Predicting clinical risk: Artificial Intelligence at the service of patients

Andrea Pazienza Innovation Lab Exprivia

The world is facing a major medical problem, with an increasing number of patients and not enough doctors to treat them. Can Artificial Intelligence be the cure?

The research carried out by theExprivia Innovation Lab and published in the international Evolving Systemsjournal analyses how to exploit the best Machine Learning model to predict the clinical risk category of patients with a limited number of vital parameters.

AI has the power to transform the way in which patients are diagnosed and treated, and to make the testing of new medical procedures more efficient and effective. It is the symbiotic relationship between man and machine that brings the magic of AI to medicine in a meaningful way.

WE NOW HAVE THE TOOLS TO MAKE A HUGE DIFFERENCE IN HUMAN LIFE.

By 2027, the value of AI in the healthcare market is expected to increase eightfold compared to 2022 (The Economist, https://www.economist.com/films/2022/02/15/the-future-of-medical-ai).

It is no longer science fiction; AI is transforming Healthcare.

AI is becoming more sophisticated at doing what humans do, but in a more efficient, quicker, and cheaper way. The potential for AI in the healthcare sector is vast: going from medical imaging analysis to precision medicine, with telemedicine and predictive diagnostic analysis in between. Just as in our daily lives, AI is increasingly part of our healthcare ecosystem.

In particular, the evaluation of clinical risk is a practice adopted in the healthcare sector to identify possible early interventions, in order to analyse quickly the clinical condition of patients, especially in accident and emergency units or as part of intensive care.

The possibility of setting up an Early Warning Score (EWS) system signalling the onset of pathological events or serious conditions can therefore be useful for doctors and healthcare professionals in order to give an overview of a patient's condition, obtain information for more effective treatment, and choose the most effective treatment. In this particular scenario, evaluating a patient's clinical risk can be considered a predictive analysistask.

AI-based approaches can be powerful tools as a decision support system when available information is uncertain.

 

The research carried out by theExprivia Innovation Labcontains the results of a project aimed at developing a system on Edge capable of providing an early clinical risk evaluation even in the event of a reduced number of vital parameters.

The Internet of Medical Things in support of Edge Computing

The new generation of AI capabilities will allow even more health data to be integrated into these systems, helping to detect patterns pointing to potential problems. One of the biggest potential benefits of AI is to help people to stay healthy so they don't need a doctor, or at least not that often. The use of IA andInternet of Medical Things (IoMT) in consumer health applications is already helping people.

As part of this concept, connected wearable sensors enabled by IoMT, with the aim of increasing the reliability of diagnoses, ensure continuous data collection and analysis adapting to the evolution of a patient's condition.

Consequently, in a clinical and operational context, Exprivia has set itself the research objective of developing integrated solutions for continuous care in which AI and IoMT are used at the Edge of the network, with a person-centred approach that adapts and evolves according to the needs of healthcare professionals, invoking a “system in constant evolution”.

The solution conceived by Exprivia ponders an Edge device, connected to one or more wearable medical devices via IoMT, which exploits the best performing Machine Learning (ML) model to predict clinical risk similar to an EWS, according to:

  • the patient's available vital parameters, appropriately detected by one or more medical devices;
  • a specific EWS protocol with three levels of risk (such as NEWS2[1], i.e. low-medium-high risk) guiding the frequency of monitoring and the level of care;
  • the patient's conditions mapped based on two different risk settings that are easier to classify, in order to distinguish correctly between a triage setting and an emergency scenario.

Exprivia's innovative approach in order to predict evolving and explainable clinical risk

In the readily available Paper, published in the international Evolving Systems journal (Evolving and explainable clinical risk assessment at the edge | SpringerLink), Exprivia carries out an in-depth study aimed at analysing the general conditions for exploiting the most effective ML model, both in terms of performance metrics and computing efficiency at the Edge, capable of predicting the clinical risk category of patients monitored in a particular condition in which a limited number of vital parameters are available.

Lastly, while the integration of AI and IoMT supports the development of a wide range of potential applications in theeHealtharea, at the same time an ethical issue arises concerning the lack of transparency and clarification in predictive and algorithmic black-box procedures. For these reasons, knowledge representation techniqueswith a Ontology Web Language (OWL) taxonomy are used to make the predictive result more interpretable, in a perspective referencing the Semantic Web of Things (SWoT). In this way, explanations integrated with the clinical knowledge of healthcare professionals can be used to give rise to further valuable information and to guide the treatment course even more effectively.

With the power of AI, we can pursue the unknown. And we can be sure to find something that no one expected.

The potential of AI in healthcare and life sciences runs deep. It has the capability to help doctors and researchers prevent disease, speed up recovery and save lives by unlocking complex data. It can also free them up from ordinary tasks, so they can focus on their patients or research.

 

 

[1] https://www.mdcalc.com/national-early-warning-score-news-2