Purchase this article with an account.
Owhofasa O. Agbedia, Mohamed Ibrahim, Yasir J. Sepah, Millena Bittencourt, Eman Magdy, Raafay Sophie, Anam Akhlaq, Hong T. Liu, Ahmed S. Fahmy, Quan D. Nguyen; Quantitative Assessment of Diabetic Macular Edema using Digital Subtraction of Early and Late Acquisitions of Fluorescein Angiograms. Invest. Ophthalmol. Vis. Sci. 2012;53(14):3111.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Traditional method for visualizing and evaluating diabetic macular edema (DME) has been fluorescein angiography (FA) where assessment is often based on subjective interpretation. The advancement of optical coherence tomography (OCT) has increased the capacity to provide detailed visualization of the retina. A novel method using FA that enables automatic segmentation and quantitative evaluation of leakage in DME is described. We evaluate the correlation between the segmentation of DME on FA images employing a novel method that we have developed and images obtained by spectral-domain OCT (SD OCT) in patients with DME.
FA images (early and late time frames) and images of central retina obtained by SD OCT in 20 patients (20 eyes) with DME were analyzed. Images were obtained at the baseline visit of participants currently enrolled in a clinical trial for DME. We modeled the macular image in the early time frame using 2D Gaussian surfaces to correct irregular hyper-fluorescent areas. The model was subtracted from the late time frame image of the macula to enhance DME areas. Intensity values of the resulting Difference Image (DI) were mapped to predicted retinal thickness using a linear transformation that maps minimum and maximum gray levels to 250 and 600 microns respectively. Correlation between predicted thickness and retinal thickness as measured by SD OCT was calculated. FA images obtained with 30 and 35-degree field of view were utilized.
For the 9 subfields combined, a mean correlation coefficient of 0.77 (SD: 0.10) was observed between retinal thickness as measured by SD OCT and thickness generated by the proposed method. The correlation between individual subfields in the standardized grid across all patients is summarized in Figure 1.
In our novel work, we have proposed a digital subtraction method to quantify DME leakage in FA sequences. Our results have shown that valuable quantitative clinical information such as active retinal vascular leakage may be extracted from the DI, which correlates strongly with the retinal thickness generated by the SD OCT.
This PDF is available to Subscribers Only