Generation of Fundus Fluorescein Angiography Videos for Health Care Data Sharing

Text-to-video generation of dynamic FFA sequences with a Wavelet-Flow VAE and Diffusion Transformer, for privacy-aware medical data sharing. Published in JAMA Ophthalmology (2025).

Generative AIMedical Video GenerationComputer VisionDiffusion TransformerWavelet-Flow VAEPythonPyTorch

Paper

Official JAMA Ophthalmology webpage for the FFA video generation publication — Official publication page

Overview

A text-to-video model that generates dynamic fundus fluorescein angiography (FFA) videos from clinical text descriptions, supporting privacy-aware medical data sharing. The work was published in JAMA Ophthalmology (2025; 143(8): 623–632).

Problem

Real patient imaging cannot circulate freely, which constrains research, education, and model development. Generated dynamic imaging offers a path to share clinically meaningful data without exposing patients.

My Role

Co-first author. Contributed to model development with Python and PyTorch, including the integration of the Wavelet-Flow VAE and Diffusion Transformer for spatiotemporal modeling of fundus lesion features.

Approach

The model integrates a Wavelet-Flow VAE with a Diffusion Transformer, and was trained and evaluated on 3,625 anonymized FFA videos. Generation quality and privacy protection were validated with objective metrics, text–video semantic consistency, ophthalmologist review, and image retrieval metrics.

Results

The study was peer-reviewed and published in JAMA Ophthalmology (2025). The research offers a new technical approach for clinical diagnosis and treatment support, medical education, and privacy-preserving medical data sharing.