Assessment of tear film using videokeratoscopy based on fractal dimension

TL;DR

This study introduces a novel fractal dimension-based method for assessing tear film stability using high-speed videokeratoscopy, demonstrating improved robustness and correlation with manual evaluations over previous techniques.

Contribution

The paper presents a new fractal dimension approach for tear film assessment that outperforms existing gray level co-occurrence matrix methods in stability analysis.

Findings

Fractal dimension method is less affected by eye movements.

High correlation (r2 > 0.8) with manual tear film stability measurements.

Potential for more detailed tear film dynamics characterization.

Abstract

Purpose: To develop and test a new method for characterizing tear film surface quality with high speed videokeratoscopy utilizing a fractal dimension approach. Methods: The regularity of the reflected pattern in high speed videokeratoscopy (E300, Medmont) depends on tear film stability. Thus determining tear film stability can be addressed by estimating the fractal dimension of the reflected pattern. The method is tested on 39 normal subjects. The results of the fractal dimension approach are compared with those obtained using previously proposed automated method, based on a gray level co-occurrence matrix approach, and with subjective results obtained by two operators which were assessing the video recordings in ideal conditions. Results: fractal dimension method was less affected by eye movements and changes in the videokeratoscopic image background than gray level co-occurrence matrix method. Median difference of the non-invasive break-up time between manual and automated methods was 0.03 s (IQR = 4.47 s.) and 0.0 s (IQR = 2.22 s.) for gray level co-occurrence matrix and fractal dimension approaches, respectively. Correlation coefficient with manual non-invasive break-up time was r2 = 0.86 (p < 0.001) for gray level co-occurrence matrix approach, and r2 = 0.82 (p < 0.001) for fractal dimension approach. Significant statistical difference was found between non-invasive break-up measurements of manual and gray level co-occurrence matrix method (p = 0.008). Conclusions: The proposed method has the potential to characterize tear film dynamics in more detail compared to previous methods based on high speed videokeratoscopy. It showed good correlation with manual assessment of tear film.