Structure-function mapping: variability and conviction in tracing retinal nerve fiber bundles and comparison to a...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Retinal nerve fiber bundles (RNFBs) are crucial for visual function.
• Accurate mapping of RNFBs to visual field defects is essential for diagnosing and monitoring optic neuropathies.
• Current methods for tracing RNFBs can be subjective and variable.

Purpose of the Study:

• To evaluate the variability and repeatability of human tracing of RNFBs.
• To compare human tracing of RNFBs with predictions from an anatomically-customized computational model.
• To assess the utility of a computational model for structure-function mapping in ophthalmology.

Main Methods:

• Eight clinicians and six naïve observers traced RNFBs from 24 visual field locations to the optic nerve head (ONH) in ten retinal images.
• Repeatability of tracing was assessed by three clinicians and two naïve observers over multiple sessions.
• A computational model predicted ONH sectors corresponding to visual field locations, and its predictions were compared to human tracings.

Main Results:

• Median between-observer variability in tracing was 27° for clinicians and 33° for naïve observers.
• Repeatability of tracing was better for clinicians (10°) than naïve observers (15°).
• The computational model showed good concordance with human tracings, with median absolute differences of 17° for clinicians and 20° for naïve observers.

Conclusions:

• The computational model demonstrates good agreement with human RNFB tracing.
• RNFB tracing variability is a critical factor to consider when performing structure-function mapping.
• The model shows potential for improving the accuracy and consistency of structure-function analysis in ophthalmology.