ARVO 2026: 3D imaging and deep learning for retinal disease

In this interview, Jay Chhablani, MD, a retina specialist at the UPMC Vision Institute and director of the Choroidal Analysis and Research Lab at the University of Pittsburgh, discusses his team’s cutting-edge work in 3D choroidal imaging and AI-driven retinal research. Using the Plex Elite platform from Zeiss, his group has developed tools to visualize the choroidal vasculature in three dimensions from in vivo imaging, enabling a more nuanced understanding of choroidal structure and disease.

Chhablani explains that their research shows detectable choroidal changes in early age-related macular degeneration (AMD), even before overt disease manifestations. These alterations in the choroidal vasculature continue to evolve as the disease progresses, positioning the choroid as a promising biomarker for early detection, disease monitoring, and treatment response. The team is also examining choroidal changes in central serous chorioretinopathy, a condition where choroidal vessels play a critical pathogenic role, including comparative analyses of fellow eyes.

A key strength of the lab is its multidisciplinary structure, combining machine learning engineers with clinical research fellows to advance artificial intelligence applications in ophthalmic imaging. Among their featured projects are efforts to differentiate real OCT scans from synthetic OCT data and an oral presentation on inter-device AMD classification. In that work, a student trained a deep learning model on Cirrus OCT data and then adapted it to Spectralis scans through fine-tuning, improving AMD classification performance while using relatively limited Spectralis data. Collectively, these initiatives highlight how advanced imaging and AI can bridge real-world challenges, such as mixed-device OCT datasets, and move retinal care toward more precise, data-driven decision-making.