Ocuscreen — Diabetic Retinopathy Detection Platform

Unprecedented accuracy in diagnosis secured by AI Machine Learning
Python, PyTorch, Google cloud, Docker, Django, Qt 5, ReactJS

Challenge


Diabetic Retinopathy is a medical condition in which damage occurs to the retina due to diabetes and might lead to blindness. Proper screening can determine harmful signs of disease and prescribe treatment to slow down this process.

ENBISYS needed a solution for multiple Clients that can be used as a new software feature for ophthalmology diagnostic equipment. This feature may be easily integrated into a device circuit and may become a great competitive advantage.
The neural netrowk was trained on 120,000+ images to achieve outstanding accuracy:
Specificity: 91%
Accuracy: 96.5%
Response time: 4 seconds

Approach

We considered various approaches that helped us build an efficient and precise AI platform to detect the signs of Diabetic Retinopathy. Our internal model was developed based on open data (EYEPACS, Messidor-1). Machine Learning is used to perform the following tasks:
Define if retinal pictures are gradable/non-gradable
Identify left/right eye
Detect the stage of diabetic retinopathy, 1 to 4

Solution

A professional ophthalmologist will spend 2 minutes on a diagnosis instead of 5
With the help of OcuScreen AI platform:
A grader will miss less early-stage cases of disease
Diabetic retinopathy screening becomes accessible to more patients
Technology helps to focus on important parts of the image
Good accuracy in diagnosis
Lesions segmentation is helpful
Interface usability is nice
In most of 'False Positive' cases experts tend to believe the model
— What doctors say
prediction rather than the original label
Let's discuss your Case!
Let's discuss your Case!
Let's discuss your Case!