# Wavelet-Based Pattern ERG Biomarkers Outperform Temporal Amplitude Measures for Functional Stratification in Optic Nerve Disease

## TL;DR

This study shows that wavelet-based biomarkers from pattern electroretinography better distinguish optic nerve disease from healthy eyes than traditional measures.

## Contribution

Symlet-2 wavelet features outperform canonical amplitude measures in detecting retinal ganglion cell and macular cone dysfunction in optic nerve disease.

## Key findings

- Sym2–A6-4 wavelet feature showed better group separation between healthy and optic nerve disease subjects than traditional amplitude measures.
- Sym2–D6-2 wavelet coefficient strongly correlated with macular cone-specific amplitudes in both healthy and diseased groups.
- Previously defined Daubechies 8-based energy index failed to provide effective biomarkers for optic nerve disease.

## Abstract

To extend wavelet analysis of pattern electroretinography (PERG) from macular cone to retinal ganglion cell (RGC) dysfunction in optic nerve disease (OND) by validating Symlet-2 (sym2) discrete wavelet transform (DWT) features.

From the open access PERG–Institute of Applied Ophthalmobiology (IOBA) dataset, 58 recordings from OND subjects and 262 recordings from healthy volunteers (HVs) were analyzed. Five pre-selected sym2 coefficients (D5-2, D6-2, D6-3, A6-3, A6-4) were quantified. Their correlations with canonical amplitudes (|P50–N35|, |N95–P50|) and group separation (rank-biserial effect size, |rrb|) were analyzed. We also assessed a previously defined DWT energy index based on the Daubechies 8 mother wavelet (7N), capturing RGC activity.

The macular cone–specific sym2–D6-2 correlated tightly with |P50–N35| in HVs (rcorr = 0.95) and OND subjects (rcorr = 0.97). In contrast, sym2–A6-4 (112–150 ms, 0–13 Hz) was best suited to capture differences between the HV and OND groups (|rrb| = 0.549), compared to |N95–P50| (|rrb| = 0.358). Bootstrap benchmarking confirmed that sym2–A6-4 outperformed |P50–N35| and |N95–P50| (Δ|rrb| = 0.362 and 0.187; Pboot = 0.005 and 0.036, respectively). The 7N feature failed to yield effective results on all measures (|rrb| = 0.084).

Sym2 DWT features provide compartment-specific, multidimensional biomarkers that outperform traditional canonical peaks for both macular cone (sym2–D6-2) and RGC (sym2–A6-4) assessment. Future work should validate these biomarkers in a large, diverse, genetically and phenotypically characterized external cohort to confirm generalizability and clinical utility.

Sym2 wavelet indices provide robust and sensitive PERG biomarkers that could serve as quantitative endpoints in clinical trials.