Optometry and Vision Sciences - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemCustomized, automated stimulus location choice to improve visual field procedures( 2014)It is well accepted in the literature that test-retest variability is a key limitation in current automated perimetric procedures, especially with moderate to advanced visual field loss. This obfuscates the ability to discriminate between true progression of the visual field and measurement noise. Studies have shown that sampling the visual field at a greater spatial resolution not only reduces global measures of test-retest variability, but can also describe the spatial extent of scotomata with more accuracy and detect scotomata otherwise missed in conventional perimetry. However, to date there has been no automated application of these principles, and previous approaches which have employed high resolution sampling have sacrificed test duration in order to measure more locations than standard stimulus distribution patterns. The primary objective of this thesis was to address each of these problems by developing a novel, automated, perimetric approach which customises test location choice based on spatial information and observer response during the examination. It was not only important that new procedures did not increase test times compared to current procedures, but also are computationally feasible to implement. The experimentation reported in this thesis concentrated on computer simulation (Experiments 1 to 3) in order to develop and tune these new procedures before testing their performance on human observers (Experiment 4). The outcome measures were precision of threshold estimates (test-retest variability), accuracy (absolute error) and efficiency (number of presentations). It was discovered that one of the developed approaches, Gradient-Oriented Automated Natural Neighbour Approach (GOANNA), improved accuracy and precision in areas surrounding scotoma borders without increasing test duration compared to current procedures (Chapter 3, Experiment 2). This led to further exploration of GOANNA, whereby it was demonstrated (through computer simulation) that the improvement in scotoma characterisation seen in Experiment 2 gave rise to earlier and more accurate detection of glaucomatous progression. It was assumed that the assumptions made in the simulation studies hold true for real data. This thesis lends support to previous findings that implementing high resolution grids is beneficial in glaucomatous progression detection, and that the conventional 6° rectangular grid of fixed locations may not be the most suitable stimulus arrangement for characterising and monitoring all visual field defects. It also reports on initial approaches that didn’t work, which may be useful for future investigators in this field of research. Most importantly, it provides a novel, principled, automated approach of locally increasing sampling spatial resolution without having to sacrifice efficiency.