The pseudotemporal bootstrap for predicting glaucoma from cross-sectional visual field data.

Area of Science:

• Ophthalmology
• Data Science
• Biomedical Engineering

Background:

• Glaucoma causes irreversible blindness due to progressive visual field loss.
• Longitudinal data for studying disease progression is scarce and expensive.
• Existing datasets often lack the temporal dimension crucial for understanding disease dynamics.

Purpose of the Study:

• To develop a method for creating artificial time series (pseudo time series) from cross-sectional data.
• To enable temporal modeling and forecasting of visual field progression in eye diseases like glaucoma.
• To address the challenge of limited longitudinal data in disease progression studies.

Main Methods:

• Developed a pseudo time series approach to reconstruct temporal trajectories from cross-sectional patient data.
• Utilized resampling techniques to generate multiple disease trajectories, handling outliers and varied disease paths.
• Applied the method to synthetic data for property demonstration and to visual field data for glaucoma prediction.

Main Results:

• Successfully generated pseudo time series from cross-sectional data.
• Demonstrated the method's ability to capture key disease trajectories.
• Achieved promising results in predicting glaucoma using visual field data.

Conclusions:

• Pseudo time series offer a viable solution for temporal modeling with cross-sectional data.
• The approach facilitates forecasting of disease progression, particularly for conditions like glaucoma.
• This method enhances the utility of existing datasets for understanding and predicting visual deterioration.