Structure-Function Correlation Across The Central Visual Field Using Pointwise Comparisons and Ganglion Cell Isocontours Derived From Pattern Recognition

We presented an abstract for 23rd International visual field and imaging symposium in Kanazawa, Japan on May 12th titled, Structure-function correlation across the central visual field using pointwise comparisons and ganglion cell isocontours derived from pattern recognition.  Authors are:

Kalloniatis M1,2, Tong J1,2, Yoshioka N1,2, Khuu SK2, Phu J1,2, Choi A1,2, Zangerl B1, Nivison-Smith L1,2, Bui BV4, Marc RE3, Jones BW3

  1. Centre for Eye Health, University of New South Wales (UNSW), Kensington, NSW, Australia.
  2. School of Optometry and Vision Science, UNSW, Sydney, NSW, Australia.
  3. Moran Eye Center, Univ of Utah, Salt Lake City, UT, United States of America.
  4. Department of Optometry and Vision Science, Univ of Melbourne, Parkville, Victoria, Australia

Purpose: To establish the correlation between visual field sensitivity and ganglion cell density within the central 20 degrees.  We hypothesized that the use of a test stimulus within complete spatial summation (Goldmann II, GII) would display improved correlation compared to the standard GIII test stimulus.

Methods:One eye of 40 normal subjectswas included in this study. The Humphrey Field Analyser (HFA) was used in full-threshold mode for the 10-2 test grid and 12 points from the 30-2 grid that matched the outer Spectralis grid. Spectralis OCT posterior pole scans for each subject was extracted and the average ganglion cell layer (GCL) thickness values were obtained for each of the 64 grid location within the measurement area ~6880µmx6880µm.  HFA sensitivity in dB was plotted against GC density/mm3(calculated from GCL thickness and GC density from histological data, also converted into dB). Both visual field and OCT data were converted to a 50 year-old equivalent for analysis. The Drasdo et al VR 2007 correction was applied to visual field data to allow comparison of structure and function (Fig. 1). Linear regression analysis was conducted at each test location using individual data or grouped data derived using the 5, 6, 7 and 8 GC iso-density theme classes of Yoshioka et al IOVS 2017 (Fig. 1). A non-parametric bootstrap was used to determine the 99% distribution limits of the slope and correlation parameters.

Results: Table 1 shows the structure-function correlation slope parameters and coefficients of determination (R2) for point-wise and theme class-based comparisons, using GII and GIII. The use of 5 or 6 theme classes resulted in a slope close to unity and high R2values for GII. Table 2 shows the 99% distribution of the slope parameters and R2values for point wise comparisons and those using 5 theme classes again demonstrating superior correlations for GII (both slope and R2 significantly different p<0.01 compared to pointwise analysis). Correcting the data for test size difference (6dB) did not result in data superposition confirming that GIII test size is not within complete spatial summation within the central 20 degrees.

Conclusions:Using a test stimulus within complete spatial summation (GII) and grouping sensitivities according to GC density test grids derived using pattern recognition (7 or fewer GC theme classes), revealed correlations close to unity with coefficients of determination (R2) >0.90. The high correlations achieved when using theme classes even when using individual datasets, suggests that an approach would provide a useful method to predict alterations of visual field sensitivity from OCT data.

Commercial Relationships Disclosure:MK and SKK commercial Relationship(s):2014/094035 A1 (USA) and 13865419.9 (EU):Code P (Patent): REM, JT, BZ, L N-S, BJ, RF: none

Grant support:  NHMRC 1033224;Guide Dogs NSW/ACT; NIH EY02576, EY015128, EY014800, an Unrestricted Grant to the Moran Eye Center from Research to Prevent Blindness.

Leave a Reply

Your email address will not be published. Required fields are marked *