Abstract
Purpose: :
Detection and analysis of changes from retinal images is important in clinical practice, computer–assisted reading centers, and in medical research. In these applications, it is desirable to develop automatic algorithms to detect and describe the changes. Algorithms have been developed to analyze changes from longitudinal time series of color fundus image, fluorescein angiograms and dual–wavelength retinal fundus images.
Methods: :
The described methods are robust to: (i) spatial variations in illumination resulting from changes both within, and between patient visits; (ii) imaging artifacts such as dust particles; (iii) outliers in the training data; and (iv) segmentation and alignment errors. Robustness to illumination variation is achieved by a novel iterative algorithm to estimate the reflectance of the retina exploiting automatically extracted segmentations of the retinal features. Robustness to dust artifacts is achieved by exploiting their spectral characteristics, enabling application to film–based, as well as digital imaging systems. False changes from alignment errors are minimized by sub–pixel accuracy registration.
Results: :
For the structural changes in the non–vascular regions, a multi–observer validation on eyes involving NPDR and PDR indicated a 97% change detection rate and 99.3% classification accuracy. For vascular changes, the algorithms had 82% detection rate. For analysis of structural changes from fluorescein angiograms, the algorithms had a detection rate of 83%. For the functional changes, the sensitivity for detecting saturation change when breathing air vs. pure oxygen was found to be 0.0226 Optical Density Ratio (ODR) units, which is in good agreement with previous manual measurements.
Conclusions: :
The importance of automated retinal image analysis procedures in general, and change analysis in particular are due to the fact that most retina–related clinical diagnostic and treatment procedures are largely image driven. This work has resulted in the development of software tools for higher–level, quantitative, and highly–automated retinal image analysis, with a focus on change analysis. Reliable, illumination–invariant, and fully–automated detection and analysis of changes in retinal images can form a valuable additional diagnostic resource for the clinician and researcher by mapping the dynamic nature of diseases.
Keywords: imaging/image analysis: clinical • diabetic retinopathy • image processing