Evaluation of Automatic Following of Anatomical Structures for Live Intraoperative OCT Repositioning

Abstract

Purpose : The scan location of microscope integrated optical coherence tomography is primarily controlled through the foot panel joystick. We study the feasibility of automatically following an arbitrary structure by tracking the surgical view and updating the OCT scan location during porcine wet labs and on recorded operations.

Methods : Two Lumera 700 microscopes equipped with RESCAN 700 (Zeis, Oberkochen) were modified to allow the surgeon to set the scan location of the live iOCT at any position such as macula to be automatically followed and acquire cross-sectional scans continuously. The system needed to be robust to illumination changes, instrument occlusion, out of focus views, and other common issues during actual surgeries. Two groups of surgeons including 5 posterior-segment surgeons and 1 anterior segment surgeon tried the modified system during wet labs. Following wet lab with porcine eyes, all surgeons answered yes or no to the questions:
1- Is the feature beneficial during surgery?
2- Is the system sufficiently fast?
The second group of surgeons also answered yes or no to the questions:
3- Is the tracker robust to the illumination change?
4- Is the tracker robust to the occlusions by surgical instrument?
In addition to the subjective tests, the performance of the tracker was evaluated by reviewing individual frames from several recorded operations. Landmarks were manually annotated every 5 frames and their position was compared with the estimated position from the algorithm.

Results : The answers of the surgeons to the questionnaire are reported in the following table. A total of 10250 fundus view frames from 38 operations were reviewed for accuracy and figure 1 shows the performance of the algorithm in following structures with different magnifications. The accuracy of the algorithm is measured as 11.19 ± 9.48 pixels and its Bland-Altman plot is shown in figure 2.
Q1 (5 Y 1 N)
Q2 (6 Y 0 N)
Q3 (1 Y 1 N.A.)
Q4 (2 Y 0 N)

Conclusions : Most surgeons participating in this study agree on the performance of the proposed feature for following anatomical structures. This is supported by the quantitative analyses.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.

 

Figure 1 Performance of following structure on different surgical videos.

 

Figure 2 Bland-Altman plot of the tracking accuracy.