|Title||AUTOMATED MEASUREMENT OF LENS THICKNESS USING OPTICAL COHERENCE TOMOGRAPHY|
|Author, Co-Author||Yuanjie Zheng, James Gee, Benjamin Backus, Kathryn Richdale|
|Abstract|| PURPOSE: To develop and evaluate an automated image analysis algorithm for measuring lens thickness (LT) from Visante Anterior Segment Optical Coherence Tomography (OCT) images.
METHODS: Measures of the right eyes of 91 subjects from a previous study were made using OCT (un-accommodated to 0D target and after cycloplegia), magnetic resonance imaging (MRI, un-accommodated only) and ultrasound (US, cyclopleged only). US data were provided automatically and MR images were made manually with digital calipers. An algorithm was developed to measure LT from OCT raw image files. The algorithm detected the anterior and posterior lens surfaces and computed max LT along the scan acquisition line. Visual inspection was conducted on marked images. A conversion of 128 pixels/mm and a refractive index (RI) of 1.4 were applied. Data from up to 6 images were averaged. Methods were compared using Bland-Altman.
RESULTS: The algorithm took < 0.35 seconds/image and < 10% of images failed inspection. There were small but significant difference between OCT & US (4.00±0.30 vs 3.96±0.29; p<0.001) and OCT & MRI (4.04±0.30 vs 3.97±0.29; p<0.001). Both methods were well correlated (OCT vs MRI, r=0.94, p<0.001; OCT vs US, r=0.99, p<0.001). There was no bias between methods (p>0.39).
CONCLUSIONS: The algorithm provided fast, reliable measures of LT. The small differences compared to US and MRI were likely due to assumptions of speed of sound and RI. The images for which the algorithm failed were mostly of poor quality and could not be reliably measured using manual methods. The automated algorithm will aid in studies of myopia and presbyopia where numerous, objective, accurate LT measurements are required.
ADDITIONAL COMMENTS: NIH K23-EY019097
|Affiliation of Co-Authors||University of Pennsylvania|