Multivariate Longitudinal Modeling of Macular Ganglion Cell Complex: Spatiotemporal Correlations and Patterns of...

Area of Science:

• Ophthalmology
• Medical Imaging
• Biostatistics

Background:

• Glaucoma is a progressive optic neuropathy characterized by retinal ganglion cell loss.
• Ganglion Cell Complex (GCC) thickness measurements are crucial for monitoring glaucoma progression.
• Understanding spatiotemporal patterns of GCC thickness change is vital for accurate diagnosis and treatment.

Purpose of the Study:

• To investigate spatiotemporal correlations among GCC superpixel thickness measurements.
• To explore underlying patterns of longitudinal change across the macular region in glaucoma patients.
• To refine a Bayesian hierarchical model for analyzing these changes.

Main Methods:

• Developed and applied a Bayesian hierarchical model for longitudinal GCC thickness analysis.
• Incorporated global priors for macular superpixel parameters.
• Utilized Bayesian residual analysis and Principal Component Analysis (PCA) to assess correlations and patterns.

Main Results:

• The Bayesian model demonstrated significant correlations among nearest-neighbor superpixels, particularly in the superior macula.
• PCA revealed distinct patterns of longitudinal change, including a global intercept component and regional contrasts (superior/inferior, inner/nasal vs. temporal/peripheral).
• Correlations were strongest for random intercepts, followed by slopes and residuals, with similar patterns observed across different model components.

Conclusions:

• The developed Bayesian model enhances the estimation of population and subject parameters for longitudinal GCC thickness data.
• Incorporating cross-superpixel random effects and correlations can further improve the analysis of spatiotemporal relationships in glaucoma progression.
• This approach offers a more nuanced understanding of macular changes in glaucoma, aiding in disease management.