About

Alison Marsden is the Douglass M. and Nola Leishman Professor of Cardiovascular Diseases, Professor of Pediatrics (Cardiology) and of Bioengineering, and also holds a courtesy appointment in Mechanical Engineering at Stanford University. Her research focuses on cardiovascular diseases, integrating bioengineering principles to advance understanding and treatment of cardiovascular conditions. She is recognized for her contributions to the field of bioengineering, particularly in the context of cardiovascular health, and holds a prominent academic and research position within Stanford's bioengineering community.

Research topics

• Medicine
• Cardiology
• Machine Learning
• Surgery
• Computer Science
• Internal medicine
• Physics
• Biomedical engineering
• Mathematics
• Materials science
• Radiology
• Mathematical optimization
• Composite material
• Statistics
• Pathology
• Mechanics
• Algorithm
• Structural engineering
• Anatomy

Selected publications

• Accelerated Patient-Specific Hemodynamic Simulations with Hybrid Physics-Based Neural Surrogates

  ArXiv.org · 2026-04-02

  articleOpen accessSenior author

  Physics-based 0D reduced-order models provide computationally lightweight predictions of cardiovascular flows, resolving bulk hemodynamics in fractions of a second that would take days to solve using traditional 3D finite-element techniques. However, the accuracy of 0D models is limited as a result of the dramatic simplifications made in their derivations. In this work, we use 0D parameters learned from high-fidelity 3D data to improve 0D model accuracy without sacrificing its low computational cost or interpretability. We use the resistor-quadratic resistor-inductor (RRI) model to predict pressure drops over 0D vessels and bifurcations, where the resistances and inductance (0D parameters) are predicted from the bifurcation or vessel geometry using neural networks trained on high-fidelity 3D simulations. We validate the hybrid physics-based data-driven framework in three types of patient-specific vasculature - aortic, aortofemoral, and pulmonary anatomies. Use of learned 0D parameters reduces error by at least 50% compared to baseline 0D parameters across all anatomical cohorts. The improvements are especially marked for the more complex pulmonary anatomies, where 0D models with learned parameters reduced error from 30% to 7%. Exclusion of the quadratic resistor in the RRI model improved convergence compared to using the full RRI model. The resulting hybrid model presents a means of real-time (personal laptop runtime of <2 seconds for the most complex pulmonary anatomies), interpretable, and accurate cardiovascular flow modeling, enabling digital twins that support clinical decision-making as well as cardiovascular science and engineering research.

  PublisherOA PDF

• Cardiac Mechanics Modeling: Recent Developments and Current Challenges

  Journal of Elasticity · 2026-04-28

  preprintOpen accessSenior author
  PublisherOA PDFDOI

• FalconBC: Flow matching for Amortized inference of Latent-CONditioned physiologic Boundary Conditions

  arXiv (Cornell University) · 2026-03-18

  preprintOpen access

  Boundary condition tuning is a fundamental step in patient-specific cardiovascular modeling. Despite an increase in offline training cost, recent methods in data-driven variational inference can efficiently estimate the joint posterior distribution of boundary conditions, with amortization of training efforts over clinical targets. However, even the most modern approaches fall short in two important scenarios: open-loop models with known mean flow and assumed waveform shapes, and anatomies affected by vascular lesions where segmentation influences the reachability of pressure or flow split targets. In both cases, boundary conditions cannot be tuned in isolation. We introduce a general amortized inference framework based on probabilistic flow that treats clinical targets, inflow features, and point cloud embeddings of patient-specific anatomies as either conditioning variables or quantities to be jointly estimated. We demonstrate the approach on two patient-specific models: an aorto-iliac bifurcation with varying stenosis locations and severity, and a coronary arterial tree.

  PublisherDOI

• Deep reinforcement learning for automatic anatomic CT landmark localization in Stanford Type B aortic dissection

  Radiology Advances · 2026-01-16 · 1 citations

  articleOpen access

  Abstract Background Long-term aortic dissection monitoring requires consistent, landmark-based measurements over time. Purpose To evaluate the performance of deep reinforcement learning (DRL) agents for the detection of anatomic landmarks in patients with Stanford Type B aortic dissection (TBAD). Materials and Methods This is an international retrospective study of 396 CT angiography scans of patients with TBAD from 9 participating sites (mean age 57.6 years ± 13.7/[SD]; 236 male, 160 female). Aortic landmarks, including the aortic annulus and 8 aortic branch vessels, were manually labeled. Additionally, interobserver variability data were collected between 2 observers for 30 scans. DRL agents were trained independently for each landmark with the manual labels serving as the reference standard. Unique landmark locations were obtained from (1) single agents’ predictions and (2) clusters of landmark predictions using the DBSCAN clustering algorithm. The performance was analyzed based on distance metrics (mean, median, quantiles) and failure rates, defined as a distance error of more than 10 mm. Interobserver variability data were analyzed with a pairwise Wilcoxon test. Results On the internal test set, DRL single agents predicted landmark locations with median errors of 2.7 (95% CI, 2.2-3.3) mm and 4.8% failure rate. Cluster-based predictions resulted in a median error of 2.5 (95% CI, 2.4-2.7) mm and 4.0% failure rate. Pooled over all landmarks, cluster-based predictions outperformed single-agent predictions (P &amp;lt; 1e-5). In the external test set, cluster-based DRL models demonstrated significantly lower localization errors and fewer failures compared to single-agent DRL models (P &amp;lt; .01), and were either not significantly different (single agents) from or significantly better (cluster-based, P &amp;lt; .05) than human interobserver variability. The median processing time for a single agent’s prediction was 1.0 second (IQR, 0.7–1.4 seconds). Conclusion Single-agent and cluster-based DRL predict aortic landmarks in patients with TBAD with high accuracy and precision, comparable to the variability between human observers.

  PublisherDOI

• Simulations Predict Improved Valve Performance Without Direct Leaflet Intervention After Neonatal Truncus Arteriosus Repair.

  PubMed · 2026-01-13

  article

  Truncus arteriosus (TA) is a rare and severe congenital heart disease. Quadricuspid valve morphology occurs in 25% of all TA patients and is linked to regurgitation and increased risk of re-operation. It remains unclear how hemodynamic changes after TA repair alter valve performance. This study simulated pre- and postoperative conditions in a neonatal TA patient to investigate valve performance without direct intervention. We hypothesize that valve performance before and after truncal repair can be predicted in-silico, matching in-vivo imaging and identifying mechanisms how hemodynamic changes after repair will reduce valve regurgitation without direct intervention. Pre- and postoperative CT images of a neonatal patient with quadricuspid valve were segmented. Free edge length and geometric height from the patient's echocardiogram were used to model the valve. For the preoperative condition, ventricular pressures were set equal modeling an unrestricted ventricular septal defect. Systemic and pulmonary resistances were tuned based on the patient's Qp:Qs ratio. For the postoperative condition, boundary conditions were modified to mimic patient-specific hemodynamics after TA repair. The preoperative simulation confirmed mild valve regurgitation seen in-vivo. Interaction between asymmetric flow and surrounding vessel resulted in asymmetric opening and closing. Poor central coaptation led to a central regurgitant jet toward the septum. Altered postoperative hemodynamics improved coaptation and eliminated regurgitation, as seen in-vivo. This modeling approach reproduced in-vivo pre- and postoperative valve performance and identified mechanisms improving coaptation after TA repair. TA repair led to elimination of regurgitation due to enhanced central coaptation. Thus, altered postoperative hemodynamic conditions after TA repair may improve valve performance without direct leaflet intervention.

  Publisher

• Simulations Predict Improved Valve Performance Without Direct Leaflet Intervention After Neonatal Truncus Arteriosus Repair

  arXiv (Cornell University) · 2026-01-13

  preprintOpen access

  Truncus arteriosus (TA) is a rare and severe congenital heart disease. Quadricuspid valve morphology occurs in 25% of all TA patients and is linked to regurgitation and increased risk of re-operation. It remains unclear how hemodynamic changes after TA repair alter valve performance. This study simulated pre- and postoperative conditions in a neonatal TA patient to investigate valve performance without direct intervention. We hypothesize that valve performance before and after truncal repair can be predicted in-silico, matching in-vivo imaging and identifying mechanisms how hemodynamic changes after repair will reduce valve regurgitation without direct intervention. Pre- and postoperative CT images of a neonatal patient with quadricuspid valve were segmented. Free edge length and geometric height from the patient's echocardiogram were used to model the valve. For the preoperative condition, ventricular pressures were set equal modeling an unrestricted ventricular septal defect. Systemic and pulmonary resistances were tuned based on the patient's Qp:Qs ratio. For the postoperative condition, boundary conditions were modified to mimic patient-specific hemodynamics after TA repair. The preoperative simulation confirmed mild valve regurgitation seen in-vivo. Interaction between asymmetric flow and surrounding vessel resulted in asymmetric opening and closing. Poor central coaptation led to a central regurgitant jet toward the septum. Altered postoperative hemodynamics improved coaptation and eliminated regurgitation, as seen in-vivo. This modeling approach reproduced in-vivo pre- and postoperative valve performance and identified mechanisms improving coaptation after TA repair. TA repair led to elimination of regurgitation due to enhanced central coaptation. Thus, altered postoperative hemodynamic conditions after TA repair may improve valve performance without direct leaflet intervention.

  PublisherDOI

• Towards personalized multiphysics heart models for clinical application

  Stanford Digital Repository · 2026-03-13

  dissertationOpen access
  PublisherDOI

• Patient‐Specific Computational Flow Simulation Reveals Adverse Hemodynamic Factors Associated With Occlusion of Directional Branches After Fenestrated‐Branched Endovascular Aneurysm Repair

  Journal of the American Heart Association · 2026-01-30

  articleOpen access

  Background Fenestrated and branched endovascular aneurysm repair can be complicated by branch vessel occlusion in the absence of structural stenosis. We hypothesized that computational flow simulation could identify adverse hemodynamic features associated with postfenestrated and branched endovascular aneurysm repair branch occlusion. Methods Patients undergoing 4‐vessel fenestrated and branched endovascular aneurysm repair for Extent II to IV thoracoabdominal aortic aneurysms were retrospectively reviewed. Branches that occluded without identifiable kinking or stenosis on computed tomography were included, along with an equal cohort of anatomy‐matched patent controls. Patient‐specific pulsatile rigid‐wall simulations were performed using SimVascular with individualized geometries and boundary conditions. Abnormal time‐averaged wall shear stress (TAWSS) was defined as &lt;10 or &gt;70 dynes/cm 2 . Results Nine patients (36 target vessels) experienced 10 branch occlusions. Postoperative pressure and flow did not differ significantly between occluded and patent renal or mesenteric branches. However, occluded renal and mesenteric branches demonstrated significantly larger postoperative areas of abnormal TAWSS compared with controls (renal: 14.5% versus 5.9%, P =0.003; mesenteric: 17.7% versus 9.9%, P =0.035). Logistic generalized estimating equation modeling showed abnormal TAWSS to be a significant predictor of renal branch occlusion ( P =0.0085). Model estimates suggested occlusion probabilities of 1.1%, 31%, and 94% at 0%, 10%, and 20% abnormal TAWSS surface area, respectively. A cluster‐bootstrapped receiver operating characteristic curve (area under the curve, 0.876) identified a &gt;10.2% threshold that correctly classified 92% of renal occlusions. Abnormal TAWSS frequently localized to distal stent–artery interfaces. Conclusions Elevated abnormal TAWSS within stented renal branches is associated with subsequent branch occlusion after fenestrated and branched endovascular aneurysm repair. Computational flow simulation‐derived TAWSS thresholds may help identify high‐risk branches before failure, warranting prospective validation.

  PublisherDOI

• Accelerated Patient-Specific Hemodynamic Simulations with Hybrid Physics-Based Neural Surrogates

  arXiv (Cornell University) · 2026-04-02

  preprintOpen accessSenior author

  Physics-based 0D reduced-order models provide computationally lightweight predictions of cardiovascular flows, resolving bulk hemodynamics in fractions of a second that would take days to solve using traditional 3D finite-element techniques. However, the accuracy of 0D models is limited as a result of the dramatic simplifications made in their derivations. In this work, we use 0D parameters learned from high-fidelity 3D data to improve 0D model accuracy without sacrificing its low computational cost or interpretability. We use the resistor-quadratic resistor-inductor (RRI) model to predict pressure drops over 0D vessels and bifurcations, where the resistances and inductance (0D parameters) are predicted from the bifurcation or vessel geometry using neural networks trained on high-fidelity 3D simulations. We validate the hybrid physics-based data-driven framework in three types of patient-specific vasculature - aortic, aortofemoral, and pulmonary anatomies. Use of learned 0D parameters reduces error by at least 50% compared to baseline 0D parameters across all anatomical cohorts. The improvements are especially marked for the more complex pulmonary anatomies, where 0D models with learned parameters reduced error from 30% to 7%. Exclusion of the quadratic resistor in the RRI model improved convergence compared to using the full RRI model. The resulting hybrid model presents a means of real-time (personal laptop runtime of &lt;2 seconds for the most complex pulmonary anatomies), interpretable, and accurate cardiovascular flow modeling, enabling digital twins that support clinical decision-making as well as cardiovascular science and engineering research.

  PublisherDOI

• Optimized biomechanical design of a tissue engineered pulsatile Fontan conduit

  npj Regenerative Medicine · 2026-01-15

  articleOpen access

  Children with congenital heart defects increasingly survive to adulthood, but the non-physiological Fontan circulation imposed by current surgical palliation leads to significant sequelae and reduced lifespan. Restoring subpulmonic pumping function remains a long-standing goal, and there have been several attempts using regenerative medicine approaches. These efforts have lacked biomechanical rigor, however, and have not achieved the requisite functionality. Here, we introduce an analytically based framework that grounds pulsatile conduit design in biomechanical principles, coupling the architecture and properties of a passive matrix with embedded myofibers to optimize performance within pediatric anatomical constraints. Parametric exploration of matrix properties and myofiber orientations yields biomechanically feasible designs. Sensitivity analyses demonstrate design robustness and highlight parameters critical for reproducible biomanufacturing and surgical implementation. To illustrate clinical potential, a patient-specific lumped-parameter hemodynamic model shows that an optimized pulsatile conduit can generate physiologically meaningful pressures and flows and outperform passive grafts.

  PublisherOA PDFDOI

Recent grants

• Improving Tissue Engineered Vascular Graft Performance via Computational Modeling

  NIH · $6.4M · 2018–2027

• Collaborative Research: SI2-SSI: A Sustainable Open Source Software Pipeline for Patient Specific Blood Flow Simulation and Analysis

  NSF · $1.2M · 2013–2015

• Enabling reliable cardiovascular simulations via uncertainty quantification

  NIH · $388k · 2016–2022

• Integrating Volumetric Light-Field with Computational Fluid Dynamics to Study Myocardial Trabeculation and Function

  NIH · $492k · 2021–2025

• CAREER: Optimization and Parameterization for Multiscale Cardiovascular Flow Simulations Using High Performance Computing

  NSF · $332k · 2015–2018

Frequent coauthors

• Irène Vignon-Clémentel

  120 shared

• Giancarlo Pennati

  99 shared

• Richard Figliola

  Clemson University

  94 shared

• Anthony M. Hlavacek

  Medical University of South Carolina

  84 shared

• Gabriele Dubini

  Politecnico di Milano

  83 shared

• G. Hamilton Baker

  Medical University of South Carolina

  81 shared

• Jeffrey A. Feinstein

  Stanford University

  80 shared

• T.-Y. Hsia

  Great Ormond Street Hospital

  68 shared

Education

• PhD, Mechanical Engineering

  Stanford University

  2005

• MSE, Mechanical Engineering

  Stanford University

  2000

• BSE, Mechanical and Aerospace Engineering

  Princeton University

  1998

Awards & honors

• Burroughs Wellcome Fund Career Award at the Scientific Inter…
• NSF CAREER award (2011)
• UCSD graduate student association faculty mentor award (2014…
• MAE department teaching award at UCSD (2015)
• Van C. Mow Medal from the ASME (2023)