Diabetic Retinopathy Detection using Intel® Distribution of OpenVINO™ Toolkit.

Pranab Sarkar

Pranab Sarkar

Jalpaiguri, West Bengal

15 0
  • 0 Collaborators

This application can predict different stages of Diabetic Retinopathy with a decent accuracy. The reason behind this research work is to create a cheap and reliable DR detection solution which can be easily accessed irrespective of proper internet connectivity. ...learn more

Project status: Published/In Market

Artificial Intelligence

Groups
Artificial Intelligence Europe, Artificial Intelligence India, Early Innovation for PC Skills, Internet of Things

Intel Technologies
OpenVINO, AI DevCloud / Xeon

Code Samples [1]

Overview / Usage

In India, there is a shortage of 127,000 eye doctors and 45% of patients suffer vision loss before diagnosis. 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. There is a need for a comprehensive and automated method of DR screening an automated detection system to the limit of what is possible – ideally resulting in models with realistic clinical potential.

In this project our model can give a prediction on the various stages of Diabetic Retinopathy:

  • No DR
  • Mild
  • Moderate
  • Severe
  • Proliferative DR

Deep learning model demands high computation power for the model inference or alternatively cloud computation could also be leveraged if there is proper internet connectivity available. But, there are many remote places throughout the globe which doesn’t have access to the internet, in these type of scenarios we can use this remote IoT device to fill the gaps.

This solution uses the model optimizer and inference engine of the Intel® Distribution of OpenVINO™ toolkit which gives another dimension to this project, using which we can achieve state of the art performance with our IoT device.

Hope this type of innovation saves many lives!

Methodology / Approach

  1. Data Collection: The images are was collected from a dataset hosted by Asia Pacific Tele-Ophthalmology Society (APTOS).
  2. Data Preparation: The dataset contains many imbalanced classes, various data augmentation and up-sampling techniques were applied to achieve a balanced dataset.
  3. Image Processing: To gain superior model accuracy the images were processed using filters.
  4. Defining the model architecture: A custom model architecture was build using DenseNet BC 169 in the initial layers which was trained over the entire dataset.
  5. Training the Model in DevCloud.
  6. Deployment: The model could be deployed inside a Flask Web App or the IOT device using the OpenVINO™ toolkit.

Technologies Used

  1. Intel Dev Cloud.
  2. Intel® Distribution of OpenVINO™.
  3. Tensorflow 2.0
  4. Python 3.6
  5. ReactJs

Repository

https://colab.research.google.com/drive/1OgjdF09DiUuek47IhKP2nYfOs4aw0HJ5

Comments (0)