Search


Professor Tariq Aslam on how human intelligence can be a superpower in the age of AI

The consultant ophthalmologist shared tips on using OCT–A and highlighted the limitations of artificial intelligence during his presentation at HSOC

Professor Tariq Aslam delivers a presentation at Hospital and Specialty Optometrists Conference
OT

Manchester Royal Eye Hospital consultant ophthalmologist, Professor Tariq Aslam, explored the potential optical coherence tomography angiography (OCT-A) during his presentation at the Hospital and Specialty Optometrists Conference (14–15 November, London).

Aslam shared that becoming familiar with OCT-A can be a learning curve because of the number of artifacts and the need to interpret multiple images.

“It’s a multimodal system. Within a single OCT-A, there’s lots of different images that you can see,” he said.

He noted that the structural image is the equivalent of a normal OCT scan.

“The structural image is important because sometimes if you have gaps missing or extra bright areas, you think ‘Is this an artifact or not?’,” Aslam said.

Aslam observed that artifacts can make it more challenging to interpret an OCT-A scan.

He explained that projection artifacts can occur when signals from a blood vessel above is seen in a lower layer.

“The more complicated thing is that some of the imaging companies have developed algorithms to try and move these,” Aslam explained.

“Sometimes you see what looks like a dark blood vessel, but it is the algorithm cancelling out that projection,” he said.

Aslam shared that a masking artifact occurs when a view is obstructed by overlying structures.

He highlighted the importance of reading an OCT-A scan methodically – taking into account the en face angiogram, B-scan with flow and structural en face.

Aslam emphasised that clinicians should rule out projection, masking, segmentation and movement artifacts before determining that something is a clinical sign.

The power of humans in the age of AI

Reflecting on the limitations of artificial intelligence within his presentation, Aslam shared that there can be a lack of transparency in the way that AI makes decisions.

“We have to have a way of understanding what the outputs are,” he said.

As the use of AI technology advances, Aslam highlighted that there is still an important role for clinicians to play in looking after patients.

“The key aspect where you will be harnessing your own superpowers is by playing an interpretive role – understanding the patient and empathising with them,” he said.

He highlighted that AI may be appropriate for some applications but not others – for example, it could be useful for screening but not for managing a patient with vision-threatening disease.

“There is a need for monitoring and being aware – not just of the patient’s ocular condition but of their holistic wellbeing. Anxiety and depression can be quite high in these patients,” Aslam noted.

Aslam shared that there is a broader question about the type of future that we are heading towards with the growing influence of AI technology.

“Are we heading towards a system where we are just acting as transmitters – we get information, the scan is done, it is interpreted and we transmit that to the patient?” he said.

“Or do we want a future where the technology is used for administrative tasks that will make more time for us to have face-to-face conversations with patients?” Aslam said.

Reflecting on the value of communication skills for patient wellbeing, Aslam noted that there is a free RNIB Mind Your Eyes course online – with modules on communication and the emotional toll of sight loss.

Aslam shared with the audience that before the pandemic, research was published showing that an AI algorithm could analyse OCT scans with a similar level of accuracy to consultants.

“However, it’s still not being used in practice,” he said,

Regulatory issues, the need for external validation and privacy considerations have all contributed to delays in translating AI technology into practice.

Aslam noted that there is a risk of deskilling if clinicians are only overseeing AI algorithms rather than applying clinical judgement.

He referred to a US study where the skill level of radiographers in identifying brain tumours dropped after an AI system was implemented.