Latest News
-- particles

SELENA+, the Intelligent Deep Learning System to Prevent Diabetic Blindness

08.01.2019

Background

fundus

  • Diabetes mellitus (DM) is the world’s fastest growing chronic disease. Today, there are approximately 415 million diabetics and this number is projected to increase to 642 million by 2040.
  • Diabetic retinopathy (DR) is a major consequence of diabetes caused by “leaky” blood vessels in the retina, and is a leading cause of vision loss. Early detection via screening and prompt treatment of DR allows for the prevention of visual impairment.
  • Retinal photography is the most common and cost-effective DR screening method. However, the assessment (grading) of photographs is commonly performed by humans (ophthalmologists, optometrists or professional graders). With the growing diabetes population, assessment of hundreds of millions of photographs will be unsustainable in the future.
  • Our team has developed a state-of-the-art artificial intelligence (AI) system to automatically perform primary assessment of retinal photographs, thus significantly reducing the public health care cost while maintaining its standard.

SELENA+, the Intelligent Deep Learning System to Prevent Diabetic Blindness 

AI

SELENA+ is an intelligent deep learning system (DLS) designed to perform automated image analysis for diabetic eye diseases. DLS, a breakthrough machine learning technology, utilizes representation-learning methods with multiple levels of representation to process natural data in their raw form, recognizing the intricate structures in high-dimensional data.

DLS develops from previous machine learning technology using “feature-based” and “pattern recognition” techniques. SELENA+ ‘learns’ from experience like a human brain so it is critical to have a large input data to enable the mining and extraction of meaningful patterns or features via the convolutional neural network (CNN) – to build a dictionary or an encyclopaedia. In other words, the greater amount of “training” SELENA+ goes through, the more accurate it becomes.

SELENA+’s excellent image processing capabilities was developed and originally trained on anonymised retinal images acquired from diabetic patients who attended the Singapore Integrated Diabetic Retinopathy Programme (SIDRP) over 10 years. Subsequently, a multi-centre global collaborative effort involving more than 10 centres such as Mexico, United Kingdom, New Zealand, Australia, United States, China and Thailand was conducted to validate the program. No other AI software in the world can claim to have undergone the same stringent validation process.

Read about The Deep Eye Study in December 2017 Journal of the American Medical Assosciation.

SELENA+, a deep learning system that yields unparalleled results

Test measurements as published in The Journal of the American Medical Association (12 Dec 2017) shows very promising and consistent results at over 90% sensitivity and specificity in detecting referable DR and over 95% for vision-threatening DR. The repeatability (same image tested twice) and negative predictive value were 100% and more than 99%, respectively.

 

 

AUC

Receiver Operating Characteristic Curve and Area Under the Curve of the DLS for Detection of Referable Diabetic Retinopathy and Vision-Threatening Diabetic Retinopathy in the Singapore National Diabetic Retinopathy Screening Program (SIDRP 2014-2015: Primary Validation Dataset). Compared with Professional Graders Performance, with Retinal Specialists' Grading as Reference Standard

 

This has proven that SELENA+ with its DLS is a reliable and accurate tool that can consistently provide a timely and cost effective strategy to assess retina images from DR screening programs.