City Research Online

Abstract

A digital imaging system has been developed for the objective assessment of severity and morphology of posterior capsule opacification (PCO). The system is based on a Nikon FS-2 photo slit-lamp and a Kodak DCS 100 digital camera system. The slit-lamp was used with conventional (forward) illumination and studies were carried out to determine the best combination of slit-lamp and camera parameters to optimize the images. These were found to be a slit beam width of 3.5mm, slit angle of 45°, film speed (ISO) of 1600 for the camera and 800 for the digital storage unit (DSU), exposure compensation 0, flash intensity 2 and a magnification of 25x. A fixation stimulus was developed since it was shown to decrease the variability of the images and reduce the possibility of unwanted reflexes affecting the area of interest. In all cases a central 3mm x 3mm area of the posterior capsule, centred on the pupil, was analysed. The image processing software used throughout the work was Image Pro+ v4.51 (Media Cybernetics, Silver Spring, MD).

For validating the system, subjects who had previously had uncomplicated cataract surgery were recruited from the Nd:YAG capsulotomy clinics of the Royal Eye Unit at Kingston. Intra-observer repeatability was assessed by taking 6 repeat images for each of 10 subjects. The coefficient of variation for total spectral power in the image, a measure of the severity of opacification, was 1.5%. Inter-observer repeatability was tested for 3 observers, an ophthalmologist, an optometrist and a non-clinician, by recording 2 images for each observer for each of 9 subjects. There was a 6 fold increase in the limits of agreement when the non-clinician was involved indicating that suitable training is required to use the system in a repeatable fashion.

Severity of opacification was measured by calculating the log of the total spectral power in the Fourier Transform image of the area of interest. Results on 45 eyes correlated well with subjective grading of PCO by an experienced ophthalmologist: Within the 3 morphological groups, fibrotic, mixed and pearl, correlation coefficients of 0.909, 0.864 and 0.713 were found, which were all statistically significant.

Morphological classification was carried out by developing a simple expert system. The rules in the knowledge base were derived from known histological features of PCO and clinically observed changes. The weighted kappa score for agreement between the objective classifier and an experienced ophthalmologist was 0.61 indicating a good level of agreement. However 14 out of 45 cases did not exhibit agreement. It may be possible with further work to improve the classification rules to reduce this number.

New methods have been presented for objectively measuring the severity and morphologically classifying PCO. The system developed in this work shows distinct potential for use within primary care environments since it uses techniques and instruments whose operation is familiar to all practitioners involved. The validation studies presented here demonstrate variability in the measurements that is largely in agreement with previously published systems.

Publication Type:	Thesis (Doctoral)
Subjects:	R Medicine
Departments:	School of Health & Psychological Sciences > Optometry & Visual Sciences School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads