Automatic measurement of the angle of strabismus in different directions of gaze with unrestrained head

Background Inaccuracy in the measurement of the angle of strabismus is an important cause of reoperations in strabismus surgery. The current methods to estimate the angle of strabismus have limited accuracy, the duration of the test is long and a high level of cooperation of the patient is required. To alleviate this the Delft Assessment Instrument for Strabismus in Young children (DAISY) is being developed. DAISY aims to measure the angle of strabismus accurately (error <1 degree) in different directions of gaze within a short time span (<1 minute) of young children with unrestrained head. Methods Strabismus angles in different directions of gaze with free head movement were estimated using the rotations of both eyes and the head pose. The head rotation was estimated with a face tracking system (FaceAPI) and a webcam, which allowed free head movement. Stereo images were obtained with two Prosilica GC2450 cameras. The eye rotations were estimated with these images, using the coordinates of the pupil centers, fixation objects (LEDs) and corneal reflections centers. These reflections were created by two infrared light sources. Validation experiments were performed on orthotropic test subjects. All experiments were performed at a measurement distance of approximately 1 meter by roomlight. I. The error of the head rotation method was determined with a high precision secondary measurement system on ten adults and ten children. Eye rotation was described relative to the optical axis through the midpoints of cornea and pupil center. II. The errors of the estimation of the angle of strabismus on one adult using a fixed head pose were identified. III. The errors of the estimation of the angle of strabismus with free head movement were determined for gaze ahead and for different directions of gaze on three adults. Results I. The error of the head rotation estimation with the FaceAPI was smaller than 5 degrees for 99.5 percent of the adults' and for 99.2 percent of the children's measurements. The median of the error of the adults' data was 1.40 degrees and was 1.23 degrees for the children's data. II. The strabismus angles with fixed head were estimated with an error smaller than 0.5 degrees for gaze directions up to 15 degrees relative to gaze ahead. III. The error of the estimated horizontal angle of strabismus with unrestrained head was 0.34 degrees for gaze ahead and 0.53 degrees for different directions of gaze. The run-time for a single measurement was 2 milliseconds. Conclusion The head rotation could accurately be assessed for both adults and young children. The measurement time was reduced considerably, from 10 minutes to 2 milliseconds for a single measurement. The angle of strabismus could be measured accurately in different directions of gaze while allowing free head movement. The error was reduced from 5.44 to 0.53 degrees. This research shows that the variance introduced by the inaccurate measurement could be reduced with a factor ten.

Subject

strabismus angle
eye rotation
head rotation
angle of gaze
unrestrained head

To reference this document use:

http://resolver.tudelft.nl/uuid:ae132060-f1a8-4b7f-b5fe-4965a29270a9

Embargo date

2015-10-08

Part of collection

Student theses

Document type

master thesis

Rights

Files

PDF

Thesis_Nicole_Bakker.pdf

63.57 MB

Close viewer