Early Detection of Diabetic Retinopathy (DR) From Fundus Images

Project Title

Early Detection of Diabetic Retinopathy (DR) From Fundus Images

Project Area of Specialization Biomedical EngineeringProject Summary

Diabetic Retinopathy (DR) is complication of diabetes that is caused due to high sugar level. In DR, the formation of retinal blood vessels. There are two types of changes in retinal blood vessels. In first type, the blood vessels walls become narrow and occlusion occur between them. The narrowing of blood vessels cause changes in retina. In second type, the blood vessels become feeble and start leaking. The leaking blood vessels causing thickening of retina and blurring a vision or new blood vessels grow thick and bleed which leads to blindness. This disease is the second largest disease which is occurring amongst the human beings as per the WHO – United Nations survey. DR patient suffers from blur vision, floaters, fluctuating vision, and distorted vision, dark areas in vision, poor night vision, impaired color vision and total loss of vision. DR symptoms doesn’t show at early stage so people affected doesn’t notice the changes in their vision or in structure of retina. People consult ophthalmologist when their condition become worsen and it became difficult to treat and most of people loss their vision. So there is no early detection of DR which leads to blindness and massive workload on ophthalmologist. The timely detection of DR will guard people from vision loss. Medical imaging technology become major tool for detection and examination of various diseases. These technologies are based on deep learning or machine learning techniques and image processing techniques. In this research, we use these techniques for early detection of DR using blood vessels to prevent from vision loss and also help Ophthalmologist for screening of DR.

Project Objectives

The timely detection of DR will guard people from vision loss. Medical imaging technology become major tool for detection and examination of various diseases. These technologies are based on deep learning or machine learning techniques and image processing techniques. In this research, we build an automated screening system which is based on these techniques for early detection of DR using blood vessels to prevent from vision loss and also help ophthalmologist for screening of DR.

Project Implementation Method Methodology:

The proposed methodology is consist following of steps:

Image Acquisition
Preprocessing
Blood Vessel Extraction
Post Processing
Diabetic Retinopathy Detection

Image Acquistion:

In today’s research on medical fundus images, there are widely accessible resources are available. Some Standard benchmark databases of fundus images are openly available for computerized screening and analysis of diabetic retinopathy. The openly available databases used in this research are DRIVE, STARE and CHASE.

Preprocessing:

Preprocessing is a step that will eliminate errors caused during taking of the image and to reduce brightness effects on the image. For this step, some filters and methods will be used to remove noise and errors from fundus images. Some filter and methods that can be used for this task are mean filter for smoothing and reducing noise, median filter to remove salt and pepper noise, weiner filter to remove noise and blurring effects, order statistic filtering to remove salt and pepper and gaussian noise, dividing method for illumination correction, homomorphic filtering (HSI) to decouples intensity, color information and illumination correction techniques based on a statistical evaluation, CLAHE to enhance local contrast and polynomial grey level transformation for contrast enhancement. 

Blood Vessel Extraction:

To extract blood vessels segmentation will be performed. Segmentation is process of image processing which is used to locate objects and boundaries in images. In this research, segmentation will performed to extract blood vessel and remove unrequired data like optical disc, macula etc.

Post Processing:

In this step, the output images from blood vessel extraction will be processed to clear contours of the vessels. Some filters that can be used for this purpose are spatial filter for sharpening, Ypmean filter for FFT smoothing, length filter to remove falsely detected isolated pixels regions and morphological operations to remove imperfections in the structure of image.

Benefits of the Project

The significance of this research is to detect and segment the retinal blood vessels for DR detection using automated screening system. DR early detection and ophthalmologist treatment can significantly reduce the risk of vision loss in diabetic patients. This project is an attempt towards finding an automated way to detect DR in its early phase.

Technical Details of Final Deliverable

A software-hardware integrated system that will perform early detection of DR with a high degree of accuracy.

Final Deliverable of the Project HW/SW integrated systemCore Industry MedicalOther Industries Education Core Technology OthersOther TechnologiesSustainable Development Goals Good Health and Well-Being for PeopleRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	70000
GPU Server	Equipment	1	30000	30000
Fundus Camera	Equipment	1	40000	40000