Alzheimer’s disease – the silent pandemic and the positive impact of AI

Using AI, future eye exams could be used to detect early signs of Alzheimer’s and heart disease

By Joachim Behar

Alzheimer’s disease has rightly been characterised as a ‘silent pandemic’, with the number of Americans over 65 who are living with Alzheimer’s projected to double by 2060 as the population ages, to 13.8 million.

More broadly, dementia affects over 55 million people worldwide and the economic impact is extensive as well, with the burden of dementia estimated at $2.8tn globally and US costs for health and hospice payments alone estimated at $360bn, in addition to billions of hours of free care provided by over 11 million relatives and unpaid caregivers. The silver lining is that this affords us an impactful way to use artificial intelligence (AI) to make healthcare better, by detecting Alzheimer’s disease – and, eventually, other diseases too – at earlier stages, in a non-invasive and low-cost way that can be as simple as an eye exam.

Current methods for detecting asymptomatic Alzheimer’s, like a lumbar puncture to analyse patients’ cerebrospinal fluid, or PET or MRI imaging, tend to be expensive, invasive or have limited availability. Researchers have been investigating other methods as well, but this one is exciting: what if someone could walk into an ophthalmologist’s office for a routine check-up and receive an eye exam that also screens for Alzheimer’s disease – and ultimately for other significant but silent conditions, like cardiovascular disease.

The good news is that the technology for AI-supported retina scans that can detect Alzheimer’s is already mature. Now it’s time to lay the groundwork to start deploying this technology at doctors’ offices, healthcare clinics and hospitals.

In the world of metaphor, the eyes are the windows to the soul. In the world of science, the retina is the window to human vasculature and neural connections. It’s the only place on the body that doesn’t require advanced imaging equipment or invasive procedures in order to get a glimpse of the blood vessels. As such, the retina holds a great deal of potential as an entry point for the detection of systemic diseases like neurological disorders and heart disease.

In October 2024, an article in the scholarly journal npj Digital Medicine reported that the deep learning framework Eye-AD excelled at detecting early-onset Alzheimer’s disease and mild cognitive impairment, using a combination of retinal imaging and artificial intelligence.  The multi-centre study involved 1,671 participants, forming by far the largest data set for use of the non-invasive imaging procedure known as optical coherence tomography angiography (OCTA) in detecting Alzheimer’s and mild cognitive impairment.

‘Retinal imaging is less expensive, simpler and faster, and has greater feasibility for smaller hospitals or community screening programmes, when compared with a conventional [Alzheimer’s disease] diagnosis protocol,’ the study found, while a literature review published last year reported that ‘the growing application of AI in medicine promises its future position in processing different aspects of patients with [Alzheimer’s disease]’.

The merits of non-invasive retinal scans bolstered by AI capabilities go beyond the lower cost, greater convenience and lack of pain or side effects. Another big benefit is that the detection is opportunistic, meaning that it doesn’t require a patient to specifically come in for an Alzheimer’s diagnosis, but can be performed on any patient who goes for an eye exam – which is particularly important for those at an early stage who are asymptomatic.

At the same time, there are multiple challenges that the medical community still needs to work on in order to turn the technology from functional viability to common practice.

The fact that the technology is mature is a huge milestone, but the process of making it accessible to patients still lies ahead. The next few years will require regulators, retinal scan and AI technology providers, healthcare facilities, medical personnel, insurers and investors to work out the best way to approve, develop, monetise and deploy the technology, and possibly build new healthcare or laboratory processes that make the new technology part of the day-to-day routine.

Translational medicine approaches in particular can help make eye-based diagnostics more common by increasing collaboration among university researchers, healthcare facilities and medtech companies to ensure the technology moves from research to practical application. This will help accelerate innovation, improve technology and help new diagnostic tools reach real patients faster.

Indeed, AI-driven analysis of retinal scans for diagnosing and staging ophthalmic and cardiovascular conditions has already achieved clinically relevant performance, making it suitable for screening and monitoring programmes.

For instance, the AI platform the lab at The Technion Israel Institute of Technology has built to identify biomarkers for specific cardiovascular and ophthalmic conditions is already available to researchers. Called Lirot.ai, the platform is based on the associations that researchers have discovered between the retina and the cardiovascular system.

Multiple studies have shown that blood vessels can be seen in non-invasive scans of the retina. Such scans are captured by medical devices used to photograph the back of the eye, known as the fundus, which includes the retina.

Given that the cardiovascular system is made up of the heart and the blood vessels, researchers have found that certain characteristics of the vessels seen in the retina serve as an important biomarker for cardiovascular disease – which causes about one-third of all deaths globally. Variations in the retinal microvasculature can be used to screen individuals at risk of developing heart failure, coronary artery disease or stroke.

At the lab, we’ve taken that research further by developing deep learning algorithms trained and evaluated on large data sets from various geographical locations that enable it to automatically segment the retinal microvasculature and generate digital vasculature biomarkers. The biomarkers can be used to identify changes in small arteries (arterioles) and small veins (venules) that can indicate cardiovascular disease. This means we can now use AI to find associations between heart conditions and eye scans. That’s what translational medicine is all about – taking our theoretical understanding of the body and applying it to real-life medical situations.

In addition to cardiovascular disease detection, Lirot.ai can also be used for the automated flagging of eye diseases whose early detection is critical – without the patients needing to visit an ophthalmologist. For instance, the deep learning algorithms can detect diabetic retinopathy, a complication of diabetes that can lead to rapid vision degradation and, eventually, irreversible blindness – yet can be slowed down or halted if detected in time.  Liort.ai can also use retinal scans to detect glaucomatous optic neuropathy, which can also lead to irreversible vision loss if not treated early. Additional modules are in development and will be announced soon.

We’ve hit a big milestone on how to combine retinal scans with AI to detect ophthalmic and cardiovascular conditions as well as early-onset Alzheimer’s disease, but there’s a long way to go. Research must continue to see how to expand the use cases to other types of diseases as well as other neurological disorders. This continued research requires more data on which to train and test the AI models as well as more financial resources. Even for the Alzheimer’s findings, the authors note that their data set was relatively small for deep learning purposes and lacks ethnic diversity, as it was limited to Chinese patients.

As wonderful as it is to make progress on early Alzheimer’s detection, it will be far more meaningful when Alzheimer’s treatment improves such that detecting it at an early stage can one day prevent the onset of the disease or cure it.

Although there is still a long road ahead before we see AI-assisted retinal scans in day-to-day use to detect Alzheimer’s disease during routine eye exams, the technology is there, and implementing it has the potential to significantly improve the early detection of this silent pandemic, along with many other diseases.

Find out more about the Artificial Intelligence in Medicine Laboratory (AIMLab.) at the Technion-Israel Institute of Technology’s Faculty of Biomedical Engineering at aimlab-technion.com/ and the Zimin Institute for AI Solutions in Healthcare at zimininstitutes.org/

Joachim Behar, Associate Professor, is the head of AIMLab. and a member of the Zimin Institute