Abstract
Purpose :
This Prospective, observational study aimed to to develop a Computer-Aided Detection/Diagnosis (CAD) system that enables an automatic detection and diagnosis of early microvasculature changes in the retina of diabetic patients based on Optical Coherence Tomography Angiography (OCTA) scans.
Methods :
This was a prospective, observational study that included 50 diabetic patients without clinical evidence of diabetic retinopathy and 50 age-matched controls. OCTA was performed using Cirrus HD-OCT 5000 Angioplex (Carl Zeiss Meditech, CA. USA).Patients underwent 3x3 mm and 6x6 mm macular scans that were captured at ~840nm wavelength and 68,000 A-scans/second, and the Split-spectrum amplitude decorrelation angiography algorithm were utilized. A mathematical CAD software was developed to automatically detect retinal microvasculature on OCTA images, including superficial and deep retinal cuts. The developed software consisted of three main stages. First, it reduces the noise and improves the contrast of the OCTA images. Second, a novel retinal segmentation technique is performed to separate the vessels from the other tissues. Finally, a complete retinal microvasculature classification and analysis was performed by applying machine learning techniques.
Results :
CAD system was implemented and tested on normal and diseased patients. The performance of the proposed system was evaluated by using some standard performance measures, such as accuracy of classification, specificity, and sensitivity. This system outperformed other systems that are based on fundus scans. The proposed system detected early development of microaneurysms that were not apparent by other investigation modalities as well as areas of capillary loss. Alterations in the vascular structure were noted as well as increased vessel and capillary tortuosity.Enlargement of the foveal avascular zone appeared to be one of the earliest changes the system detected in diabetic patients. The automatic software was able to accurately detect areas of non-perfusion and decreased vessel density.
Conclusions :
An automatic software was developed for the detection of early changes in the microvasculature of diabetic patients using OCTA images. It carries the promise of better monitoring of diabetic patients to allow prevention of progression towards advanced diabetic retinopathy.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.