DIABETIC RETINOPATHY DETECTION USING CV
Budget: ₹1,500 – ₹12,500 INR
bjectives :The main objective of this method is to detect DR (Diabetic Retinopathy) eye disease using Image
Processing techniques. The tool used in this method is Open CV and it is widely used in image processing. It extracts the Blood Vessels from the medical image of human eye-retinal fundus image which is used for detecting DR. This method utilizes an approach of Adaptive Histogram Equalization using CLAHE (Contrast Limited Adaptive Histogram Equalization) algorithm with open CV (Computer Vision) framework implementation. Contrast-Limited Adaptive Histogram Equalization (CLAHE) filter was proven to be effective in improving the micro-aneurysms pixels. CLAHE filter manages to produce adequate enhancement in veins. The expected result shows that affected DR is detected in fundus image and the DR is not detected in the healthy fundus image . Dataset has 3200 eye images and their corresponding severity scale
No DR - 0
Mild -1
Moderate - 2 Severe - 3 Proliferative DR - 4
Brief Methodology :
Automating the process using Convolutional neural networks to speed up blindness detection in patients before its too late. Eye retina images contain artifacts, some of them are out of focus, underexposed, or overexposed etc. Also, some of the images have low brightness and low lightning conditions thus making it difficult to assess the difference between the images. First we do image preprocessing by applying Gaussian blur to bring out distinctive features in the images. In Gaussian Blur operation, the image is convolved with a Gaussian filter which is a low-pass filter that removes the high-frequency components . Then convert the RGB scale images to Gray scale images and circular cropping from gray scale images is used. After applying some image augmentation, model architecture is used. Mostly I am planning to use VGG-16.
Processing techniques. The tool used in this method is Open CV and it is widely used in image processing. It extracts the Blood Vessels from the medical image of human eye-retinal fundus image which is used for detecting DR. This method utilizes an approach of Adaptive Histogram Equalization using CLAHE (Contrast Limited Adaptive Histogram Equalization) algorithm with open CV (Computer Vision) framework implementation. Contrast-Limited Adaptive Histogram Equalization (CLAHE) filter was proven to be effective in improving the micro-aneurysms pixels. CLAHE filter manages to produce adequate enhancement in veins. The expected result shows that affected DR is detected in fundus image and the DR is not detected in the healthy fundus image . Dataset has 3200 eye images and their corresponding severity scale
No DR - 0
Mild -1
Moderate - 2 Severe - 3 Proliferative DR - 4
Brief Methodology :
Automating the process using Convolutional neural networks to speed up blindness detection in patients before its too late. Eye retina images contain artifacts, some of them are out of focus, underexposed, or overexposed etc. Also, some of the images have low brightness and low lightning conditions thus making it difficult to assess the difference between the images. First we do image preprocessing by applying Gaussian blur to bring out distinctive features in the images. In Gaussian Blur operation, the image is convolved with a Gaussian filter which is a low-pass filter that removes the high-frequency components . Then convert the RGB scale images to Gray scale images and circular cropping from gray scale images is used. After applying some image augmentation, model architecture is used. Mostly I am planning to use VGG-16.