Inflammation in retinal disease has long been inferred from structural imaging and clinical signs. Direct observation at the cellular level has remained limited. Recent advances in adaptive optics imaging are beginning to close that gap, enabling visualization of immune cells and their activity in the living human eye.
At the 2026 meeting of the Association for Research in Vision and Ophthalmology (ARVO) in Denver, Ethan A. Rossi, PhD, director of the Advanced Ophthalmic Imaging Laboratory (Rossi Lab) at the University of Pittsburgh, described the use of multimodal adaptive optics scanning laser ophthalmoscopy (AOSLO) to image inflammation in patients with infectious and noninfectious uveitis, as well as retinal degenerations. His presentation focused on both the imaging hardware and the processing approaches required to detect and interpret cellular features.
AOSLO systems correct for ocular aberrations and enable high-resolution imaging of retinal structures. Nonconfocal detection methods, including split-detector and multi-offset techniques, extend this capability by capturing scattered light from structures that are not visible with conventional confocal imaging, Dr. Rossi explained. “What all of these different techniques have done has given us a new dimension to be able to explore the different structures of the retina,” he said.
In the Rossi Lab’s approach, a fiber-based detection system collects light from multiple apertures and combines the signals to generate images with improved contrast. However, Dr. Rossi emphasized that image processing has become central to recent progress. “The image processing really was the key last missing piece, I think, to really enable doing this work,” he said. New methods that integrate information across channels can reduce noise and directional bias while providing additional contrast based on refractive properties, he explained.
The study was exploratory, enrolling patients from uveitis clinics for imaging during active disease and follow-up. The goal was to capture inflammatory changes at different stages and across etiologies. Cases included ocular toxoplasmosis, syphilitic uveitis, sarcoidosis, and birdshot chorioretinopathy. Across these conditions, investigators identified a range of cellular and structural features, including circular immune cells, clusters, ramified cells, and cyst-like structures.
Some of these features appeared consistently across diseases, while others were more specific. In infectious uveitis, for example, the group observed dynamic cellular activity and patterns suggestive of pathogen-related changes. “We saw a really striking pattern of immune cell activity in this patient,” Dr. Rossi said, describing a case of ocular syphilis in which cell morphology and movement changed over short time intervals. In some instances, structures resembling macrophages or microglia were identified, although definitive classification remains uncertain.
The imaging also revealed noncellular changes associated with inflammation, including microcystic structures and vascular abnormalities. In several cases, vascular wall inflammation and immune cell adherence were visible. Longitudinal imaging demonstrated that these features could change with treatment. In one patient with birdshot chorioretinopathy, follow-up imaging showed “drastic changes…including remodeling in vasculature” and reductions in visible cellular activity over several weeks.
Dynamic imaging over seconds to minutes captured cell movement within and around vessels. In some cases, clusters of cells appeared to form or disperse over time. The ability to observe these processes directly may provide insight into disease mechanisms and treatment response. At the same time, Dr. Rossi noted that interpretation remains a challenge. “We need to really understand what we’re seeing in these patients,” he said.
Efforts are under way to quantify these observations using segmentation tools and machine learning. Early work involves building image data sets and applying algorithms to identify and track cellular features. “These powerful AI tools…have really accelerated our ability to segment and quantify these cells,” Dr. Rossi said, while noting that further data are needed to improve performance.