Real-time open-field pupil tracking algorithms and their performance in optical coherence tomography for retinal imaging

Accurate pupil tracking is essential for minimizing motion artifacts and enhancing image quality in optical coherence tomography (OCT), particularly in dynamic or uncontrolled imaging environments. In this study, we evaluated the performance of six feature-based pupil-tracking algorithms, namely Blob, Swirski, ExCuSe, ElSe, PuRe, and PuReST, within a custom-built frequency-domain OCT (FD-OCT) system equipped with a pupil imaging module using 940 nm near-infrared illumination. Each algorithm was assessed using both artificial eye and real-eye datasets across three metrics: tracking accuracy, repeatability, and runtime. Results indicate that PuRe and PuReST achieved sub-millimeter accuracy and demonstrated robust tracking performance under varying lighting conditions. PuReST outperformed others in computational efficiency, while the Blob algorithm showed resilience against visible light interference. The integration of pupil tracking with OCT imaging significantly reduced motion-induced artifacts and improved retinal image quality. These findings support the use of PuReST as a practical solution for real-time motion correction in OCT systems deployed in clinical and open-field settings.