Peer Reviewed

1

DOI

10.17077/omia.1045

Conference Location

Athens, Greece

Publication Date

October 2016

Abstract

Age-related macular degeneration (AMD) is a leading cause of blindness in developed countries. The presence of drusen is the hallmark of early/intermediate AMD, and their sudden regression is strongly associated with the onset of late AMD. In this work we propose a predictive model of drusen regression using optical coherence tomography (OCT) based features. First, a series of automated image analysis steps are applied to segment and characterize individual drusen and their development. Second, from a set of quantitative features, a random forest classifiser is employed to predict the occurrence of individual drusen regression within the following 12 months. The predictive model is trained and evaluated on a longitudinal OCT dataset of 44 eyes from 26 patients using leave-one-patient-out cross-validation. The model achieved an area under the ROC curve of 0.81, with a sensitivity of 0.74 and a specificity of 0.73. The presence of hyperreflective foci and mean drusen signal intensity were found to be the two most important features for the prediction. This preliminary study shows that predicting drusen regression is feasible and is a promising step toward identification of imaging biomarkers of incoming regression.

Rights

Copyright © 2016 the authors

Included in

Ophthalmology Commons

Share

COinS
 
Oct 21st, 12:00 AM Oct 21st, 12:00 AM

Predicting Drusen Regression from OCT in Patients with Age-Related Macular Degeneration

Athens, Greece

Age-related macular degeneration (AMD) is a leading cause of blindness in developed countries. The presence of drusen is the hallmark of early/intermediate AMD, and their sudden regression is strongly associated with the onset of late AMD. In this work we propose a predictive model of drusen regression using optical coherence tomography (OCT) based features. First, a series of automated image analysis steps are applied to segment and characterize individual drusen and their development. Second, from a set of quantitative features, a random forest classifiser is employed to predict the occurrence of individual drusen regression within the following 12 months. The predictive model is trained and evaluated on a longitudinal OCT dataset of 44 eyes from 26 patients using leave-one-patient-out cross-validation. The model achieved an area under the ROC curve of 0.81, with a sensitivity of 0.74 and a specificity of 0.73. The presence of hyperreflective foci and mean drusen signal intensity were found to be the two most important features for the prediction. This preliminary study shows that predicting drusen regression is feasible and is a promising step toward identification of imaging biomarkers of incoming regression.