List of Papers Browse by Subject Areas Author List
Abstract
Our work introduces a novel framework for generating a digital twin of the lung by leveraging diagnostic CT data and motion planning based on electrocardiogram (ECG) signals from wearable sensors. The simulation and real-time visualisation (3D+t) of lung motion is driven by automated predictions of respiratory signals. This dynamic framework provides an interface to visualise a patient’s live breathing, enabling reliable, real-time monitoring of cardiorespiratory function. We evaluate our framework using two public datasets, BIDMC and CapnoBase, comprising 95 subjects with synchronised PPG, ECG, and reference respiratory signals. Central to our approach is a novel adaptive ECG-Derived Respiration (EDR) method, which serves as the input vector for kinetic simulation of the lung’s digital twin. On the BIDMC dataset, our method achieves a Mean Adaptive Error (MAE) of 2.84 bpm with 98% accuracy within ±5 bpm. On the CapnoBase dataset (42 subjects), our method substantially reduces error to 3.54 bpm and improves accuracy to 79.3% within ±5 bpm. The results show significant improvements over spectral and decomposition-based baselines, including Welch Power Spectral Density (PSD) and Ensemble Empirical Mode Decomposition with Spectral Density Fusion (EEMD-SDF). These results demonstrate that ECG-based respiratory extraction provides robust and accurate motion signals that can drive real-time lung digital twin simulations. The framework supports reliable, wearable-driven respiratory monitoring and the simulation of cardiopulmonary behaviour. Our code is publicly available at https://github.com/Nikhil-Rao20/Digital_Twin.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/5574_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026/supp/5574_supp.zip
Link to the Code Repository
https://github.com/Nikhil-Rao20/Digital_Twin
Link to the Dataset(s)
BIDMC Dataset: https://physionet.org/content/bidmc/1.0.0/
CapnoBase Dataset: https://ppg-beats.readthedocs.io/en/latest/datasets/capnobase/
BibTex
@InProceedings{AchPar_Digital_MICCAI2026,
author = { Acharya, Partha AND Sulake, Nikhileswara Rao AND Chakraborty, Soutrik AND Mandal, Subhamoy AND Chakraborty, Suman},
title = { { Digital Twin of the Lung from Wearable Biosignals for Real-Time Respiratory Monitoring } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16889},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The paper proposes a novel framework that integrates wearable ECG-derived respiration (EDR) with a CT-based lung model to construct a real-time digital twin of the lung.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.Combines wearable biosignal processing with 3D digital twin modeling, which is relatively unexplored for lung applications. 2.The three-stage framework is well-structured and easy to follow 3.Uses two public datasets (BIDMC and CapnoBase) with diverse patients.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.Limited methodological novelty in signal processing. 2.The 3D lung simulation is visually demonstrated but lacks quantitative validation against ground-truth lung motion or clinical imaging. 3.The optimized filter (0.1–0.4 Hz) excludes higher respiratory rates (>24 bpm), requiring fallback strategies 4.Limited comparison with recent deep learning-based respiratory estimation methods.
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The methodological novelty is moderate, and the digital twin component has not been rigorously validated. Although the results are encouraging, further experiments and a more convincing explanation of the novelty would strengthen the paper. Overall, it is slightly above the acceptance threshold.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
This work offers an insightful interdisciplinary contribution with potential clinical significance, especially in wearable-driven healthcare monitoring, though additional validation is necessary.
Review #2
- Please describe the contribution of the paper
The paper introduces a framework for generating a real-time lung digital twin from EDR for dynamic 3D lung motion visualization. The pipeline has three parts: extraction of a continuous respiratory waveform from a wearable ECG, estimation of respiratory rate from that waveform, and conversion of the signal into lung-mesh motion.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
- The idea of using a single wearable ECG lead to drive a live breathing visualization is intuitive and potentially useful for clinical monitoring.
- The three-stage pipeline is clearly presented in the overview figure and method section, making the paper easy to follow.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
- The paper claims a lung digital twin, but what is actually built is close to a signal-driven visualization method.
- The evaluation mainly tests respiratory-rate estimation, which validates signal feature extraction only.
- The paper repeatedly emphasizes real-time monitoring, but there is no evidence for it.
- The visualization itself is not related to a clinically meaningful downstream use case.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The overall idea is interesting, though I do not think the experimental evidence fully supports their claims. The authors claimed the method as a digital twin, but what is actually demonstrated is closer to a visualization framework. The experiments do not validate their digital twin claim. The paper also repeatedly emphasizes real-time monitoring, but does not provide actual evidence for it.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Reject
- [Post rebuttal] Please justify your final decision from above.
The rebuttal clarifies that the framework is better viewed as a proof-of-concept and that the real-time claim is understandable. It also explains why direct quantitative mesh validation is difficult with currently available public datasets. I find this explanation reasonable, but it also means that the submitted evidence mainly supports a proof-of-concept visualization framework rather than a validated patient-specific lung digital twin. Therefore, I keep my score unchanged.
Review #3
- Please describe the contribution of the paper
This paper proposes a framework for constructing a real-time lung digital twin driven by wearable ECG signals. The method combines an ECG-derived respiration (EDR) pipeline with a 3D CT-based lung mesh to generate dynamic breathing visualizations. The technical contribution lies in an adaptive EDR method using a biquad bandpass filter, which improves respiratory rate estimation over standard PSD, EEMD-based, and Butterworth baselines on the BIDMC and CapnoBase datasets. The work aims to bridge wearable biosignal processing and organ-level simulation, targeting continuous and non-invasive respiratory monitoring.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.Novel cross-domain integration: Combines wearable monitoring and digital twin modeling, which is interesting in conception.
2.End-to-end Real-time framework: Integrates signal extraction and 3D visualization into a coherent pipeline.
3.Empirical evaluation: Demonstrates consistent improvements over multiple baselines across two public datasets.
4.Well Written and Clear Illustration: The paper well described the EDR and pipeline with clear figure illustration. The experiments provide clear analysis and support for the conclusion.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.Limited methodological novelty: The EDR approach mainly relies on incremental improvements (filter design and parameter tuning) over established pipelines. The main gain may come from parameter optimisation rather than algorithmic innovation.
2.Insufficient validation of the digital twin: The 3D lung simulation is only qualitatively demonstrated, with no quantitative evaluation of anatomical or physiological accuracy
3.Concerns on overstated claims and clarity issues: I have concerns that claims such as “first framework” and “digital twin” are overstated in the manuscript, as the system is closer to a visualization-driven simulation rather than a true patient-specific predictive digital twin.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper addresses a relevant and emerging topic (wearable-driven digital twins) and presents a coherent end-to-end system with promising empirical results in respiratory signal extraction. However, The methodological novelty is limited, and the central ``digital twin’’ conception lacks rigorous validation and physiological fidelity. As a result, the work appears more as a signal processing enhancement coupled with a proof-of-concept visualization, rather than a fully developed medical image computing contribution.
- Reviewer confidence
Not confident (1)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Author Feedback
We sincerely thank the Area Chair and all reviewers for their careful reading of our manuscript and for the constructive feedback. This work integrates non-invasive data to generate a patient specific lung digital twin. By combining deep learning-based segmentation from static CT scans with dynamic, ECG derived respiratory rates, our workflow reconstructs a 4D rendering of individual anatomy and pathology. Unlike standard techniques requiring cine imaging, our algorithm achieves this 4D visualization demonstrated here via a lung cancer case study by merging two multi-dimensional signals from the same subject. The Meta Reviewer, Reviewer2 and Reviewer3 characterized our model as a signal-driven visualization, but our work aligns with the Digital Twin Consortium’s 2024 definition of an integrated, data-driven virtual representation. Much like foundational cardiac digital twins that use patient-specific simulation without full bidirectionality, our framework qualifies by merging CT-derived anatomy and pathology with ECG-based respiratory dynamics. This continuous mirroring of the patient’s actual disease state creates a personalized virtual counterpart. However, we agree that framing this work as a proof-of-concept framework rather than a fully verified clinical digital twin is more accurate and equitable. We acknowledge the omission of latency figures and clarify that our framework operates in two phases. The initial phase CT segmentation and 3D mesh construction is computationally intensive but occurs only periodically, using a pre-trained model ensures rapid inference. The second phase, continuous respiratory rate extraction, is designed for live integration. Validated on the BIDMC and CapnoBase datasets (representing ICU and anesthesia environments), our pipeline demonstrates a real-time latency under 100 ms. This confirms the framework’s generalizability and readiness for processing live wearable data in diverse clinical settings. Direct quantitative mesh validation is currently impossible due to the lack of synchronized ECG and 4D CT public datasets. To address this, we anchored the mesh motion in established respiratory biomechanics and physiological literature. Additionally, a senior pulmonary specialist validated the framework’s animation and segmentation, confirming its clinical face validity and significance. While this does not replace quantitative imaging metrics, it provides expert-level verification of the model’s anatomical and functional accuracy, which we will clarify in the revision. The lack of tachypnea cases is a dataset limitation, not a methodological one. Our integrated fallback system addresses this by using Butterworth EDR for rates outside the biquad’s ideal range, covering 52 of 53 BIDMC participants with negligible accuracy loss. Unlike deep learning models that output 30-second scalar rates, our end-to-end solution generates the continuous waveforms necessary for real-time 3D mesh animation. We validated our filter design against Welch PSD, EEMD-SDF fusion, and Butterworth EDR. Finally, due to page constraints and double-blind requirements, we have limited the discussion of our previously reported deep learning segmentation methods. This work’s novelty lies in unifying disparate bio signals and images into a lung digital twin. While EDR principles are known, we contribute a complete workflow combining adaptive EDR, wearable-compatible extraction, and CT-based motion simulation. We also show that a narrower respiratory band improves performance a key finding for biomedical imaging. This framework facilitates industry-academic collaboration and aims to integrate with consumer smartwatches for home-based monitoring of COPD or lung cancer risks. By enabling real-time 3D depictions for remote oversight, this study establishes a lightweight, resource-efficient healthcare solution for limited-resource settings. We will release specific code snippets under the MIT License, full code remains closed-source.
Meta-Review
Meta-review #1
- Your recommendation
Invite for Rebuttal
- Please justify your decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
The draft offers a novel three-part pipeline to generate the lung digital twin from ECG-derived respiration.
There is a lot of reasons why having the lung twin is useful (R2/AC), has clinical relevance (AC) and remains challenging at present (AC), so the significance of the research appears bleeding obvious. Moreover, reviewers found the three stages end-to-end pipeline to be clear and well structured (R1/R2/R3). Results are consistent over two datasets covering diverse patients (R1/R3), suggesting good generalizability.
I would have bet that the main contribution was the lung digital twin per se, but if I understood the authors correctly, they seem to be putting the emphasis on the process to get to it. In particular, in the pipeline, to build it. This perception is important I suppose. For instance, R1 and R3 coincide that there is limited methodological novelty.
Besides, quantitative evaluation is missing (R1,R3/AC) and further it focuses solely on respiratory rates (R2). Since the pipeline is unidirectional, there is no evidence of supporting the claim of “real-time monitoring” (R2). In fact, R3 also suggests that there is a bit of overselling. And with limited comparison with SOTA e.g., deep models, assessment of performance is incomplete (R1). Finally, R1 points out something relevant; the optimised filter excludes higher respiratory rates, hence requiring fallback strategies.
Overall, I got the impression that the reviewers and myself do agree on the potential of the work, but remain cautious about the operationalization.
Reproducibility is compromised without access to source code.
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Reject
- Please justify your recommendation.
Perhaps the most difficult to decide in my package; whilst almost everything in my guts tell me that this could/should be accepted as there are a lot of good things in this submission, but I can’t ignore that the authors’ response did not solve the concerns of R2.Further, R2 has lower the score after the rebuttal clearly indicative that their concerns have not been properly addressed. Also, while I may have some discrepancy on this harsh assessment by R2, I still can see why such frustration; the concern about quantitative mesh validation was ducked rather than tackled, and some of the claims needed waterish down e.g. “real time”.
So considering that neither R1 or R3 move their scores upwards, I’m voting rejection (against my guts). Notwithstanding, this being a truly difficult decision, I look forward to see this work published elsewhere soon.
Meta-review #2
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
During the initial review, as also summarized by the meta-review, the key concerns were with method novelty, particularly in the process of developing the twin, and quantitative validation with experiments focusing on respiratory rates (excluding higher rates) and lack of support for real-time claim. The authors clarified the work as proof-of-concept, which more closely matches the meta-review and reviewers’ perception, and acknowledge the noted limitations. The rebuttal provided justifications for methodological choices and a fallback strategy for handling a wider range of respiratory rates. The lack of quantitative validation of the mesh was attributed to the lack of synchronized ECG and 4D CT public dataset. While the post-rebuttal discussion did not arrive at a clear decision, the consensus is leaning towards acceptance. AC agrees that the presented integration of the ECG and CT for personalized lung representation shows promise for (remote) health monitoring via wearables. AC also notes that authors promised on releasing code for key modules to aid reproducibility and transparency. Thus, AC recommends acceptance, contingent on ensuring that the final version must emphasize the proof-of-concept nature, clearly state validation limitations, and detail the real-time performance and fallback strategy.
Meta-review #3
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
The authors introduce a novel framework for generating a digital twin of the lung by leveraging diagnostic CT data and motion planning based on ECG signals from wearable sensors. The overall pipeline looks more like a standard engineering pipeline that integrates wearable signals acquisition, processing, and rendering, but the results look promising under a runtime close to real time (latency under 100 ms) and application scenario has its clinical significance in home-based monitoring or ICU monitoring. Regarding the lack of downstream evaluation, the authors mentioned that a radiologist verification of the anatomical and functional accuracy will be added into the final version. As a proof-of-concept paper, it can bring the MICCAI community more discussions and insights into lung digital twin from wearable device signals.
