AI Ireland-Stories from the Source Issue 2: Health

AI-supported health innovation is flourishing in Ireland, enhancing everything from diagnostic accuracy to resource allocation in the health sector. This month's edition of AI Ireland - Stories from the Source comes from five researchers who are reimagining the future of healthcare - from life-saving screening techniques for mothers and babies to AI decision-making that will transform palliative care.

All projects are supported by the Insight Research Ireland Centre for Data Analytics.

Bringing tech sector efficiency to palliative care

My journey from business information systems to digital healthcare was a very personal one. After the birth of my first child, I was surprised to discover how much of my baby’s health data was handed over to external parties. It seemed problematic to me that health data should not be owned by, and used for the direct benefit of, the patient. The systems I had been working on for years could be applied in healthcare to empower patients, improve efficiency, and provide decision support for clinicians.

The new pathway led me to the project I am now working on with the Insight Research Ireland Centre for Data Analytics:?CommPAL (Community PalliAtIve Care).

CommPAL is a web and mobile AI-driven platform aimed at healthcare workers who are strained for time related to patient triaging, daily travel, planning and resource management. Specialist palliative care is a resource intensive service, and nurses are often spread thin trying to decide where they are most needed and when. AI can never replace humans in palliative care, but it can help them to make optimal use of their time and deliver gold standard care to patients.

My PhD was in the field of knowledge management for small to medium software companies and it has become clear to me that services like palliative care can benefit from the same commitment to efficiency, customer-centred service and data-driven decision making.

We are building CommPAL using data captured by nurses on the ground – each patient visit provides new insights into how best to keep people comfortably at home and out of Emergency Departments. We are also consulting with patient advocacy groups – people who have experience of their loved ones accessing palliative care services and who have knowledge to offer about what works.

We were concerned at first that nurses, patients and their families might be resistant to introducing AI into such a person-centred activity as palliative care. We have been surprised to discover that there is support for systems that can reduce the cognitive burden on nurses and potentially allow patients and their families to contribute to the data pipeline.

What does AI-supported palliative care look like? Intelligent triage is a major pillar – taking various data points such as geography, health data, patient availability, and nurse availability and using them to ensure the best decision-support for nurses when determining who is most in need. We are using machine learning techniques, in partnership with the Insight SFI Centre, to build these intelligent triage systems.

Optimisation techniques of the sort used routinely in business systems allow us to support smart decision making in palliative care by recommending the optimum patient caseload for a given day.

The CommPAL team is sensitive to the discomfort that some readers might feel at the comparison between palliative care services and business systems. The first? one is a sensitive and critical, person centred, highly human-skilled activity while the other can sound impersonal and technical. However, in my years as a Professor of Business Information Systems I have always been most interested in the complex interaction component between systems and people. My work has largely focused on socio-technical systems – how humans use and interact with technology and data and how that technology can be developed to keep humans in the loop and support our skills and decisions, not the other way around.

Palliative care services are under pressure and under-resourced all over the world. AI will not bridge that gap and cannot be a substitute for investment in people. However, the enthusiasm I have encountered from the nurses we are working with tells me that data-driven systems are long overdue in palliative care. The need is getting more complex with increases in medication, comorbidity and a range of other factors. AI-driven decision support for nurses is part of the future for palliative care.

CommPAL is due to be tested with real patient data in the coming months. To learn more visit?https://commpal.ai/

Professor Ciara Heavin is Co-Director of Health Information Systems Research Centre, Cork University Business School University College Cork . She is Chair of the University Ethics Committee, University College Cork and lectures in systems analysis and design, data modelling, database development and opportunity assessment and recognition in high technology firms.

AI Premie – life-saving data for mothers and babies

Preeclampsia affects expectant mothers and can cause dangerously high blood pressure. One in every 12 expectant mothers will experience it. Globally, a mother or her baby dies every 60 seconds from the condition. The only cure for preeclampsia is delivery and it is still diagnosed by taking blood pressure and urinalysis.

Diagnosis for the condition hasn’t changed for hundreds of years. It can be difficult for clinicians to identify which cases are likely to become life threatening. When expectant mothers are diagnosed with preeclampsia, they often have to wait out the rest of the pregnancy in a tertiary hospital, which may be far from home.

Professor Patricia Maguire is a biomedical scientist from UCD and has recently joined Insight as a Funded Investigator. Her many years of work with the blood samples of expectant mothers means she has amassed a huge amount of data about this condition.

‘After years of working with Excel spreadsheets I became interested in advanced statistics,’ says Professor Maguire. ‘My team and I started to use Machine and Deep Learning to make sense of the all the data that we have collected.’

Through her work leading the UCD Institute for Discovery, Patricia Maguire learned the value of bringing teams of different disciplines together to tackle global challenges. She realised that it would be useful to assemble a broader range of disciplines to attack the challenge of using health data generated by researchers to find new solutions in healthcare. She pitched the project, called AI PREMie, to SFI (now Research Ireland) and won a special prize in the AI for Social Good Challenge Funding.

AI PREMie seeks to use the biomarkers in the blood of sick expectant mothers to risk stratify preeclamspia. Combined with a range of other patient data, Prof Maguire suspected it could yield a powerful clinical decision support system.

‘My project colleague, Professor Fionnuala Ní áinle (Consultant Haematologist in both the Rotunda and the Mater Hospitals), first approached Maguire to understand if new biomarkers could help in the clinical decision-making process.

At first it was just Maguire and Ní áinle working on the project – ?now the group has 30 members working across the three Dublin maternity hospitals. The AI PREMie team now comprises obstetricians, gynaecologists, midwives, lab managers, haematologists, health economists, biomedical scientists and computer scientists. That’s where the Insight Research Ireland Centre for Data Analytics comes in. Patricia has been working closely with the Director of Insight at UCD, Professor Brian MacNamee to ensure that the data part of the AI PREMie project is sound.

The project has advanced so far in four years that the team is now in a commercialisation fund from Enterprise Ireland.? The product is an algorithm that takes values from both bloods and other variables and suggests a supportive diagnosis to guide the clinician.

Recognising that this approach could be applied across other health fields and that there are many researchers in UCD who likely have vast quantities of useful but unused data, Professor Maguire has since spearheaded the UCD AI Healthcare Hub to identify data-rich projects and guide their researchers in how to collect, process and use the data.

‘There is a massive chasm between the information being collected and the end user, the patient.? Amazing information sits on academics’ computers. We have to get it into the real world. Once it’s out there, how do we integrate it into hospital systems? If we can learn to seamlessly integrate AI PREMie into the Irish hospital system, the approach will work for other health areas too.’

The potential benefit for expectant mothers is considerable – right now sick expectant mothers could spend weeks or even months of their pregnancy on bedrest at home or in one of a handful of city-based hospitals.

‘The neonatal care system in Ireland is excellent,’ says Maguire, ‘and while many women have to spend a lot of time in hospital, mothers and babies rarely die from preeclampsia here. However, in other countries like the US and the developing world the death rates are tragically high. In the US, women of colour are at particular risk.

‘We really haven’t moved on that much in our understanding of the aetiology of preeclamspia. Research into women’s health globally has been very poor. We know aspirin can help if prescribed before 16 weeks of pregnancy but it is only used when a woman has notable risk factors. Unfortunately, however, most of the time a woman will have no obvious risk factors.

‘Insight is moving more into the realm of data-driven health care. Connecting the health researchers with the data scientists is the way forward. We all need to be speaking the same language. Working with Insight we can refine this data that is giving us, finally, new information about diagnosis and treatment and improving quality of life.’

in interview with Louise Holden

Professor Patricia Maguire is Professor of Biochemistry the University College Dublin School of Biomolecular and Biomedical Science, Director of the UCD Institute for Discovery and Scientific Director of UCD Conway Institute SPHERE Research Group and a Funded Investigator with the Insight Research Ireland Centre for Data Analytics

US-style health insurance will accelerate AI uptake in our hospitals

Healthcare systems the world over are under huge pressure as rates of chronic disease rise, along with the cost of drugs and procedures.? Meanwhile skills shortages in key specialisms, particularly in Europe, are creating further challenges to the provision of care. The result? Long waiting lists for? procedures, delayed diagnostics and treatments, poor outcomes for patients.

Ageing populations in the west herald an increase in pressure on limping systems over the coming decades. Change is coming – albeit slowly – and there are two main change agents: insurance companies and digital health development.

Traditional systems of reimbursement by insurance companies go a little something like this. When a patient is admitted to hospital for a procedure, the payout from their insurance company is based not just on the procedure but on the number of nights spent in hospital afterwards and any follow-on care?that might result from poor outcomes.

With this model of reimbursement, a poor outcome is not just disadvantageous to the patient, it puts the insurance company out of pocket too. While healthcare providers want speedy recovery for patients from a professional point of view it has not, until now, been a financial consideration. If the patient has to spend ten nights in a hospital bed, that bed will be paid for.

This model is starting to change. In the US, the model now is to pay a fixed sum for a procedure and let the hospital find ways to make sure that money covers the entire cost of the treatment. In that instance, ensuring that the patient is discharged quickly, and does not end up back again because of a poor outcome, is a matter of financial life or death for the healthcare provider.

That’s where digital health and AI are starting to play a role, at least in the US – more slowly in Europe. Hospital administrators are motivated now to seek ‘value-based’ systems and they are increasingly turning to medtech companies to provide these. A value-based system is designed to optimise outcomes for patients by improving and shortening the clinical pathway.

There are four main ways that AI can support this – I call them the four Ps: prediction, prevention, personalisation and participation.

Prediction is easy to understand: early diagnosis of disease results in fewer interventions and more impactful ones. AI is a very sophisticated spotter of anomalies, and with expert human oversight, can deliver more accurate screening results.

Prevention, again, is straightforward. AI’s role here is a holistic one. The rapid development of wearable and implantable sensors means that we have the technology to monitor our physical biomarkers, such as blood pressure, frequently and outside clinical settings. Many of us are already monitoring ourselves via smartwatches. This info can inform local health clinicians and flag markers that might need attention before disease sets in.

Once disease is present and treatment required, personalisation is the pathway to better outcomes. We are all different, and require different dosages, interventions, therapies. Again, digital health and AI have a role to play here, employing sensors to supply personalised information to clinicians about a patient’s reaction to treatment. Sensor tech does not just measure?traditional physical biomarkers such as heart rate and blood pressure, it is increasingly effective at measuring biochemical markers (for cancer, for example) and electrophysiological markers (cardio). A ‘closed-loop’ system is one where AI-on-the-edge (a sensor that has the capacity to perform its own AI-decision making) is attuned to the individual and capable of informing decision-making without needing?leave the system (to the cloud, for example) at all.

The final P, participatory, empowers the patient to monitor their own progress via AI-enabled sensors, and adjust their behaviours accordingly. Recovery at home with a supportive digital health system feeding into the dashboard of a local healthcare provider is the holy grail here.

Each of these digital and AI interventions has the capacity to shorten the healing pathway and reduce costs. Add to these the rapid development of AI-powered surgical tools, robotics, augmented reality in theatres and the scope for remote treatment of patients by specialists anywhere in the world, and it all adds up to long term savings. Little wonder then that US hospitals are proactively partnering with medtech companies to help them with, or in some cases take over, aspects of patient care.

It’s a slower transition in Europe for a number of reasons but here are two. Centralised reimbursement in the US means that the changing payment protocols are impacting right across the system. In Europe the insurance sector is far more fragmented; system-wide change of the sort that drives investment decisions is more challenging.

The other obstacle is an important one – data protection regulation in Europe is tighter and when it comes to personal data it doesn’t get much more personal than health data. GDPR regulation is a valuable protective principle here, as well as the incoming European AI regulation.

How soon will we see change? A €30 catheter that isn’t functioning optimally should be replaced by a smart catheter that costs a great deal more. Right now, when making investment decisions, hospitals administrators have bigger fish to fry. However, once the reimbursement models start to change, that cheap catheter will start to look like a financial liability rather than a money saver. That works for everyone.

By Dr Paul Galvin, ?head of the ICT for Health Strategic Programmes, head of the Life Sciences Interface Group, head of the Bioelectronics Cluster at the Tyndall National Institute and Funded Investigator with the Insight SFI Research Centre for Data Analytics

Using AI to predict loneliness and depression

Smartphones and wearable devices? are equipped with powerful sensors; such as accelerometers, GPS and gyroscopes; that can track our daily routines and behaviours passively. For example, accelerometers measure movement and physical activity while GPS can track locations and travel patterns. Collectively, these sensors provide data on how active we are, where we spend our time, how much we sleep and even social interactions – like how often we are near other people or in social gatherings –? all without requiring us to actively input any information.

This research focuses on using this passive data to detect signs of mental health issues like loneliness or depression. The idea is simple but powerful: as our devices collect data in the background, we also ask users how they are feeling through standard mental health surveys. By comparing their actual responses with the data collected from sensors, we train artificial intelligence models to recognise early signs of mental health conditions. The main goal is to spot these issues early, so we can step in before things get worse and help improve mental health. This approach could one day lead to mental health support that is personalised, timely and seamlessly integrated into everyday life

Malik Qirtas is a postdoctoral researcher in Machine Learning at Insight University College Cork

Galway researchers teaching AI to show medical expertise

Large Language Models like ChatGPT have become household names, capable of generating text and answering general questions. However, when it comes to more technical queries—especially in fields like healthcare – these models often struggle. The problem lies in their general training: while they know a little bit about a lot of things, they don’t have the specialised knowledge to answer domain-specific questions accurately. This is where the Insight team’s work comes into play.

Researchers at the Insight Research Ireland Centre for Data Analytics, University of Galway, are working to make AI smarter and more accurate in answering complex, specialised questions. Professor Paul Buitelaar, Ghanshyam Verma and Devishree Pillai from the Insight Centre Galway, in collaboration with Dr Bogdan E. Sacaleanu from Accenture Labs Dublin, are making significant contributions in the field of Natural Language Processing and Large Language Model (LLM)-based question-answering by addressing one of the key limitations of current LLMs: their inability to provide precise answers in fields that require deep, domain-specific knowledge such as healthcare.

LLMs like ChatGPT have become household names, capable of generating text and answering general questions. However, when it comes to more technical queries—especially in fields like healthcare – these models often struggle. The problem lies in their general training: while they know a little bit about a lot of things, they don’t have the specialised knowledge to answer domain-specific questions accurately. This is where the Insight team’s work comes into play.

The Insight team developed an approach called SKnowGPT, designed to enhance LLMs ability to handle domain-specific questions by integrating them with Knowledge Graphs (KGs). KGs are specialised databases that store information in a structured way, such as medical knowledge about diseases, symptoms, treatments and tests. SKnowGPT doesn’t just add knowledge from a KG; it also carefully filters it to remove irrelevant information, ensuring that only the most useful data is used to answer the question.

For example, in the field of medicine, where accuracy is paramount, irrelevant information can confuse the LLM and lead to incorrect answers. SKnowGPT tackles this by finding the most relevant knowledge for a given question, while also pruning away unnecessary data to avoid distractions. This dual-filtering process makes the LLM much more reliable when answering complex medical queries.

Another contribution of this project is the creation of a KG called DisTreatKG by enhancing an existing KG (EMCKG). This expanded KG enables the LLM to draw on a broader range of diseases, symptoms and treatments, making it even more effective in providing accurate answers.

The research has far-reaching implications for improving domain-specific natural language understanding and question-answering systems. Not only does SKnowGPT outperform existing methods like MindMap, but it also demonstrates that smaller LLMs—like Mixtral—can perform just as well as larger, resource-heavy models such as GPT-3.5. This makes the technology both accessible and scalable, opening the door for its use in industries where cost and efficiency are key concerns.

The Insight team’s work is a prime example of how academia and industry can collaborate to solve real-world problems. By releasing their code and datasets to the public, the researchers have ensured that others can build on their findings, accelerating progress in LLMs ability to handle complex, domain-specific inquiries.

This research not only highlights the Insight Centre’s commitment to cutting-edge innovation but also promises to revolutionise how AI is used in high-stakes environments where precision and reliability are non-negotiable. The impact of SKnowGPT and its KG-driven approach will likely extend far beyond healthcare, influencing fields like finance, law and any domain that requires expert-level answers.

Paul Buitelaar is Professor in Data Analytics and Deputy Director of the Data Science Institute at the 爱尔兰国立高威大学 where he leads a team in Natural Language Processing. He is co-Director of the Research Ireland Centre for Research Training in AI and co-PI of the Insight Research Ireland Centre for Data Analytics.