In recent years, the intricate process of wound healing has intrigued biomedical researchers striving to unravel the complexities beneath the skin’s surface. While external observations provide limited insight, a revolutionary approach developed through a collaboration between Duke University biomedical engineers and Nokia Bell Labs is set to transform how clinicians and scientists understand and monitor wound repair. This partnership has yielded a cutting-edge platform leveraging multimodal optical coherence tomography (OCT) combined with advanced deep learning algorithms, ushering in a new era of non-invasive, high-resolution, and quantitative wound analysis.
Optical coherence tomography, traditionally a cornerstone technology in ophthalmology, generates three-dimensional images by capturing depth-resolved reflections of light within tissue. This imaging modality offers micron-scale resolution, enabling visualization of microstructural features without disrupting living tissue. By adapting OCT for wound examination, the Duke-Nokia team has overcome longstanding limitations in wound monitoring, bypassing the need for invasive biopsies and less informative visual assessments. This approach allows clinicians to peer beneath wound surfaces and track healing dynamics in real time.
However, imaging alone produces a flood of complex data, including structural tissue morphology and capillary blood flow patterns. To extract meaningful biological insights, the researchers integrated artificial intelligence methods within their pipeline. Custom deep learning models trained on expansive imaging datasets enable automated segmentation and quantitative analysis of wound components over time, such as granulation tissue development and vascular remodeling. This advancement shifts wound assessment from subjective interpretation to objective, reproducible metrics, enhancing clinical decision-making.
The team employed this powerful OCT-AI system to evaluate hydrogel-based therapies engineered in the Gerecht laboratory at Duke. Hydrogels, biocompatible polymer networks capable of mimicking extracellular matrices, have shown promise as scaffolds that promote tissue regeneration and repair. Critically, the mechanical properties of these hydrogels influence healing trajectories, but the precise relationship remains incompletely understood. Through longitudinal imaging in murine wound models, the researchers contrasted the effects of soft versus stiff hydrogels on granulation tissue formation and vascularization.
Remarkably, wounds treated with stiffer hydrogels exhibited accelerated formation of granulation tissue characterized by a smooth, glassy morphology, indicative of rapid maturation toward functional tissue architecture. These wounds also showed quicker restoration of blood vessel integrity and flow, underscoring the critical role of mechanical cues in orchestrating healing. Conversely, softer hydrogels, while fostering more abundant but slower remodeling vasculature, demonstrated prolonged healing timelines. This nuanced insight challenges conventional views and underscores the importance of tuning hydrogel stiffness for optimal therapeutic outcomes.
The imaging modalities employed integrate structural OCT with angiographic capability, allowing simultaneous visualization of tissue morphology and dynamic blood flow within microvessels. This multimodal approach captures temporal evolution of both cellular and vascular components of wound healing with unprecedented detail. The AI algorithms analyze these datasets to quantify parameters such as tissue thickness, vessel density, and perfusion, generating comprehensive profiles of how biomaterials influence regeneration in vivo.
One of the most compelling aspects of this platform is its capacity for real-time monitoring without compromising wound integrity. Traditional biopsies, while informative, disrupt the delicate balance of the healing environment and cannot be performed repeatedly. In contrast, this OCT-AI system enables continuous, non-destructive surveillance of healing dynamics, providing a powerful tool for both research and clinical practice. It sets the stage for personalized medicine approaches that tailor interventions based on ongoing wound response.
Looking ahead, the researchers envision extending the platform’s capabilities to address clinically challenging scenarios such as chronic wounds frequently encountered in diabetic patients. These wounds are notoriously difficult to heal due to impaired vascularization and systemic complications. By refining predictive models using diverse datasets, the OCT-AI system has the potential to forecast healing trajectories and guide timely therapeutic adjustments, improving patient outcomes and reducing complications.
The success of this endeavor relies heavily on the interdisciplinary synergy between biomedical engineering, optics, computer science, and clinical insight. Collaboration with Nokia Bell Labs provided expertise in instrument design and AI development, while Duke’s biomedical engineering team ensured biological relevance and translational focus. This confluence of skills demonstrates the transformative power of combining cutting-edge technology with deep domain knowledge to solve complex biological problems.
Further technical details illuminate the imaging system architecture: the team engineered a custom OCT device optimized for skin wound imaging, balancing spatial resolution, penetration depth, and acquisition speed. The device’s light source and detection modules were calibrated to maximize contrast for both structural and angiographic signals. Meanwhile, the deep learning pipeline utilized convolutional neural networks tailored for volumetric data, trained with annotated ground truth from histological comparisons, ensuring robust performance across experimental conditions.
Together, this innovative research opens exciting avenues not only for enhancing wound care but also for broadening applications of multimodal OCT in regenerative medicine and bioengineering. It exemplifies how merging imaging innovation with artificial intelligence can yield powerful diagnostic tools capable of revealing the hidden choreography of tissue repair, ultimately paving the way toward faster, more effective healing for patients worldwide.
Subject of Research: Animals
Article Title: Multimodal OCT with Deep Learning Reveals In Vivo Healing Dynamics in Hydrogel-Treated Wounds
News Publication Date: 20-Mar-2026
Web References:
https://www.cell.com/cell-biomaterials/fulltext/S3050-5623(26)00078-4
http://dx.doi.org/10.1016/j.celbio.2026.100422
References:
Jiyeon Song, Shreyas Shah, Makenzie Bushold, Michael S. Crouch, Sohini Sarkar, Nicole Hanson, Bibek R. Samanta, Ya Guan, Michael S. Eggleston, Sharon Gerecht. Multimodal OCT with Deep Learning Reveals In Vivo Healing Dynamics in Hydrogel-Treated Wounds. Cell Biomaterials, 2026. DOI: 10.1016/j.celbio.2026.100422
Image Credits: Duke University
Keywords
Medical imaging, Tissue repair, Physiology, In vivo imaging, Data analysis, Image processing, Optics, Tissue regeneration, Tomography
