Development of an Automated Electroretinography Analysis Approach

Area of Science:

• Ophthalmology
• Neuroscience
• Biomedical Engineering

Background:

• Electroretinography (ERG) is crucial for assessing retinal function in clinical and research settings.
• Manual analysis of ERG waveforms is time-consuming and prone to interobserver variability.
• Objective and repeatable methods are needed to enhance the rigor of ERG analysis.

Purpose of the Study:

• To develop and validate ERGAssist, an automated approach for non-subjective and repeatable feature identification of ERG waveforms.
• To improve the efficiency and consistency of ERG data analysis in ophthalmology and ocular disease research.

Main Methods:

• ERGAssist employs denoising, low-pass filtering, and band-pass filtering to identify key ERG waveform features (a-wave, b-wave, Oscillatory Potentials - OPs).
• The method was validated using two cohorts: Brown Norway rats (Coherence cohort) and control/diabetic Long Evans rats (Verification cohort).
• Stimulus conditions ranged from -6 to 1.9 log cd·s/m² across dark- and light-adapted protocols.

Main Results:

• ERGAssist demonstrated strong correlations with manual markings for amplitudes (r² = 0.92–0.99) and implicit times (r² = 0.90–0.96) in the Coherence dataset.
• In the Verification cohort, ERGAssist successfully differentiated control from diabetic rats, identifying significantly longer OP implicit times in diabetic animals (P < 0.0001).
• The automated approach proved effective in identifying differences in retinal function between healthy and diseased states.

Conclusions:

• ERGAssist provides a reliable and accurate automated method for identifying features of full-field ERG waveforms.
• The tool enhances the rigor of basic science studies investigating retinal function.
• An open-source graphical user interface (GUI) has been developed to facilitate community adoption.