Detection of hemorrhage and exudates in retinal fundus image of diabetic patients
Diabetes is a disease that interferes with the body's ability to use and store sugar, which can cause many health problems. Over time, diabetes affects the circulatory system including the retina. As diabetes progress, the vision of a patient may start to deteriorate and then leading to Diabeti...
| Summary: | Diabetes is a disease that interferes with the body's ability to use and store sugar, which can cause many health problems. Over time, diabetes affects the circulatory system including the retina. As diabetes progress, the vision of a patient may start to deteriorate and then leading to Diabetic Retinopathy (DR) which further will cause blindness. So, early detection of the disease is important to avoid blindness. There are
several ways to diagnose DR and slit – lamp examination is one of the traditional method used by the ophthalmologist. This method requires the clinician to see directly into patient’s eye through an ophthalmoscope or the slit lamp machine to determine whether or not the eyes contain any abnormal features that indicate DR. However, this is not the most effective method yet. Any human can get tired and drowsy including doctors. This
natural flaws of human being can affect the diagnosis and then causing false result analysis. Besides, every individuals doesn’t hold same opinion and judgment. Therefore, this project is proposed to assist the clinicians in identifying DR. There are two main abnormal features that are formed in the retina of a diabetic retinopathy’s patient. They are hemorrhage and exudates. Hemorrhage are formed as a result due to leakage of retinal blood vessel which has similar red colour to the vessel. Whereas exudates are yellow-white deposits structure on the retina that is formed due to leakage of blood from abnormal vessels. This thesis mainly focuses on developing a Fundus Image Analysis (FIA) system that extracts the anatomical and both the abnormal features of the retina in order to diagnose the disease. This research is carried out in three phases. In the first
phase, an automated system is developed to distinguish the anatomical features of the retina from the abnormal features. This phase is called the Masking Phase. This phase involved combinations of several image processing techniques including Specify Polygonal Region of Interest (ROIPOLY), Contrast-limited adaptive histogram equalization (CLAHE), Morphological Opening and Structuring, Median Filtering and Thresholding. The second phase is the Haemorrhage Extraction phase. In this phase, Saturation Adjust Method, Morphological operations and Regional Minima technique is proposed. The third and the last phase is the Exudates Extraction phase. In this phase, Edge Detection, Gradient Magnitude and Region Of Interest techniques are combined to form a complete working algorithm. The experimented images in this project are the retinal fundus images that was taken from a public database (diaretdb1 - Standard Diabetic Retinopathy Database). It is a public database for benchmarking diabetic retinopathy detection from digital images. By using this database and the defined testing
protocol, the results between different methods can be compared. At the end of this project, the result shows that the method applied is able to detect exudates features and capable of detecting and distinguishing hemorrhage from blood vessels. Final result shows the accuracy of 48.3% for detecting images with haemorrhages and 68.5% for images with exudates. |
|---|