Higher Resolution Pupillographic Multifocal Perimetry

May 21, 2008 in Conference, Paper, Publication | No comments

IPS2008 - 18th International Visual Field & Imaging Symposium, Nara Japan, 21-24 May 2008

Authors

T Maddess, AC James, M Kolic, XL Goh. ARC Centre of Excellence in Vision Science, CVS, Australian National University, Canberra, Australia

Purpose

To investigate 4 variants of multifocal pupillographic perimetry in glaucoma that displayed 24 or 44 regions per eye.

Methods

We tested 43 normal and 44 glaucoma subjects. Glaucoma patients had moderate to severe fields in at least one eye. All subjects were examined with HFA achromatic, SWAP and Matrix 24-2 perimetry, Stratus OCT, slit lamp and tonometry. Informed written consent was obtained from all subjects under ANU ethics approval 238/04.

Multifocal stimuli were presented concurrently to both eyes, having 24 or 44 independent test regions/eye in a dartboard layout, extending to 30 deg eccentricity. Four stimulus variants were examined in which stimuli were presented at 4/s, 1/s, or 1/4s, each stimulus persisting for 33 ms. Maximum and background luminances were 10 and 290 cd/m2. All stimuli were yellow. Recording duration was 4 minutes, divided into 8 segments of 30s. Pupil diameter was monitored by video cameras under infrared illumination. Data from fixation losses and blinks was automatically excluded. Up to 15% data loss from blinks and fixation losses were permitted before a 30s segment was repeated. Measures of field loss examined the N-worst amplitudes, areas under the response, response delays, or pair-wise linear combinations of those.

Results

For these simple measures linear discriminant functions produced areas under ROC curves (AUCs) of 0.80 for the 44 region 1/s stimulus for severe fields, and 0.73 for mild fields. AUCs were consistently smaller for the 24 region, and slower 44 region, stimuli.

Conclusions

The simple N-worst measures basically examined the sum of the N-worst response measures regardless of where in the field they occurred, and were appropriate for comparing the relative performance of the 24 and 44 region methods. Having 44 regions per eye may improve the scope for detecting clusters of damage, which as not been examined so far. The method eliminates problems associated with false positive and negative errors, and fixation losses found in conventional perimetry, all of which effectively lower sensitivity and specificity.