Chronic Disease – Are we blind to the problem?

Share this post:

Written by Dr Joanna Batstone, Vice President and Lab Director, IBM Research

Blindness. Simply the word evokes a fear in many of us. In fact, blindness is the second most feared medical condition among Australians, after cancer. It steals your independence. It significantly impacts your capacity to work. And it increases your reliance on support from others and the health system. For far too many in the world, they remain unaware that this is what lies ahead for them, thanks to a global epidemic. What do many of these people have in common? Diabetes.

In the five minutes it takes you to read this, one Australian will develop diabetes. Today, approximately 1 million Australians have the disease, with this number expected to double by the year 20252. Almost all people with type 1 diabetes, and 60 percent of those with type two diabetes will develop diabetic retinopathy (DR), the leading cause of irreversible blindness in Australia1. If left undetected, DR will cause permanent loss of sight, yet early detection of the disease can reduce the risk of blindness by 95 percent. This sounds promising, yet it is estimated that 50 percent of people who have diabetes don’t know they have it, and could potentially be unaware they are slowly losing their sight.

While the economic burden of diabetes eye disease is dwarfed by the personal and social cost to the individual, it is estimated that the indirect cost of the disease for a single patient each year is $28,0002. It is no surprise then, that chronic disease, including diabetes is considered the greatest challenge that our nation faces today. So what can we do to reduce this burden?

Diagnosis of DR requires regular screening of diabetes patients, where an expert clinician examines specialised images of the back of the eye, called colour fundus photography. Clinicians manually assess the image to look for tiny signs of lesions such as micro-aneurysms, haemorrhages and exudates. These pathologies indicate the presence of the disease and how severe it is for the patient. Interpreting these images is often a manual, time-intensive and subjective process.

With the estimated growth in diabetes, our current infrastructure is not enough to tackle the challenge in preventing blindness from DR.

What if emerging computer vision technology could help create a highly accurate, automated system for DR detection?

At IBM Research we are using deep learning technology combined with pathology insights to train machines to understand what a constitutes a normal eye structure, and to automatically identify lesions which may indicate the presence of diseases like DR. Our newly proposed method achieves an accuracy score of 86 percent in classifying the severity of the disease across five levels (no DR, mild, moderate, severe, proliferative DR). Emerging computer vision methods to identify and classify lesions in an image within seconds could create new levels of efficiency, which could help clinicians screen a greater number of diabetes patients, and quickly refer those who need specialist care.

Insight into severity is an important result, given clinicians need to understand the progression of the disease to be able to inform appropriate patient care and treatment. For example, according to the Centre for Eye Research Australia, early stage DR can be managed through optimal diabetes control and maintaining a healthy lifestyle, however proliferative DR requires treatment which is much more invasive and costly, for both the patient and the health system.

We are still in early stages of our research, but these results have shown a great promise for an automated system to help clinicians be more efficient in their analysis of diabetes patients, and quickly refer those who need specialist care. We’re part of an exciting new era; an era which offers the potential to make a real difference in the lives of those who need it most, and hopefully enable us to make headway against this global epidemic of chronic disease.

More Watson Stories

The original story telling machine

Written by: Anne Nicholson, Brand Strategy Lead A/NZ A recently rediscovered article in Life magazine, has uncovered a research project that used an IBM computer to generate original stories in the early 1960s. The computer was the IBM 650 Magnetic Drum Data Processing Machine and it predates IBM’s Watson by more that 40 years. In […]

Continue reading

Artificial Intelligence helping drivers get back on the road faster

Written by: Gary Dransfield, CEO Insurance, Suncorp Data has always been at the core of the insurance industry, whether it be for calculating premiums, responding quickly to disasters or accurately determining liability for claims. Typically, when a customer interacts with an insurer, it’s because they’ve gone through a highly stressful incident like a car accident, […]

Continue reading

My thoughts can be like pieces of a mismatched jigsaw puzzle

Reproduced with the permission of the author, Joshua Gliddon IBM scientists and the University of Alberta in Edmonton, Canada, published new data in Nature’s partner journal, demonstrating that AI and machine learning algorithms helped to predict instances of schizophrenia with 74% accuracy. This retrospective analysis also showed the technology predicted the severity of specific symptoms […]

Continue reading