Can Artificial Intelligence help close the indigenous healthcare gap?

3 years ago admin Comments Off on Can Artificial Intelligence help close the indigenous healthcare gap?

Improving the health and welfare of indigenous Australians has been a longstanding challenge for our nation. Earlier this year, the government’s ninth annual Closing the Gap report showed only one of the seven targets will be met this year. And unless considerable strides are made to close the chasm, we are at a real risk of moving backward instead of forward.

The life expectancy for Aboriginal and Torres Strait Islander people is approximately 10 years less than non-indigenous people. When you think about what you’ve experienced over the past decade in your life, it’s difficult to swallow this number as a mere statistic.

Chronic illnesses such as diabetes, cancer and heart disease are at the root of the health gap, with estimates indicating that 80 per cent of the mortality gap for indigenous Australian adults being due to chronic disease. And the burden of a chronic disease like diabetes goes beyond a shortened life expectancy and the strain on the health system, with some of the most debilitating complications including amputations and diabetes eye disease which causes irreversible blindness. One third of people with diabetes have the eye disease and are at risk of losing their sight, significantly impacting a person’s independence and capacity to work.

Diabetic Retinopathy (DR) is the leading cause of preventable blindness in Australia. The good news is that 95 per cent of blindness from the disease can be prevented if detected early enough. However, in far too many instances, people are unaware that they have DR as it can progress to an advanced stage before obvious symptoms occur. Regular eye examinations by expert clinicians like optometrists and ophthalmologists are essential to catch the disease and intervene with treatment before irreversible vision loss occurs. Unfortunately, due to a lack of expertise and equipment, regular screening and treatment is something many people in remote areas don’t have access to as easily as those in major cities. The Centre for Eye Research Australia estimates that less than 40 per cent of indigenous Australians showing early stages of DR had consulted a health care provider in the preceding year. Many are going blind without even knowing it.

This is where Artificial Intelligence technology could step in.

Scientists in IBM’s Research lab in Australia have trained machines to understand what constitutes a normal eye structure, and identify the subtlest details in eye images, such as a micro-aneurysm or haemorrhage which indicate the presence of DR. Using deep learning and visual analytics, the newly published methods can pick up tiny lesions in an eye image and through analysis of where they are distributed in an eye, can determine how severe the disease is for a patient. The analysis is completed within seconds and achieves an accuracy score of 86 per cent, the highest recorded accuracy using deep learning techniques and pathology insights for machine-assisted DR severity analysis.

Being able to simply identify these tiny signs is often difficult, even for the most skilled clinician. Currently eye specialists manually find and assess each lesion to estimate the severity of the disease and advise appropriate treatment. However if we could use AI technologies to automate what is typically a time consuming and manual process, clinicians may have the potential to screen a greater number of patients at a faster rate.

It could also offer new possibilities for greater efficiency in telehealth services, which are in place today in our remote communities. Recent changes to national legislation last year allow GPs to screen diabetes patients for DR via telehealth services, capturing high resolution eye images of patients and sending the images to eye specialists for remote analysis. While this is a big step forward for remote and rural communities in Australia, the delayed results mean the chances of missed follow up appointments remain the same. However, what if we could give these GPs access to technology to not only capture these images but to accurately identify patients with signs of DR? The patients referred to specialists could be narrowed to those who have been identified as presenting with the pathologies of the disease, reducing the need for specialists to manually analyse every patient image. It could also help reduce the chances of missed follow up appointments, as patients would get an indication of their risk of DR on the spot.

Benefits won’t just be felt at the individual level but to Australia as a whole. Currently the total annual financial impact of diabetes is $14.6 billion. And with the number of people diagnosed with diabetes growing rapidly each year, the strain on the economy will also deepen, locally and globally. Now, more than ever before, governments around the world are looking to do more with less. That’s precisely where we see the future potential for this technology to make the most impact. By being smarter and more efficient through the use of AI technology for health delivery, we may be able to not only ease the burden on our health systems but also allow more people to keep their independence and capacity to earn.

It’s important to recognise this research still has a while to go before it’s in the hands of those who need it, but these latest results demonstrate that the potential is there. The potential which could help Australia make further steps forward in our efforts to close the gap on indigenous healthcare.

Dr. Joanna Batstone is Vice President and Lab Director, IBM Research — Australia

Reader comments on this site are moderated before publication to promote lively and civil debate. We encourage your comments but submitting one does not guarantee publication. We publish hundreds of comments daily, and if a comment is rejected it is likely because it does not meet with our comment guidelines, which you can read here. No correspondence will be entered into if a comment is declined.