Diabetic Retinopathy – Microanalytical Blogs

Diabetic Retinopathy(DR) is a disease caused by diabetes. It is one of the most common problem occurred in diabetic persons

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.

So here to resolve this problem we at MICROANALYTICAL have created an AI based approach where we can deploy a Kiosk machine with a camera which captures retina images and a edge device or some computer to deploy the model, user would take the snapshot of their eyes retina and model should be able to inference and detect level of diabetes in the captured retinal image .

Methodology

Our team have used a Deep Learning Technology to provide a full software to determine whether the patient is affected with Diabetic Retinopathy or is Healthy. We have spent a larger amount of time in building a best in class Classification System.

Neural Network used has been developed from the scratch using Classification algorithm.

Classification

Classification is an algorithm where the output variable is a categorical data or a variable as a category such as cat or dog, red or blue, diabetic or no diabetic.
Classification problem is mainly of two types:

Binary classification: eg: email spam detection, which each email is spam (1) spam; or isn’t (0).
Multi-class classification: eg: recognizing handwritten characters from 0-9.

The below example will represent binary classification:
We have 2 classes, circles and squares and two features , X1 and X2.the model is able to find the relationship between the feature of each data point and it’s class, and to set a boundary line between them ,so when provided with new data it can estimate the class where it belongs, given it’s features. Here , the new data point tend to fall in the portion where circle lies. Therefore, the model will predict it to be a circle.

We have used Convolutional Neural Network(CNN) to resolve image related problems.

Convolutional Neural Network (CNN)

A Convolutional neural network (CNN) is the foundation for most computer vision technologies. Unlike traditional multi-layer perceptron architectures, two operations generally used are “convolutions” and “pooling” to reduce an image Into it’s essential features, and these features will be used for understanding and classifying the images.
The basic building blocks of CNN are :

Convolutional layer
Activation layer
Pooling layer
Fully connected layer

There may be multiple activation and pooling layer, depending on the CNN architecture.

We have used ResNet50 architecture. ResNet , short for residual network is a classic neural network used as a backbone for many computer visions tasks. The fundamental break through with ResNet was t allowed us to train extremely deep neural network with 150+layers successfully.

Kaggle is the place where you can get a huge amount of data images. Generally every year Kaggle competitions occurs where lot of companies post data and data miners tend to find useful models for predicting and data description. We can use this dataset for our analysis and prediction, its freely available to all community for use.
Here total we have 16,798 images in which 26% images are of DR and 74 are of NoDR.

How it works

Deploy Diabetic Retinopathy Detection

MICROANALYTICAL offers an enterprise solution that is pre-configured and ready to deploy. MICROANALYTICAL’s Diabetic Retinopathy Detection System is a proactive and ready to determine whenever the input is feeded. This helps in determining about persons health in a fraction of time.

Diabetic Retinopathy Determination

Once a result is appeared, it sends image feed to your staff, administrator or whomever you designate. At the same time, it can also sent to other enterprise security platform.