Sensor-driven digital motion correction of robotically-aligned optical coherence tomography retinal volumes

Optical coherence tomography (OCT) has revolutionized diagnostics in retinal ophthalmology. Traditional OCT requires minimal relative motion between the subject and scanner, which is difficult to achieve with handheld devices and/or non-stabilized subjects. We recently introduced robotically-aligned OCT (RAOCT) as an alternative that promises to alleviate these minimal-movement requirements by tracking the subject and compensating for their motion with dynamic hardware components in real-time. However, hardware and image processing delays lead to residual motion artifacts even after automatic alignment and motion compensation. Here, we introduce a novel sensor-driven digital motion correction approach that overcomes these shortcomings. Our method leverages synchronized sensing of both the subject's eye and the scanner hardware to continuously estimate the imaging system state during acquisition. The A-scans are then remapped using a ray-tracing model of the system at the precise moment of acquisition. We demonstrate that, in addition to motion compensation from RAOCT, our method further reduces residual artifacts by 88.3 %, 80.4 %, and 62.6 % across axial, lateral, and rotational motions, respectively. We also show our correction in human retinal OCT images where residual errors from acquisition were reduced down to 12.4 µm, 0.11°, and 0.39° for axial, lateral, and rotational motion, respectively.