Spatio-Temporal Signatures to Predict Retinal Disease Recurrence.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Longitudinal spectral domain optical coherence tomography (SD-OCT) provides rich data for understanding disease progression.
• Predicting treatment response and disease recurrence remains a challenge in managing retinal conditions.

Purpose of the Study:

• To develop a method for predicting treatment response patterns and disease recurrence using spatio-temporal signatures from SD-OCT images.
• To accurately forecast the timing and likelihood of future disease recurrence.

Main Methods:

• Extraction of spatio-temporal disease signatures from total retinal thickness maps in a joint reference coordinate system.
• Formulation of prediction using a multi-variate sparse generalized linear model regression on aligned signatures.
• Utilizing a time-to-event based Cox regression model for predicting time to recurrence.

Main Results:

• The model achieved an ROC AUC of 0.99 for predicting recurrence vs. non-recurrence.
• The spatio-temporal survival model predicted time to recurrence with a Mean Absolute Error (MAE) of 1.25 months.
• Identified predictive and interpretable features within the spatio-temporal signature.

Conclusions:

• The proposed method effectively predicts disease recurrence and timing from SD-OCT data.
• Spatio-temporal analysis of retinal morphology offers significant advantages over traditional regression models.
• This approach has the potential to enhance clinical decision-making and patient management for retinal diseases.