Study On The Method Of Microaneurysms Detection In Fundus Image

Diabetes is a chronic disease and a serious threat to human health.The disease would bring many complications.Among them,diabetic retinopathy(DR)is one of the most common and serious complications of diabetes.DR is an important factor of blindness.Retinal fundus images is an important basis for judging the eye disease.But relying on the doctor working in the Department of Ophthalmology observing the fundus images through the naked eye,there are too many subjective factors and the efficiency is low.Therefore,with the help of computer,if we can identify the DR lesions in fundus images quickly and reliably,we not only reduce the influence of subjective factors,but also improve the reading efficiency of doctors.In the retinal fundus images,microaneurysms are an early mark of DR.If we can detect microaneurysms accurately,we can effectively inhibit and evaluation the development of disease.Therefore,accurate microaneurysms detection is very important for the early diagnosis and treatment of DR.Research on the method of microaneurysms detection in this paper,we obtain the following achievements:(1)Before the detection of microaneurysms in the fundus images,the first step is preprocessing.In the preprocessing step,the method to extract the fundus image histogram differentiation and morphological erosion operation based on the field.Provide field mask image for the microaneurysms detection in later chapters.Through the brightness balance and contrast gray level grouping enhancement method,we can remove non-uniform illumination phenomenon and enhance the fundus lesions and background contrast in the fundus image.In order to make sure microaneurysms detection can be accurate in later chapters.(2)Because of the optic disc area or the optic disc border there may be some dark spots,they could interfere the detection of microaneurysms.So we must remove the optic disc area before we do microaneurysm detection.Based on obtain the image features in the practical application,using a detection method combining brightness and circularity of the optic disc.The algorithm does not need to the retina blood vessel segmentation,reduces the complexity of the algorithm and get a better balance in the accuracy and the real-time performance.(3)In the aspect of microaneurysms detection,microaneurysms detection is influenced because of tiny blood vessels,vascular cross or the background texture and other factors.A method combining Canny edge detection based on histogram analysis concavity,and feature detection method for microaneurysms detection.For the fundus images,the first step is preprocessing.Then remove the edge of the field and optic disc area.Then through the edge detection and region filling to obtain candidate microaneurysms.Finally,according to the area and shape characteristics of microaneurysms detected real microaneurysms.Verified by the open HEI-MED data sets,this method can get a high evaluation index.(4)The method of microaneurysms detection is applied to the remote DR screening system.Through the microaneurysms detection extract microaneurysms point automatically.And through the artificial interpretation to adjust some false or missing microaneurysms.So make sure to detect the accurate microaneurysms and improve the interpretation efficiency.