DevMesh

Diabetic Retinopathy Detection

Raushan Verma

Raushan Verma

Noida, Uttar Pradesh

0 0
  • 0 Collaborators

Diabetic retinopathy is the leading cause of blindness in the working-age population of the developed world. It is estimated to affect over 93 million people. In this project we create an automated analysis system capable of assigning a score of presence of diabetic retinopathy in each retina image on a scale of 0 to 4. ...learn more

Project status: Concept

Artificial Intelligence

Intel Technologies
AI DevCloud / Xeon, Intel Opt ML/DL Framework

Overview / Usage

Diabetic retinopathy is one the leading causes of blindness in the working-age population of the developed world. It is estimated to affect over 93 million people. Currently, detecting DR is a time-consuming and manual process that requires a trained clinician to examine and evaluate digital color fundus photographs of the retina. By the time human readers submit their reviews, often a day or two later, the delayed results lead to lost follow up, miscommunication, and delayed treatment.

The need for a comprehensive and automated method of DR screening has long been recognized, and with this POCs we are trying to address this problem using image classification, pattern recognition, and machine learning.

Methodology / Approach

  1. Pre-processing - In this stage, we first resize all the images to 256 X 256 pixels and convert it into grayscale. Then we perform blurring, normalization, and convert it to binary to get better insight at the veins and arteries of the eye. We also rotate the images by some degree to increase the dataset. We use both the processed and unprocessed images in our models to get better accuracy. This enhances our image quality with respect to our deep learning models.

  2. Features extraction - In this stage, we apply deep Convolutional Neural Network (CNN) to extract features from the images. CNN automatically detects the features required for identification. We also use dropout, batch normalization and other techniques to get better features and control the CNN. Once the features have been extracted, we apply a combination of Fully Connected layers to learn those features and find patterns to classify classes. At the last, there is a softmax layer that evaluates the probabilities of the classes.

  3. We created two models: The first model only classifies whether the image has DR or not, the second model takes the DR images and classifies it into different stages of DR.

Technologies Used

Programming Language: Python3
Libraries: pandas, numpy, opencv, keras, intel tensorflow, skimage
Platform: Intel® AI DevCloud

Comments (0)