Physics-Informed Conditional Diffusion for Motion-Robust Retinal Temporal Laser Speckle Contrast Imaging
RetinaDiff is a physics-informed conditional diffusion model designed to reconstruct retinal blood flow from very few laser speckle frames. It stabilizes raw data using phase correlation to correct for eye motion and uses a physical prior derived from speckle contrast to guide the reconstruction, resulting in robust flow maps despite severe data limitations.