Detecting glaucoma in biomedical data using image processing
Physical copy available from RIT's Wallace Library at RE871 .B42 2005
This thesis addresses the problem of the early detection of an eye blinding disease, glaucoma. It presents new approaches for analysis of the biomedical scan data of Retinal Nerve Fiber Layer (RNFL) thickness obtained through Scanning Laser Polarimetry that can lead to better tools for early diagnoses of glaucoma. The thickness maps of the RNFL obtained from a Scanning Laser Polarimeter (Gdx-VCC) were used to draw features as opposed to the circular ring one-dimensional data (TSNIT graph) in previous approaches. Fourier analysis and wavelet analysis were performed on the 90° projections of the thickness map data to emphasize the shape contained in the RNFL around the optic disc. Another approach was to analyze the shape of the entire 2 dimensional thickness maps through 2D Fourier Transform. A pattern image based on the shapes observed in the scans was generated and used to draw features. Principal Component Analysis was performed on the combined feature set for dimension reduction of feature space. Finally Fisher's linear discriminant function (LDF) was used as a classifier. A Receiver Operating Characteristic (ROC) curve analysis of the developed parameters has been performed for all the feature sets used and has been compared with one of the currently used technique of Fourier analysis of TSNIT graph data available from similar eye scan images. The analysis tools implemented and used for the classification gives comparable results with the existing techniques and hence offer an effective tool for enhancing diagnostic abilities and can add to the sensitivity of the existing techniques to improve performance.