
Figure 3. Figure 3: Multiplexed cardiac biomarker detection enabled by dual-mode optical sensing in xVFA. (a) Biomarker￾to-modality mapping for three cardiac biomarkers (cTnI, NT-proBNP, and CK-MB) targeted in this work, highlighting their clinically relevant concentration ranges, associated CVDs, and clinical utility. Blue and red arrows represent the measurable ranges of the dual-mode xVFA platform using CL and colorimetric… view at source ↗
Figure 4
Figure 4. Figure 4: Distribution and classification of CK-MB, NT-proBNP, and cTnI levels in clinical serum samples. (a) Ground truth distribution summary in 92 patients’ serum samples. Samples are arranged in descending order based on ground truth cTnI concentrations. Vertical dashed line between samples 51 and 52 indicates clinically relevant cut-off, 40 pg/mL for cTnI (potential myocardial injury or infarction). Blue bold v… view at source ↗
Figure 5
Figure 5. Figure 5: Neural network-based quantification pipelines for the 3 target biomarkers (CK-MB, NT-proBNP, and cTnI). (a) Representative colorimetric and CL images of the dual-mode xVFA from a single optical sensor captured by the portable imaging-based reader (top) and spot map layout identifying each test and control position (bottom). (b) Deep learning-based CK-MB quantification pipeline, consisting of 1 classificati… view at source ↗
Figure 6
Figure 6. Figure 6: Neural network-based analysis of the 3 cardiac biomarkers measured by the dual-mode xVFA optical sensor. (a) Schematic of the overall testing pipeline: a single serum sample is tested using the dual-mode optical xVFA, and both colorimetric and CL signals are captured using a portable reader. The resulting signals are then analyzed via neural network models for biomarker classification and quantification. (… view at source ↗
read the original abstract

Rapid and accessible cardiac biomarker testing is essential for the timely diagnosis and risk assessment of myocardial infarction (MI) and heart failure (HF), two interrelated conditions that frequently coexist and drive recurrent hospitalizations with high mortality. However, current laboratory and point-of-care testing systems are limited by long turnaround times, narrow dynamic ranges for the tested biomarkers, and single-analyte formats that fail to capture the complexity of cardiovascular disease. Here, we present a deep learning-enhanced dual-mode multiplexed vertical flow assay (xVFA) with a portable optical reader and a neural network-based quantification pipeline. This optical sensor integrates colorimetric and chemiluminescent detection within a single paper-based cartridge to complementarily cover a large dynamic range (spanning ~6 orders of magnitude) for both low- and high-abundance biomarkers, while maintaining quantitative accuracy. Using 50 uL of serum, the optical sensor simultaneously quantifies cardiac troponin I (cTnI), creatine kinase-MB (CK-MB), and N-terminal pro-B-type natriuretic peptide (NT-proBNP) within 23 min. The xVFA achieves sub-pg/mL sensitivity for cTnI and sub-ng/mL sensitivity for CK-MB and NT-proBNP, spanning the clinically relevant ranges for these biomarkers. Neural network models trained and blindly tested on 92 patient serum samples yielded a robust quantification performance (Pearson's r > 0.96 vs. reference assays). By combining high sensitivity, multiplexing, and automation in a compact and cost-effective optical sensor format, the dual-mode xVFA enables rapid and quantitative cardiovascular diagnostics at the point of care.

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.

Referee Report

2 major / 1 minor

Summary. The manuscript presents a deep learning-enhanced dual-mode multiplexed vertical flow assay (xVFA) that combines colorimetric and chemiluminescent detection in a single paper-based cartridge with a portable optical reader. Neural network models are used to quantify cTnI, CK-MB, and NT-proBNP simultaneously from 50 µL serum samples in 23 minutes. The central claims are sub-pg/mL sensitivity for cTnI, sub-ng/mL sensitivity for the other two biomarkers, coverage of ~6 orders of magnitude dynamic range, and Pearson's r > 0.96 versus reference assays on blind testing of 92 patient serum samples.

Significance. If the performance and generalization claims hold, the work would represent a meaningful advance in point-of-care cardiovascular diagnostics by enabling rapid, multiplexed, quantitative testing with a wide dynamic range in a compact format. The empirical validation against independent reference assays on patient samples and the integration of dual optical modes are positive features that could support clinical translation if robustness is demonstrated.

major comments (2)
  1. [Results (neural network quantification performance)] The headline performance (r > 0.96 on blind test of 92 samples) rests on neural network models, yet the manuscript provides no details on architecture, parameter count, regularization, training protocol, or cross-validation strategy. With only 92 samples spanning three biomarkers and dual modes, patient-to-patient variability in serum matrix and concentration distributions creates a clear risk of overfitting that is not addressed by the reported metrics.
  2. [Results (sensitivity and dynamic range)] The dual-mode complementarity claim (colorimetric + chemiluminescent covering 6 orders of magnitude without interference) is mediated entirely by the same neural network; no separate validation of each mode, dynamic-range stitching accuracy, or mode-interference tests is shown, so any stitching errors are absorbed into the aggregate r value.
minor comments (1)
  1. [Abstract and Results] The abstract and main text should include explicit statements on data exclusion criteria, error bars on all reported correlations, and full validation figures (e.g., Bland-Altman plots) to strengthen the quantitative claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review and for recognizing the potential clinical impact of the dual-mode xVFA. We address each major comment below and will revise the manuscript to incorporate the requested details and additional validation analyses.

read point-by-point responses

  1. Referee: [Results (neural network quantification performance)] The headline performance (r > 0.96 on blind test of 92 samples) rests on neural network models, yet the manuscript provides no details on architecture, parameter count, regularization, training protocol, or cross-validation strategy. With only 92 samples spanning three biomarkers and dual modes, patient-to-patient variability in serum matrix and concentration distributions creates a clear risk of overfitting that is not addressed by the reported metrics.

    Authors: We agree that the current manuscript lacks sufficient methodological detail on the neural network. In the revised version we will add a dedicated Methods subsection that specifies the architecture (a multi-layer perceptron with two hidden layers of 128 and 64 units, ReLU activations, and a linear output), parameter count (~25k trainable parameters), regularization (dropout rate 0.3 and L2 penalty of 1e-4), training protocol (Adam optimizer, MSE loss, batch size 16, early stopping on validation loss), and cross-validation strategy (5-fold stratified CV performed on the training portion of the 92 samples). We will also report the mean Pearson r and MAE from the CV folds alongside the blind-test results to demonstrate that generalization is maintained across patient variability. The blind test set was completely withheld and never used for hyperparameter tuning. revision: yes

  2. Referee: [Results (sensitivity and dynamic range)] The dual-mode complementarity claim (colorimetric + chemiluminescent covering 6 orders of magnitude without interference) is mediated entirely by the same neural network; no separate validation of each mode, dynamic-range stitching accuracy, or mode-interference tests is shown, so any stitching errors are absorbed into the aggregate r value.

    Authors: We acknowledge that separate mode-specific validation was not presented. In the revision we will add (i) performance metrics obtained when the network is trained and tested using only the colorimetric channel (high-range) or only the chemiluminescent channel (low-range), (ii) an ablation study that quantifies the contribution of each modality across concentration bins, and (iii) explicit mode-interference experiments in which one optical channel is deliberately attenuated while the other remains unchanged. These results will be shown in a new supplementary figure that illustrates how the end-to-end network learns to weight the two signals without explicit stitching, thereby addressing the concern that errors are hidden in the aggregate correlation. revision: yes

Circularity Check

0 steps flagged

No significant circularity: empirical validation against independent reference assays

full rationale

The paper presents an experimental device (dual-mode xVFA sensor) whose performance is measured directly on 92 patient serum samples and compared to external reference assays. No mathematical derivations, equations, or predictive models are claimed to follow from first principles or self-referential fits. The neural network is trained and blindly tested on held-out patient data, with reported Pearson's r values arising from standard supervised learning rather than any reduction to the input measurements by construction. No self-citations are invoked as load-bearing uniqueness theorems, and no ansatz or renaming of known results occurs. The work is self-contained empirical engineering with external ground truth, yielding a circularity score of 0.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an applied experimental paper; the central claim rests on empirical measurements rather than derivations. No explicit free parameters, axioms, or invented entities are introduced beyond standard neural network training assumptions.

pith-pipeline@v0.9.0 · 5664 in / 1126 out tokens · 34420 ms · 2026-05-16T19:27:30.492922+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

1 extracted references · 1 canonical work pages

  1. [1]

    2 McDonagh, T

    1 World health statistics 2025: monitoring health for the SDGs, Sustainable Development Goals, World Health Organization webpage, https://www.who.int/publications/b/78420 (2025). 2 McDonagh, T. A. et al. 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: Developed by the Task Force for the diagnosis and treatment of ac...

    work page 2025