Automated grading and diagnosis system for evaluation of dry eye disease

Today, Dry Eye Syndrome (DES) is a widely seen health problem that is a disorder of the tear film due to tear deficiency. The number of dry punctate dots occurred on corneal surface can be used as a diagnostic indicator of DES severity. Grading of DES severity exactly by counting these dots is a really difficult task for human eye. Taking into account that current methods are also subjectively dependent on the perception of the ophtalmologists in addition to spending time and resource intensively, the enhancement of diagnosis techniques would significantly contribute to clinical DES analysis. Computer-aided diagnosis systems can potentially provide more objective and reliable diagnostic results for health care systems. Moreover, they would also benefit as remote diagnosis systems in places where there may not be well-trained ophtalmologists and modern testing techniques. A computerized diagnosis system by utilizing image processing techniques can be developed and used as an automated grading system in clinical decision making, also speed up evaluation, diagnosis and treatment processes. The work presented in this thesis consists of developing an automated grading system in order to provide more reliable and accurate diagnosis for DES. For determining the recognition performance of the system, an orginal clinical database has been formed during a year in eye clinic. A computerized diagnosis system is performed at the region of interest (ROI) level by applying computational methods on the fluorescein-stained corneal images. These images were gathered via slit lamp photography after sodium fluorescein staining and labeled based on the clinical Oxford Grading Schema (OGS) traditionally implemented by ophtalmologists that uses a 0-none to 5-severe grading scale. This study shows that automatic DES diagnostic kits can be developed by implementing computational methods on the fluorescein-stained cornea images to assist investigators for a more objective and faster DES diagnosis in real life.