CompHealth Corner August 2026
Computational and digital health ideas evolve quickly. Our monthly research roundup highlights the latest publications — from our Center faculty and other leaders in the field. Here’s what you should be reading this month!
Can virtual reality help patients recover while still in the ICU?
Rhythms of Recovery: Patient-Centered Virtual Reality Exergame for Physical Rehabilitation in the Intensive Care Unit
Sangjun Eom; Tianyi Hu; Wenyi Xu; Liheng Zou; Ernesto Escobar; Gabriel Streisfeld; Anna Mall; Bradi Granger; Maria Gorlatova. “Rhythms of Recovery: Patient-Centered Virtual Reality Exergame for Physical Rehabilitation in the Intensive Care Unit.” IEEE Transactions on Visualization and Computer Graphics (2026).
A new study published in IEEE Transactions on Visualization and Computer Graphics from the lab of CCDHI Associate Director of XR Maria Gorlatova explores how virtual reality could support early physical rehabilitation for patients in intensive care.
The team developed a patient-centered VR exergame specifically designed around the challenges of the ICU environment. Using Meta Quest 3 body tracking, the system guides patients through progressively more challenging upper-body movements while providing an immersive game experience and allowing the care team to monitor patients and quantify their movement.
The researchers evaluated the system with 18 cardiothoracic ICU patients. Patients reported high enjoyment, and the ICU care team found the system easy to use, safe, and effective for bedside use. As patients progressed through the game, they also demonstrated increases in movement speed, range of motion, and workspace volume.
Importantly, the work shows how XR technologies could do more than make rehabilitation engaging. By capturing quantitative measures of movement while patients exercise, these systems could provide new ways to track physical activity and recovery in critical care, where motor progress is often difficult to measure objectively.
As extended reality becomes increasingly integrated with healthcare, studies like this demonstrate how immersive technologies can be designed around real clinical workflows to support more engaging, measurable, and patient-centered rehabilitation.
What keeps therapeutic antibodies from reaching cancer cells inside solid tumors?
Single-cell spatial pharmacobiology identifies conserved stromal barriers to therapeutic antibody delivery in human solid tumors
Guolan Lu, John W. Hickey, Maximilian Haist, Xulei Qin, Emily Zhao, Abdullah Naveed, Erna Forgo, Marc-A. Baertsch, Lucas Mani, Xavier Rovira-Clavé, Andrey Finegersh, Yury Goltsev, Chiara Caraccio, Nynke S. van den Berg, Marisa Hom, Deana R. Colburg, Brock A. Martin, Christina S. Kong, Natalie S. Lui, George A. Fisher, A. Dimitrios Colevas, Robert B. West, Greg M. Thurber, George A. Poultsides, Garry P. Nolan & Eben L. Rosenthal. “Single-cell spatial pharmacobiology identifies conserved stromal barriers to therapeutic antibody delivery in human solid tumors.” Nature Biotechnology (2026): 1-15.
A new study published in Nature Biotechnology, co-authored by CCDHI member and Co-Associate Director of AI John Hickey, introduces single-cell spatial pharmacobiology (SSP), a new approach for mapping how therapeutic antibodies move through human tumors and engage their targets at single-cell resolution.
The researchers combined imaging of a systemically infused, fluorescently labeled therapeutic antibody with highly multiplexed spatial proteomics, allowing them to simultaneously visualize the drug, its molecular target, individual cell types, and the surrounding tumor microenvironment.
Applying the approach to head and neck squamous cell carcinoma and pancreatic ductal adenocarcinoma revealed striking differences in antibody delivery not only between tumors, but from cell to cell within the same tumor. The team identified conserved stromal barriers associated with reduced antibody penetration, including periostin-rich extracellular matrix structures and FAP-positive cancer-associated fibroblast neighborhoods.
Importantly, the work shows that understanding whether a therapy reaches its target may require looking beyond conventional measurements of drug concentration. By resolving drug delivery, target engagement, and the tumor microenvironment together, SSP provides a way to investigate why some regions of a tumor respond to antibody therapies while others may remain protected.
This type of spatially resolved pharmacology could ultimately help researchers uncover mechanisms of treatment resistance, optimize dosing strategies, and identify spatial biomarkers that inform more precise cancer therapy.
Congratulations to John and the entire team on this exciting work!
What if machine learning could work alongside a scientific simulation while it’s still running?
Instrumenting Lightweight, Modular Machine Learning Training and Inference in Parallel Solvers
Ayman Yousef, Corey Wetterer-Nelson, Mengjiao Han, Victor Mateevitsi, Joseph Insley, Silvio Rizzi, Janet Knowles, Michael E. Papka & Amanda Randles. “Instrumenting Lightweight, Modular Machine Learning Training and Inference in Parallel Solvers.” International Conference on Computational Science. Cham: Springer Nature Switzerland, 2026.
A new study presented at ICCS 2026, led by CCDHI member Ayman Yousef in Amanda Randles’ lab at Duke University, introduces a lightweight, modular framework for bringing machine learning training and inference directly into large-scale scientific simulations.
The team tested the framework using a mini-app of HARVEY, the lab’s high-performance blood-flow simulation code. Rather than treating simulation and machine learning as separate stages, the approach allows ML models to train and make predictions during a simulation, without requiring major changes to the underlying scientific software.
This matters because today’s large-scale simulations can generate enormous amounts of data, often requiring researchers to save results, transfer them elsewhere, and analyze them after the simulation has finished. Integrating ML directly into the running solver creates opportunities to analyze data as it is generated and ultimately build simulations that can adapt based on what they are learning.
By lowering the barrier to combining machine learning with high-performance computing, this work provides a foundation for more adaptive, data-driven scientific simulations across computational science.
Congratulations to Ayman and the entire team on this exciting work!
