Opportunities for AI in Ireland’s healthcare system
A device which can recognise 75 different abnormalities on chest x-rays, and the ability to detect Parkinson’s disease from a picture of the back of the eye, are just two examples of benefits which artificial intelligence (AI) could unlock for Ireland’s healthcare system to enhance patient care, says HSE consultant nephrologist and senior lecturer at University of Galway, Conor Judge.
Spending roughly half his time working in the HSE and the other half as a senior lecturer in applied clinical data analytics at the University of Galway, Judge’s primary focus is applying AI solutions to enhance patient care, particularly in the realm of chronic conditions like hypertension.
“We have huge unmet clinical research need whereby the number of AI publications emerging in healthcare, are not matching the number of AI or machine learning enabled medical devices approved by the FDA,” Judge states.
“There are really interesting research examples of AI in healthcare, some of which are being used in Ireland, but most are not in any of our hospitals. The experience of AI devices being used in other countries are very telling and indicate that they can deliver better patient care.”
He points to a medical device called ChestLink, which was developed by Oxipit and was the first AI device that got the CE mark for fully autonomous reporting.
“Oxipit developed this computer vision and model that was able to recognise 75 different abnormalities on chest x-rays,” Judge explains.
“If it did detect one of those 75 abnormalities, then the image was sent back to the consultant radiologist who would then report on the chest x-ray.”
Judge states that this an example of “task sharing between AI and the physician”, and emphasises that this is something that we will “see more and more of in the next few years”.
The University of Galway lecturer also notes another example of AI technology being used to determine the early changes of Parkinson’s disease. A team of researchers in Moorfields Eye Hospital in London, led by Irish ophthalmologist Pearse Kane, developed a computer vision model, training it to consider 1.6 million pictures of the retina.
“The computer vision model is able to detect early changes of multiple eye conditions, such as age related macular degeneration… which is very impressive as it is able to do a task that non-specialists physicians find hard,” Judge states.
“What is more amazing is that this model was also trained to be able to recognise conditions that are not actually associated or previously taught as associated with the eye,” he says, adding: “There was a lot of press coverage about this AI technology being able to detect the early changes of Parkinson’s disease, a disease that we did not think had any relation to the eye.”
AI in Ireland’s hospitals
Judge also hails an AI technology called Deep Resolve, which is used in several MRI scanners throughout Irish hospitals. Developed by Siemens, it has created a computer vision model that takes the raw information from the MRI scan and it recreates the final image that the consultant radiologist is able to see.
“Deep Resolve allows consultants to take a quicker scan with a lower quality of raw data and produce the same image,” Judge states. “Currently, this is in use in many hospitals in Ireland and consensus amongst consultants and academics indicates that the AI technology makes the MRI scanning process 73 per cent faster.”
Judge points to another AI technology used in treating Irish patients who have suffered from strokes or suspected strokes, which was developed by a company called Brainomix. The technology uses computer vision AI algorithms to support doctors by providing real-time interpretation of brain scans to help guide treatment and transfer decisions for stroke patients, allowing more patients to get the right treatment in a timely manner.
Judge says that the technology is “exceptional”, in calculating the ASPECTS score (which is a 10-point quantitative topographic CT scan score used in patients with middle cerebral artery stroke), and that the company have “automated this whole process using a computer vision model that is in use in many hospitals in Ireland”.
AI and the physician
As the senior lecturer in the University of Galway master’s programme for applied clinical data analytics, Judge speaks extensively about the future potential for multimodal AI, a new paradigm in which various data types are combined with multiple intelligence processing algorithms to achieve higher performances, in medicine.
He illustrates a clinical scenario in which a patient is admitted to hospital and there are multiple points of data collected about the patient.
“Patients come into hospital and their admission note is written down in an unstructured format, they have blood drawn and blood tests sent off that come back in a structured format, then they have medications prescribed – so the name of the medication, the dose, the frequency in which it is administered, is in a structured format, and then they have imaging done which is unstructured,” Judge explains.
“All of this information was reviewed by a general physician, someone that did not have nephrology training, and they failed to recognise that this patient was at risk from a condition called acute kidney injury (AKI), where the kidney function unfortunately gets worse.”
Judge indicates that in the scenario, no changes were made to the patient’s medications and then the next day it was found that the patient developed AKI. He explains that, as a result, this led to an incorrect decision to stop providing the patient with their medications. One of the medications was furosemide, which ensures that patients can release fluid from their body, thus preventing any exacerbation of heart failure.
“This is where something like single-modal AI will not solve the problem because it does not have access to the full information,” he says.
“Multimodal AI provides extra information to make the really nuanced decision where AI can read the unstructured note, read the structured data and read the image as well. So now in this scenario, nuanced changes to medications are made where all medications are stopped with the exception of furosemide and thus the AKI and heart failure is prevented.”
Summarising, while Judge emphasises that in a context where AI technology has the potential to deliver better patient care, clinicians “need the enabling infrastructure inside the HSE” to utilise the AI tools that are used in other parts of the world, to deliver better patient care in Ireland.
