Abstract
Purpose: :
Diabetic retinopathy is the leading cause of blindness in the working age population around the world. Computer assisted analysis has the potential to assist in the early detection of diabetes by efficient screening of large patient populations. We present the deployment of a telemedicine network based in primary care clinics in the Memphis, TN area and in rural Mississippi. This network addresses an underserved population and represents a valuable asset to broad-based screening of diabetic retinopathy and other diseases of the retina. A secure web-based protocol for submission of images and a database archiving system has been developed with a physician reviewing tool. We will focus on previously reported image analysis methods that have been newly implemented in our telemedical network to automatically estimate image quality and locate physiological features in the retina.
Methods: :
All images are acquired from non-dilated retinal images obtained in primary care clinics and are manually reviewed by an ophthalmologist. Images undergo an automatic quality estimation. Previously developed image analysis methods for fast quality estimation, characterization of the vascular tree, and optic nerve / macula location were implemented and operate in parallel to the image storage and manual review procedure. (CBIR) .Image and that are represented in the extensive, previously diagnosed patient archive.
Results: :
Previous results by our methods on sample data sets resulted in correctly classification of 100% of good quality images, 83% of fair quality images, and 0% poor; and optic nerve localization in excess of 90% accuracy. For this work we describe the network and present results of image analysis for a serviced population of 200 patients. We present the performance of the automated image analysis procedures on this population.
Conclusions: :
Our automated analysis methods are a key first step towards full computer-assisted diagnosis of retina diseases. Our future work includes the expansion of our analysis to lesion detection and characterization using our CBIR approach.
Keywords: diabetic retinopathy • image processing