Real-time Optical Coherence Tomography overlay in Epiretinal Membrane Peeling for enhanced intraoperative visualization: A Pilot Study

Abstract

Purpose : The integration of preoperative Optical Coherence Tomography (OCT) images into vitreoretinal surgery holds promise for optimizing surgical approaches and improving outcomes. In this pilot study, we present a real-time overlay of OCT thickness map onto surgical video to aid surgeons in intraoperative decision-making.

Methods : Five surgical videos were segmented into frames, and feature points were extracted using the Shi-Tomasi method. The optical flow between consecutive frames was calculated using the Lucas-Kanade method, ensuring corresponding feature points (Figure 1). A homography transformation matrix aligned feature points between frames and the OCT image. Corresponding feature points on the initial frame of the video and the OCT image were manually annotated, calculating the homography transformation matrix from the OCT image to the first frame. Recursively, for each video frame, we computed the homography transformation matrix from the OCT image, overlaying the transformed OCT image onto each frame. The resulting frames were reassembled to reconstruct the surgical video with the OCT overlay (Figure 2). Successful matching of two consecutive frames was manually measured, excluding frames with obscured or out-of-focus retina. Statistical analyses were not conducted due to the limited number of videos.

Results : Our method successfully achieved optical flow detection and feature point pairing in 92.7% of consecutive frame pairs. Qualitative analysis of the feature points selected between frames indicated a tendency of our algorithm to choose points near the vascular tree. Frames, with a resolution of 1024×576 WSVGA, were rendered at an average of 7.56 FPS.

Conclusions : This study demonstrates the successful and efficient overlay of OCT images onto epiretinal membrane peeling surgery videos using classical algorithms. The real-time integration of OCT data not only supports surgical decision-making but also has potential applications in surgical training and robotic-assisted vitreoretinal surgery.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.

 

Figure 1: Feature Points Used for Aligning Consecutive Frames in Surgical Video.

 

Figure 2: Overlay of Preoperative OCT Image on Surgical Video Frame. The OCT image is seamlessly superimposed onto the intraoperative view of the retina.