Statistical methods for the analysis of visual field data in glaucoma

Gardiner, SK, 2003. Statistical methods for the analysis of visual field data in glaucoma. PhD, Nottingham Trent University.

Preview

Text 10183217.pdf - Published version Download (21MB) | Preview

Abstract

Glaucoma is a leading cause of visual disability. Automated static perimetry determines the minimum contrast that a subject can perceive (called the sensitivity) at individual locations within their visual field. This instrumentation currently offers the most reliable strategy for the clinical follow-up of the disease. The accurate detection of glaucomatous change in a series of visual field results is important in the clinical management of a patient, and in the evaluation of which treatments are most effective in arresting progression. The slow, often equivocal rate of sensitivity loss, and the variability that exists between field results, makes this a difficult task.

The purpose of this thesis is to improve the statistical methods used for determining true visual field defects and progression in suspected glaucoma patients.

Pointwise linear regression models for identifying true change in the sensitivity at a point are examined. By simulating visual field data, it is shown that a testing frequency of three tests per year is optimal for a given rate of sensitivity loss. Next, the use of confirmation fields to verify suspected progression is examined, and the comparative specificity and sensitivity of different confirmation techniques assessed; a novel technique where the penultimate field is omitted is found to be the most specific.

A novel spatial filter is developed, based on pair-wise covariances between different points in the fields of patients at a tertiary glaucoma clinic. The filter is found to closely reflect the anatomical structure of the retinal nerve fibre layer. This filter appears to reduce the noise present in visual field tests, as evidenced by an improved fit to pointwise linear regression, fewer false positives and fewer false negatives when looking for progression. It has been programmed into computer software used to identify progressing defects.

The effect of applying the filter to patient data is proposed as a measure of the amount of noise present. This noise is modeled statistically to determine the suitability of assuming that the noise is normally distributed. The assumption is found to be not unreasonable but inaccurate; an improved model is developed which fits the empirical data better and over a wider range of sensitivities.

These new quantitative methods, in particular the new spatial filter, may prove clinically useful in the analysis of visual function deterioration in glaucoma.

Item Type:	Thesis
Creators:	Gardiner, S.K.
Date:	2003
ISBN:	9781369314830
Identifiers:	Number Type PQ10183217 Other
Divisions:	Schools > School of Science and Technology
Record created by:	Jill Tomkinson
Date Added:	30 Sep 2020 13:45
Last Modified:	03 Aug 2023 11:06
URI:	https://irep.ntu.ac.uk/id/eprint/41034

Actions (login required)

Edit View

Statistics

Views

Downloads