Automatic intraocular lens segmentation and detection in optical coherence tomography images

We present a new algorithm for automatic segmentation and detection of an accommodative intraocular lens implanted in a biomechanical eye model. We extracted lens curvature and position. The algorithm contains denoising and fan correction by a multi-level calibration routine. The segmentation is realized by an adapted canny edge detection algorithm followed by a detection of lens surface with an automatic region of interest search to suppress non-optical surfaces like the lens haptic. The optical distortion of lens back surface is corrected by inverse raytracing. Lens geometry was extracted by a spherical fit. We implemented and demonstrated a powerful algorithm for automatic segmentation, detection and surface analysis of intraocular lenses in vitro. The achieved accuracy is within the expected range determined by previous studies. Future improvements will include the transfer to clinical anterior segment OCT devices.