Developing spatial models for early glaucoma detection using retinal ganglion cell point patterns

Although earlier studies applied non-parametric methods to identify differences between healthy and diseased tissue, these approaches struggle with the complexity of the data, which includes biological noise, variability across samples, and differing cellular interactions.<br><br>To address these challenges, this work develops point pattern analysis methods that incorporate fractal measures, which are more robust to noise, and second-order statistics to capture subtle structural changes in cell distribution.

A model is constructed using approximate Bayesian computation (ABC), integrating aspects of the log-Gaussian Cox and Strauss processes. It is extended to include marks, spatial covariates, and replicate structures, and is used alongside change point detection to identify spatial deviations from healthy patterns.<br><br>The findings reveal retinal regions that may be more susceptible to RGC loss, offering insights into disease progression and enabling more targeted diagnostic focus.

This approach advances spatial modelling methods and provides a promising framework for early glaucoma detection, with potential clinical applications for improving diagnosis and treatment outcomes.<br>