Development of an Automated Segmentation Program to Assess Corneal Suturing Performance of Ophthalmology Residents Using 3D Printed Eye Models

Abstract

Purpose : To develop an automated segmentation program to assess corneal suturing performance of ophthalmology residents using 3-D printed eye models and to compare the results with a semi-manual program.

Methods : After cornea suturing, on BIONIKO (Miami, FL) 3D printed eye models, by ophthalmology residents at Bascom Palmer Eye Institute, one eye model was randomly selected to be analyzed using our custom developed segmentation program. A digital camera (Canon EOS 40D DSLR & EF 100mm f/2.8 Macro Lens) was mounted on a microscope stand (AmScope TS130-LED, Irvine, CA) equipped with a centration adaptor holding the 3D printed eye model (Fig 1). Images were taken and imported into MATLAB (Mathworks, Natick, MA) to assess spacing and suture length. Images were preprocessed using an averaging filter and edge detection methods were utilized to isolate sutures and points in the circle of trephination. Spacing was determined by finding the distance between intersection points of the circle of trephination and adjacent sutures. Suture length was determined by the distance between the endpoints of the suture (Fig 2). Results were compared to data obtained with the semi-manual segmentation program previously developed in the laboratory (Alawa K et al, IOVS 2016; 57(12):1224.).

Results : Segmentation of one image took 16.5 seconds with the automated segmentation program and 61.08 seconds with the semi-manual segmentation program. No significant difference was found between the automated and semi-manual programs regarding spacing and suture length. Mean spacing for the 8 sutures was 6.39±0.85 mm using the automated program and 6.33±1.04 mm using the semi-manual program (p=0.92). Mean suture length was 2.53±0.58 mm using the automated program and 2.77±0.47 mm using the semi-manual program (p=0.30).

Conclusions : The automated program was able to calculate the spacing and suture length and yielded results consistent with those obtained with the semi-manual program. Further refinements and calculation of radiality will be needed to optimize the objective assessment of corneal suturing performance provided by this novel automated program.

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

 

(A) Side view and (B) front view of the camera mount set up with a 3D printed eye model.

 

Original image (A) and a superimposed image (B) of 3D printed eye model with the endpoints and circle of trephination detected from the automated software.