Investigating the Utility of Fractal Measures as a Diagnostic for Glaucoma

Recent advances in Adaptive-Optics Confocal Scanning Laser Ophthalmoscopy (AO-cSLO) have enabled non-invasive, in-vivo imaging of retinal cell subsets, at near single-cell resolution. However, challenges such as spatial and temporal variability, limited observation windows, and segmentation errors hinder the effective use of this data for diagnosing neurodegenerative disorders like glaucoma. Common point pattern metrics used in this area, such as density and distance-based measures, are sensitive to noise and incomplete data, reducing their diagnostic utility. Fractal measures, with their scale-invariance and robustness to point pattern degradation, offer a promising alternative. In this study, we apply fractal analysis to RGC point pattern data sets from rodent glaucoma models and controls, comparing fractal dimension, lacunarity, and succolarity with traditional metrics. We find that lacunarity and succolarity provide more accurate and enable earlier detection of glaucoma than traditional metrics, demonstrating greater robustness to noisy and degraded data. These measures demonstrate significant potential for achieving an earlier and more reliable glaucoma diagnosis.