Today, glaucoma is already responsible for 12% of all cases of irreversible loss of sight and is the second biggest cause of blindness worldwide. Estimates indicate that more than 100 million people will be diagnosed with glaucoma by 2040.

Over the last few decades, we have seen a sharp rise in life expectancy. We are living longer and healthier lives, but can’t prevent chronic eye diseases from beginning to develop at a certain point. Eye diseases such as glaucoma can lead to impaired vision, causing a disability in people and making them dependent for their day-to-day activities. The crucial question is how we can detect glaucoma faster and more accurately long before the debilitating damage occurs. This is exactly the challenge that VITO has taken up, and its partner in crime is Prof. Ingeborg Stalmans, ophthalmologist and world expert in glaucoma.

What exactly is glaucoma?

Prof. Stalmans: Glaucoma is a collective name for disorders of the eye that damage the optic nerve in such a way that there is an impairment or loss of vision. The main cause is the channels for draining aqueous humour not working as they should. This leads to an accumulation of fluid. The pressure on the optic nerve increases and optic nerve fibres begin to die off. The most common form of glaucoma – chronic glaucoma – appears very gradually. With acute glaucoma, the opposite is true: it appears very suddenly, the patient feels it quite distinctly and his or her vision rapidly becomes clouded. In the latter case, action needs to be taken immediately.

Can anyone get glaucoma?

The risk of getting glaucoma increases with age. It usually develops after the age of 40. There is also a genetic component. Having a family member with glaucoma increases the risk of developing the disease yourself at a younger age. There are also children with glaucoma, and even babies can be born with glaucoma, although that is thankfully very rare.

How do you know if you have glaucoma?

Chronic glaucoma in itself is painless. Initially, it only affects peripheral vision. As the disease evolves, this zone extends further towards the centre and the patients sees less and less clearly. However, the brain can mask this and – certainly in the early stages – compensate for the lack of information from the periphery with information that it receives from the surrounding zones and via the other eye. As a result, the patient often becomes aware very late that his or her vision is worsening. That is why chronic glaucoma is also known as the “silent thief of sight”.

Does that also mean that diagnosis is often only made at a later stage?

Yes, definitely. People are often not sufficiently aware of the fact that glaucoma can develop as they age. In many cases, they only come for a consultation when they have been having persistent problems with their sight and there is often already irreversible damage when the diagnosis is made. The sad thing about late diagnosis is that we can usually stabilise glaucoma from as soon as the first signs are found. Correct and timely diagnosis is therefore crucial for avoiding or limiting as far as possible the effects of glaucoma on the field of vision.

And how can you make the diagnosis?

Making a correct diagnosis is challenging for an ophthalmologist. They measure the pressure in the eyeball, evaluate the optic nerve and test the field of vision. However, it is estimated that 25-30% of people with glaucoma have normal eye pressure. In addition, the field of vision is often still normal during the early stages of the damage to the optic nerve. That is why a thorough evaluation of the optic nerve is crucial for detecting glaucoma at an early stage. However, manual evaluation of the optic nerve by a clinical examination or by interpreting photos is time-consuming and often complicated, which is why screening for glaucoma on a large scale is currently not cost-effective. Software applications that help to accurately interpret the available information would therefore be a useful tool for solving this problem and thus more effectively combating blindness caused by glaucoma.

This brings us to the cooperation with VITO.

We are working with the MONA team of VITO’s Health unit to develop computer models for automatically analysing retina images. There is not only a doctoral partnership in progress here, but many research initiatives have also been launched at national and international level.

Patrick De Boever (VITO): During the visit to the ophthalmologist, a digital photo is taken of the retina. The partnership with UZ Leuven made many thousands of images available for the research anonymously via the electronic patient file. We then had a vast database of retina scans of both ‘good’ eyes and eyes that had already been affected by glaucoma. Using artificial intelligence (AI) and deep learning techniques, we taught the computer to accurately recognise affected eyes. This work was recently published in the specialist journal Acta Ophthalmologica. At the moment, we are extending the research even further so that we can also validate the computer models for images from other international clinics. The development carried out by a number of these partnerships is currently at an advanced stage. It continues to be an enormous challenge to obtain images from patients, since we also need to bear in mind the new GDPR legislation and privacy rules, but we are convinced that we will very soon be able to release new results.

What other challenges do you see in using the computer model?

Patrick De Boever: There is increasing demand for AI models that explain why the computer model reached a certain decision. The technical term for this is “model explainability”. The intention is to provide doctors with an explanation that could increase the level of confidence in automatic diagnosis. For this purpose, we are currently developing a number of models with UZ Leuven. The first set of results, which we will soon be submitting for publication, shows that our initial computer model looks at clinically relevant changes. This is something that we have already discussed at length, and there are already foreign experts who are enthusiastic about this and wish to continue working with us. In addition, we don’t just want to know whether or not a patient has glaucoma but also want to be able to predict how his or her disease will evolve. The clinical data that we will shortly be able to study will also allow us to tackle this challenge.

What still needs to be done?

Prof. Stalmans: The partnership between VITO and UZ Leuven is already bearing its first fruit with some really promising results. We can still expect a number of challenges when it comes to being able to use AI models in practice. Ophthalmologists currently need a number of different elements to be able to make a diagnosis: patient data, eye pressure measurements, eye photos, etc. We therefore need to refine methods for also using this information in the AI models. We also need to strive towards faster detection of glaucoma in the early stages via screening, and thus not waiting for patients themselves to indicate that there is a problem. In conclusion, a faster detection of further deterioration should also help us to further fine-tune the therapy to the individual patient and thus become even more involved in ‘personalised medicine’.

The partnership with VITO is our way of playing a part in this, and we are already convinced that we will be able to continue building on this partnership in the years to come.