17th February 2023
7 minute read
Categorised under:
Industry News

The advantages and drawbacks of using artificial intelligence to diagnose and monitor eye diseases

Artificial intelligence (AI) involves programming computers to perform tasks that would usually be carried out by humans. AI systems are trained to respond to data based on specific parameters, learning to become better at sorting, categorising and evaluating data over time. Programmers can confirm whether the computer’s interpretations are correct or incorrect, and this constant reinforcement leads to increasingly accurate results, until little to no human intervention is required. Using this method, AI can be trained to analyse data, notice patterns, solve problems, make decisions, and predict outcomes in much the same way a human would.  

AI is already a big part of our daily lives – in the Netflix algorithms that tell us what to watch next based on our previous viewing habits, and on social media, where our timelines are configured based on what we’ve liked and engaged with in the past. Millions of households across the country rely on AI-powered devices, like Alexa or Siri, to answer their questions, organise their calendars and even control their household appliances. From 2016-2019, based solely on user interactions and data input, Alexa’s “skills” jumped from 130 skills to over 100,000 skills 

One of the hottest topics in ophthalmology today is the use of artificial intelligence (AI) in the diagnosis and treatment of eye diseases. AI has the potential to revolutionise the field of ophthalmology by providing faster, more accurate diagnoses and more personalised treatment options for patients, but there are also some concerns about what it means for the future of the profession.  

What are the advantages of using AI in ophthalmology? 

On the face of it, AI has lots of advantages. It can: 

  • Help optometrists to accurately diagnose eye conditions and determine whether a patient should be monitored or referred for further treatment 
  • Reduce the number of unnecessary referrals to eye hospitals and reduce the demand on ophthalmology services – which in turn will help to reduce waiting lists for patients 
  • Reduce the administrative burden on optometrists by quickly analysing scans and identifying patterns that may be difficult – or take longer – for optometrists/ophthalmologists to detect (there is evidence that AI, through deep learning, can detect early warning signs of progressive eye conditions like glaucoma before optometrists/ophthalmologists can) 
  • Help technicians to diagnose and triage eye conditions in countries where there are high levels of eye disease, but low numbers of optometrists/ophthalmologists 
  • Remove the scope for human error 

For example, optometrists spend a lot of time reviewing and interpreting eye scans. They need to identify any abnormalities within those scans, reach an appropriate diagnosis, assess the severity/grade of the diagnosed condition, and then determine whether the patient should be monitored or referred for further medical intervention. It takes years of training and experience to be able to do this effectively and, even then, when faced with a complex or rare case, it’s often necessary to seek a second opinion. Many optometrists decide it’s better to be safe than sorry and refer a patient to their local hospital for further investigation.  

However, research carried out by the Association of Optometrists in 2019 showed that, for some eye diseases, up to half of referrals made from optometrists to hospitals turned out to be “false positives” – meaning that many of the patients referred to hospitals didn’t have the eye condition they were being tested for and were found not to require further treatment. False positives can put more strain on a healthcare system that’s already overburdened. In their report on The State of the UK’s Eye Health, Specsavers noted that “ophthalmology is the busiest outpatient speciality in the NHS” and, with over 633,000 people waiting for an ophthalmology appointment as of March 2022, AI could play an invaluable role in helping optometrists reach faster and more accurate diagnoses, prioritise urgent referrals, and only refer patients to hospital when it’s necessary to do so. In turn, patients will be seen and treated more quickly, and their visual outcomes will likely improve.  

How accurate is AI? 

In their research ‘Artificial Intelligence for Screening of Multiple Retinal and Optic Nerve Diseases,’ published in May 2022, researchers based in China developed a Retinal Artificial Intelligence Diagnosis System (RAIDS) using a dataset of over 120,000 ocular fundus photographs. The AI system was able to “accurately distinguish 10 retinal diseases in real time” and achieve a level of diagnostic accuracy that was similar – or in some cases superior to – senior retinal specialists. 

In the UK, researchers from Moorfields Eye Hospital and University College London’s Institute of Ophthalmology were able to train an AI system to identify signs of eye disease and recommend whether patients should be referred to hospital for further treatment. The AI system was able to make the correct referral decision for over 50 different eye diseases with 94% accuracy, matching the decision-making capabilities of world-leading eye experts, but much more quickly and efficiently. 

94% accuracy is impressive in its own right, but some AI systems have been able to diagnose and differentiate between complex eye diseases with 99.8% and 100% accuracy. 

Researchers are also using AI to analyse retinal scans and identify risk factors for other health conditions, such as cardiovascular disease and strokes 

Evidence would suggest that – providing an AI system has enough data to work with and is trained effectively to “think for itself” – AI is extremely accurate and reliable. 

What are the drawbacks of using AI in ophthalmology? 

While the benefits of using AI in ophthalmology are clear, it’s not without its drawbacks. In fields like tech, manufacturing and customer service, AI has already started to replace human workers and, naturally, one of the main concerns is the potential for AI to replace optometrists who have trained for years to master their profession. However, there’s a long way to go before that becomes a possibility, let alone a reality. Clinical trials of AI are still ongoing, and most healthcare providers don’t have the budget or the resources to invest in the development of AI technology. While the NHS has a dedicated AI Lab, most projects are in their early stages, and it could be years before AI technology is ready to roll out across the country, although the ARIAS project – which will gather evidence to support the commissioning and deployment of Automated Retinal Image Analysis Systems for use within the NHS Diabetic Screening Programme – is “on the horizon.”  

To be effective, AI also requires access to large-scale datasets that are representative of the population as a whole. This will often require collaborative working across several providers, which gives rise to questions surrounding patient privacy, data protection and informed consent. AI systems that aim to predict risk factors for other diseases often require access to a patient’s entire medical history – in addition to monitoring the patient’s health over time to determine whether the predictions proved accurate. 

It’s also worth noting that AI systems – during their initial development – still rely on humans to “program” them effectively and decide how the results of their algorithms will be applied. So, while AI systems can help to identify patterns and make recommendations, it’s up to the optometrist/ophthalmologist to interpret the data and make the final decision. If an incorrect outcome is validated in error, this could lead to bias within the AI system itself. 

Given that the field is constantly evolving, it’s also essential to ensure that AI remains up to date and its performance is monitored and evaluated at regular intervals to ensure continued accuracy –eating into the time that AI can potentially save. 

In conclusion, AI has the potential to revolutionise the field of ophthalmology by providing faster, more accurate diagnoses and helping to reduce waiting lists for treatment, but further investment and advances are needed before optometrists can truly benefit from its application in real world scenarios.  

Did you find what you were looking for?

Thank you for feedback