Development of an Automated Image Analysis Technique for Assessment of Macular Fluid Volume from High Resolution OCT Images of the Macula

Abstract

 

Purpose:

 

To develop an image processing algorithm to analyse optical coherence tomography (OCT) scans to determine volume of fluid leak in patients with macular oedema.

 

Methods:

 

High resolution OCT images of patients with macular oedema were assessed by a retinal specialist and areas of low reflectivity determined to be fluid were quantified and investigated. To segment these abnormal areas of low reflectivity the OCT scans are first processed to separate out normal hyporeflective inner and outer nuclear layer using mathematically defined texture-based analysis. These processed frames are subsequently used as masks in detection of the regions of interest (ROI) that contain the abnormally hyporeflective (dark) areas in the remainder of the image. A sequence of logical and morphological operations are performed on the OCT scans and boundaries around the suspected lesion areas are detected. Thus we developed algorithms to segment out areas thought to correspond to fluid in each longitudinal A scan. We calculated the total area of abnormally low reflectivity ‘fluid’ in each of 128 scans taken for each patient and calculated total macular fluid volume by integrating these areas into a total volume score.

 

Results:

 

A single segmented frame is shown below. Segmented areas are outlined in purple on the OCT image. We plan to improve accuracy by introducing region- based thresholding that will essentially vary the algorithms for detection of abnormality depending on exactly what depth in the retina any region is being assessed.

 

Conclusions:

 

The current algorithms need further refinement and programming, and assessment for reliability and validity before clinical use. Although further work is required, the current algorithms demonstrate the feasibility of using image processing algorithms to calculate macular fluid volume that would be of use in clinical assessment and particularly as objective measures in clinical trials.  

 

Keywords: macula/fovea • age-related macular degeneration • imaging/image analysis: clinical