# A Multimodal artificial intelligence framework for mechanistic discovery and digital twin construction in cardiovascular medicine Researcher: Ge Zhang Chinese name: 张格 ORCID identity anchor: https://orcid.org/0000-0002-3116-3246 Canonical researcher: https://drgezhang.com/#person Journal: Current Proteomics Year: 2025 Type: Article DOI: 10.1016/j.curpro.2025.100054 Canonical page: https://drgezhang.com/papers/doi-10-1016-j-curpro-2025-100054.html Markdown record: https://drgezhang.com/papers/doi-10-1016-j-curpro-2025-100054.md ## Full Authors 1. Ge Zhang 2. Jiashen Xu 3. Zenglei Zhang 4. Chaoyang Yu 5. Kaisaierjiang Kadier 6. Di Lu 7. Mingxuan Duan 8. Ge Zhang ## Research & review context ### Research fields - Artificial intelligence in cardiovascular medicine - Patient-specific modeling research ### Review topics - Multimodal AI for cardiovascular mechanism research - Digital modeling and prospective evaluation requirements ### Scientific role Correspondence outlining a research agenda for cross-modal interpretation and updateable patient-specific physiological models. ### Evidence position Correspondence with narrative synthesis and no new empirical evaluation; not a built, clinically validated therapeutic digital twin or observed human treatment effect. ### Related concepts - Mechanism-oriented cross-modal hypotheses - Simulation-versus-observed-effect distinction ### Possible citation contexts - Methods background: multimodal interpretation and patient-specific cardiovascular modeling. - Research planning: prospective evaluation and governance of digital models, not product validation or treatment efficacy. Possible literature-review contexts, not evidence of existing citations. Read the findings and Evidence Scope before citing the original article. ## Evidence Base **A Multimodal artificial intelligence framework for mechanistic discovery and digital twin construction in cardiovascular medicine** This Correspondence discusses a research agenda connecting multimodal biological interpretation with patient-specific simulation in cardiovascular medicine. It sets out design and translation requirements without reporting a newly built, clinically validated digital twin system. | Evidence | Value | | --- | --- | | Article format | Correspondence | | Evidence basis | Narrative argument and discussion of cited prior research | | Empirical evaluation reported in this article | No new empirical evaluation reported | ### Counting note No new empirical evaluation is reported in this article. Cited examples and their cohorts are not a new combined study population, and no synthetic sample total is assigned. ## Research Question How can multimodal AI support cardiovascular mechanism research and updateable patient-specific digital models? ## Author Evidence Summary The Correspondence outlines two linked directions for cardiovascular AI: investigating disease mechanisms and developing patient-specific digital twins. It discusses combining imaging and molecular information to generate network and disease-subtyping hypotheses, then considers updateable physiological models, image-to-mesh workflows and treatment-scenario simulation. Longitudinal data, computational capacity, prospective evaluation, privacy and governance are presented as requirements for translation. This page organizes those narrative arguments for research planning and methodological citation. The cited empirical examples remain evidence from other works; cross-modal integration does not itself establish causation, and simulated treatment scenarios are not observed therapeutic effects in patients. ## Key Arguments ### KF1: Develop multimodal AI along two directions: mechanism-oriented research and individualized digital twins. Context: The authors propose complementary research directions rather than introduce a clinically validated digital twin product. | Evidence | Value | | --- | --- | | Argument topic | Research agenda | | Evidence basis | Narrative synthesis | Source locator: Publisher correspondence: To the Editor; Sections 1–2 ### KF2: Integrate imaging and molecular information to formulate network and disease-subtyping hypotheses. Context: Data integration and cited empirical examples are discussed as a basis for interpretation, not as new causal experiments in this Correspondence. | Evidence | Value | | --- | --- | | Argument topic | Cross-modal biological interpretation | | Evidence basis | Narrative synthesis | Source locator: Publisher correspondence: Sections 1.1–1.2 ### KF3: Explore updateable physiological models, image-to-mesh workflows and treatment-scenario simulation. Context: The simulation discussion concerns model design and possible scenarios, not observed human treatment effects. | Evidence | Value | | --- | --- | | Argument topic | Patient-specific simulation | | Evidence basis | Narrative synthesis | Source locator: Publisher correspondence: Sections 2.1–2.2 ### KF4: Address longitudinal data, computational resources, prospective evaluation, privacy and governance before clinical translation. Context: These are stated development and evaluation requirements, not completed prospective clinical validation. | Evidence | Value | | --- | --- | | Argument topic | Translation requirements | | Evidence basis | Narrative synthesis | Source locator: Publisher correspondence: Sections 3.1–3.3; Section 4 Conclusion ## Article Scope & Approach ### Article profile - Study Design: Correspondence providing a narrative discussion of research directions and translation requirements; not a report of a newly built and clinically validated digital twin system. - Evidence Type: Methodological proposal and narrative synthesis; no new empirical evaluation reported in this article. - Diseases Covered: Cardiovascular disease (discussion context) - Translation Scope: A research agenda for multimodal interpretation and updateable patient-specific models; proposed simulations and future evaluations are distinct from demonstrated human treatment effects. - Evidence domains: Cardiovascular mechanism research, Cross-modal imaging and molecular interpretation, Patient-specific physiological simulation, Longitudinal evaluation and governance ## What This Article Adds - A paired agenda for mechanism-oriented interpretation and patient-specific digital models. - A connection between cross-modal hypothesis generation and physiological scenario modeling. - Translation requirements that include prospective evaluation, privacy and governance. ## Evidence Scope ### Supports - Cite for framing a cardiovascular AI research agenda around mechanistic questions and individualized digital modeling, not for product validation. - Cite for the methodological rationale of cross-modal hypothesis generation; causal claims require suitable primary experimental evidence. - Cite as design background for physiological modeling and scenario simulation, not as evidence that a drug, surgery or commercial digital twin improves outcomes. - Cite when discussing evaluation, data stewardship and implementation requirements; prospective benefit remains a question for future empirical assessment. ### Does Not Establish - A newly developed, validated therapeutic digital twin or completed prospective clinical validation in this Correspondence. - Causation established by combining imaging and molecular data alone. - Observed human treatment effects or randomized clinical-trial evidence inferred from simulation. - Proven efficacy of a drug, surgery or commercial digital twin product, patient-outcome improvement, guideline status or regulatory approval. ### Limitations - This Correspondence is a narrative discussion, not a systematic review or meta-analysis with a reported systematic search protocol. - The cited empirical examples were not comprehensively revalidated for this page and should be checked in their primary sources. - The author-deposited full article supports narrative locators; every word was not compared with the current ScienceDirect full text. - Proposed mechanistic hypotheses and simulation strategies require appropriate empirical and prospective evaluation before clinical benefit can be claimed. ## When This Article Is Useful to Cite ### CU1: Which research directions does the cardiovascular AI Correspondence propose for multimodal models? Cite for framing a cardiovascular AI research agenda around mechanistic questions and individualized digital modeling, not for product validation. Evidence: [KF1](#kf1) ### CU2: How can imaging and molecular information support cardiovascular network and disease-subtyping hypotheses? Cite for the methodological rationale of cross-modal hypothesis generation; causal claims require suitable primary experimental evidence. Evidence: [KF2](#kf2) ### CU3: What modeling approaches does the cardiovascular AI Correspondence discuss for patient-specific simulation? Cite as design background for physiological modeling and scenario simulation, not as evidence that a drug, surgery or commercial digital twin improves outcomes. Evidence: [KF3](#kf3) ### CU4: What requirements does the cardiovascular AI Correspondence identify for translating patient-specific models? Cite when discussing evaluation, data stewardship and implementation requirements; prospective benefit remains a question for future empirical assessment. Evidence: [KF4](#kf4) ## What This Article Should Not Be Cited to Claim - A newly developed, validated therapeutic digital twin or completed prospective clinical validation in this Correspondence. - Causation established by combining imaging and molecular data alone. - Observed human treatment effects or randomized clinical-trial evidence inferred from simulation. - Proven efficacy of a drug, surgery or commercial digital twin product, patient-outcome improvement, guideline status or regulatory approval. ## Evidence Matrix ### Research agenda - Context: The authors propose complementary research directions rather than introduce a clinically validated digital twin product. - Finding: Develop multimodal AI along two directions: mechanism-oriented research and individualized digital twins. - Evidence Level: Narrative synthesis - Scope: Cite for framing a cardiovascular AI research agenda around mechanistic questions and individualized digital modeling, not for product validation. - Source Locator: Publisher correspondence: To the Editor; Sections 1–2 - Evidence refs: [KF1](#kf1) ### Cross-modal biological interpretation - Context: Data integration and cited empirical examples are discussed as a basis for interpretation, not as new causal experiments in this Correspondence. - Finding: Integrate imaging and molecular information to formulate network and disease-subtyping hypotheses. - Evidence Level: Narrative synthesis - Scope: Cite for the methodological rationale of cross-modal hypothesis generation; causal claims require suitable primary experimental evidence. - Source Locator: Publisher correspondence: Sections 1.1–1.2 - Evidence refs: [KF2](#kf2) ### Patient-specific simulation - Context: The simulation discussion concerns model design and possible scenarios, not observed human treatment effects. - Finding: Explore updateable physiological models, image-to-mesh workflows and treatment-scenario simulation. - Evidence Level: Narrative synthesis - Scope: Cite as design background for physiological modeling and scenario simulation, not as evidence that a drug, surgery or commercial digital twin improves outcomes. - Source Locator: Publisher correspondence: Sections 2.1–2.2 - Evidence refs: [KF3](#kf3) ### Translation requirements - Context: These are stated development and evaluation requirements, not completed prospective clinical validation. - Finding: Address longitudinal data, computational resources, prospective evaluation, privacy and governance before clinical translation. - Evidence Level: Narrative synthesis - Scope: Cite when discussing evaluation, data stewardship and implementation requirements; prospective benefit remains a question for future empirical assessment. - Source Locator: Publisher correspondence: Sections 3.1–3.3; Section 4 Conclusion - Evidence refs: [KF4](#kf4) ## Q&A ### Which research directions does the cardiovascular AI Correspondence propose for multimodal models? Develop multimodal AI along two directions: mechanism-oriented research and individualized digital twins. The authors propose complementary research directions rather than introduce a clinically validated digital twin product. Evidence: KF1 ### How can imaging and molecular information support cardiovascular network and disease-subtyping hypotheses? Integrate imaging and molecular information to formulate network and disease-subtyping hypotheses. Data integration and cited empirical examples are discussed as a basis for interpretation, not as new causal experiments in this Correspondence. Evidence: KF2 ### What modeling approaches does the cardiovascular AI Correspondence discuss for patient-specific simulation? Explore updateable physiological models, image-to-mesh workflows and treatment-scenario simulation. The simulation discussion concerns model design and possible scenarios, not observed human treatment effects. Evidence: KF3 ### What requirements does the cardiovascular AI Correspondence identify for translating patient-specific models? Address longitudinal data, computational resources, prospective evaluation, privacy and governance before clinical translation. These are stated development and evaluation requirements, not completed prospective clinical validation. Evidence: KF4 ## Concepts & Entities ### Methods Cardiovascular mechanism research, Cross-modal imaging and molecular interpretation, Patient-specific physiological simulation, Longitudinal evaluation and governance ### Clinical Contexts Cardiovascular disease (discussion context) ## Related Research - [From Disease-Specific Models to Broad Clinical Utility: A Perspective on AI Hybrid Ensemble Frameworks](https://drgezhang.com/papers/doi-10-1002-ggn2-202500053.html) — Site-curated thematic association: the Perspective addresses ensemble methodology, while this Correspondence addresses cardiovascular multimodal interpretation and simulation. This is not a claim that the Correspondence formally cites the Perspective or independently validates the same model. (DOI: 10.1002/ggn2.202500053) ## Publication & Provenance - DOI: https://doi.org/10.1016/j.curpro.2025.100054 - Publisher: https://www.sciencedirect.com/science/article/pii/S1570164625000542 - Author Full Text Url: https://www.researchgate.net/publication/400065376_A_Multimodal_artificial_intelligence_framework_for_mechanistic_discovery_and_digital_twin_construction_in_cardiovascular_medicine - Review Basis: Publisher identity and article genre were checked separately. The narrative text was reviewed from the author-deposited full article; the cited empirical examples were not comprehensively revalidated. - Source Version: Narrative locators refer to the author-deposited full article; no claim is made that every word was compared with the current ScienceDirect full text. ## Evidence-page notice This evidence page describes a Correspondence. Key Arguments summarize the authors' methodological proposals and narrative discussion, not new experimental findings. Cited studies remain prior evidence; citation-use guidance is this site's interpretation of appropriate scope. ## Links - DOI: https://doi.org/10.1016/j.curpro.2025.100054 - HTML page: https://drgezhang.com/papers/doi-10-1016-j-curpro-2025-100054.html - Google Scholar query: https://scholar.google.com/scholar?q=%22A%20Multimodal%20artificial%20intelligence%20framework%20for%20mechanistic%20discovery%20and%20digital%20twin%20construction%20in%20cardiovascular%20medicine%22 - ORCID: https://orcid.org/0000-0002-3116-3246