Using texture based features from the continuous wavelet transform of the electroretinogram to predict glaucoma.

Glaucoma is an optic neuropathy resulting in the progressive loss of retinal ganglion cells (RCGs). The photopic negative response (PhNR) of the electroretinogram (ERG) has been used to objectively measure RCG function. This study sought to explore whether the usage of textural features extracted from the continuous wavelet transform of the ERG combined with ERG amplitude markers were more effective at predicting glaucoma severity than using the ERG markers alone. One-hundred and three eyes of 55 participants were included in this study, who underwent ERG testing with a protocol targeted at the PhNR. Predictive models for glaucoma severity based on the estimated RGC count were fitted using multiadaptive regression splines (MARS). The models informed by a combination of amplitude markers and texture analysis had a better predictive performance; R2 = 0.492, compared to models informed by markers alone having an R2 = 0.349 (p = 0.009).Clinical Relevance- As a direct measure of retinal function, the ERG has potential to determine the health of RGCs. This study demonstrates there is additional data within the ERG available to clinicians, which has the potential to improve the diagnosis and management of glaucoma.