Peer Reviewed

1

DOI

10.17077/omia.1018

Conference Location

Boston, MA, USA

Publication Date

September 2014

Abstract

Age related macular degeneration is a major cause of blindness and visual impairment in older adults. Its exudative form, where fluids leak into the macula, is especially damaging. The standard treatment involves injections of anti-VEGF (vascular endothelial growth factor) agents into the eye, which prevent further vascular growth and leakage, and can restore vision. These intravitreal injections have a risk of devastating complications including blindness from infection and are expensive. Optimizing the interval between injections in a patient specific manner is of great interest, as the retinal response is partially patient specific. In this paper we propose a machine learning approach to predict the retinal response at the end of a standardized 12-week induction phase of the treatment. From a longitudinal series of optical coherence tomography (OCT) images, a number of quantitative measurements are extracted, describing the underlying retinal structure and pathology and its response to initial treatment. After initial feature selection, the selected set of features is used to predict the treatment response status at the end of the induction phase using the support vector machine classifier. On a population of 30 patients, leave-one-out cross-validation showed the classification success rate of 87% of predicting whether the subject will show a response to the treatment at the next visit. The proposed methodology is a promising step towards the much needed image-guided prediction of patient-specific treatment response.

Rights

Copyright © 2014, Hrvoje Bogunovic, Michael D. Abramoff, Li Zhang, and Milan Sonka.

Included in

Ophthalmology Commons

Share

COinS
 
Sep 14th, 12:00 AM Sep 14th, 12:00 AM

Prediction of treatment response from retinal OCT in patients with exudative age-related macular degeneration

Boston, MA, USA

Age related macular degeneration is a major cause of blindness and visual impairment in older adults. Its exudative form, where fluids leak into the macula, is especially damaging. The standard treatment involves injections of anti-VEGF (vascular endothelial growth factor) agents into the eye, which prevent further vascular growth and leakage, and can restore vision. These intravitreal injections have a risk of devastating complications including blindness from infection and are expensive. Optimizing the interval between injections in a patient specific manner is of great interest, as the retinal response is partially patient specific. In this paper we propose a machine learning approach to predict the retinal response at the end of a standardized 12-week induction phase of the treatment. From a longitudinal series of optical coherence tomography (OCT) images, a number of quantitative measurements are extracted, describing the underlying retinal structure and pathology and its response to initial treatment. After initial feature selection, the selected set of features is used to predict the treatment response status at the end of the induction phase using the support vector machine classifier. On a population of 30 patients, leave-one-out cross-validation showed the classification success rate of 87% of predicting whether the subject will show a response to the treatment at the next visit. The proposed methodology is a promising step towards the much needed image-guided prediction of patient-specific treatment response.