Real-time Corneal Segmentation and 3D Needle Tracking in Intrasurgical OCT

Abstract

Purpose : Intra-operative microscope-integrated optical coherence tomography provides cross-sectional images that offer qualitative estimates of depth. In this work, we present methods for quantitative depth measurement using real-time segmentation and needle tracking in OCT volumes and demonstrate these methods for guidance of needle insertions in an ex vivo model of deep anterior lamellar keratoplasty (DALK).

Methods : We evaluated our segmentation and tracking in the 27-gauge needle insertion step of DALK in human donor corneas mounted on a Barron artificial anterior chamber. Three surgical fellows each completed 20 insertions on three different corneas. Surgeons were instructed to insert the needle to between 80% and 90% of the corneal thickness as in DALK. In 10 insertions, surgeons viewed the procedure through the surgical microscope. In the other 10 insertions surgeons viewed the procedure through the microscope and on a nearby monitor displaying a tracked, cross-sectional OCT view along the axis of their needle (Fig. 1) with a live estimate of the needle’s depth. We compared the number of perforations and the final needle depth between the two visualizations. This study was approved by the Duke IRB.
The OCT view was provided by a 100 kHz swept-source system with a volume refresh rate of 1.5 Hz. Volume dimensions were 5.47mm x 12.0mm x 8.0mm. The tracking and live estimate of needle depth was performed by implementing a graph search segmentation method to segment two corneal surfaces and a 3D model fitting algorithm to locate the needle. Needle depth was defined as the location of the needle tip relative to the anterior and posterior corneal surfaces.

Results : Surgeons perforated the cornea in 2 of 30 trials using the tracked OCT view with dynamically updated needle depth. With only the microscope view, perforations occurred in 15 of 30 trials. The mean percent depth for non-perforation OCT trials was 78.58% ± 6.68% (N = 28) and the mean percent depth for non-perforation microscope only trials was 61.38% ± 17.16% (N = 15) (Fig. 2).

Conclusions : Using our real-time, quantitative needle depth estimate, surgeons perforated less frequently and achieved higher non-perforated penetration depths than when using only the surgical microscope in this ex vivo model.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.

 

A) Segmented cross-section along needle axis. B) Segmented volume rendering.

 

Final needle depth for trial types. Blue X indicates mean plus/minus one standard deviation.