by Artificial intelligence has already begun to be used in medicine. For example, in some cases it is a valuable aid in the diagnosis of diseases because, if correctly practiced, it can “read” medical images with great precision, distinguishing the presence or absence of tissues pathological.
However, it is much more difficult to train an AI system to make the right treatment decisions and even in the most difficult conditions – inside intensive care units (ICUs) where critically ill patients are cared for and whose life is often at stake. This is exactly what an AI system developed by researchers from the Vienna University of Technology (TU Wien) has achieved in collaboration with experts from the Medical University of Vienna. , according to their publication in the scientific journal “Journal of Clinical Medicine”.
When the computer chooses the remedy
Using a wide range of data obtained from the intensive care units of several hospitals, the researchers developed the AI system which can make recommendations on the therapeutic approach to be taken in critically ill patients with sepsis. The analysis showed that artificial intelligence was able to make better decisions than humans. However, as the researchers point out, it is very important that the legal ramifications of using these methods within ICUs are also discussed.
Make better decisions
“In the ICU, a lot of data is collected around the clock. Patients are monitored with medical devices around the clock. We wanted to explore if this data could be used in an even better way than it is today. .” says the teacher Clement Heitzinger from the Institute for Analysis and Informatics of TU Wien and described that medical staff in intensive care make decisions about the treatment of patients according to specific rules. Most of the time, doctors know very well what parameters to take into account to provide the best possible care to their patients. However, Artificial Intelligence can take into account many more parameters than humans, which in some cases leads it to make better decisions.
reward and punishment
Professor Heitzinger explained that the research team used a form of machine learning called reinforcement learning. “This type of learning does not consist of simple categorization – for example, separating a large number of images into those that show cancerous tumors and those that do not. It refers to an ever-changing assessment of how a patient should progress. Mathematically it is different. Little research has been done in this specific area in medicine.”
Essentially, the computer makes its own decisions, depending on whether it is rewarded or punished. If the patient is progressing well, he receives a “virtual award” while on the other hand, if the patient’s condition worsens or he dies, the AI system is “punished”. What is the price ? To be able to move on to more… action. In this way, a large amount of medical data is analyzed to automatically determine a treatment strategy that will have the best possible results.
RN-doctors: tick 1
“Sepsis is one of the most common causes of death in intensive care units and poses a huge challenge for doctors and hospitals, as its early detection and treatment is vital for patient survival,” noted Professor Oliver Kimberger of the Medical University of Vienna and added that “to date there have been few achievements in this area, which makes the need for new treatments urgent. For this reason, it is very interesting to study if artificial intelligence can contribute to the improvement of medical care and treatment. The use of machine learning models and other AI technologies is an opportunity to improve diagnosis and treatment sepsis, thereby increasing the chances of patient survival.”
Expert analysis in Austria showed that artificial intelligence outperforms humans in this area: “Cure rates are now higher using the AI strategy than if only human decisions were made. In one of our studies, cure rates – defined as avoiding mortality over 90 days – increased by around 3%, to reach around 88% thanks to AI. says Dr. Heitzinger.
Valuable help around the clock at the patient’s bedside
As the professor pointed out, this does not mean that medical decisions in ICUs should be made exclusively by computer. However, artificial intelligence could be an important aid that will constantly lead patients and medical staff can consult it and evaluate its suggestions. Such systems will also be very useful in the training of students and young doctors.
Upcoming legal issues
“Of course, all of this raises important questions, mainly of a legal nature” notes Professor Heitzinger. “For example, who will be responsible if the AI makes a medical error? But there is also the opposite problem: what if the AI had made the right decision, but the doctor had chosen a different therapeutic approach, resulting in a poor outcome for the patient? In such a case, should the doctor be blamed because he should have trusted the artificial intelligence or is he entitled to ignore its advice?” The teacher concluded by saying how “Our results show that AI can be used successfully in clinical practice based on existing technology – however, the social context and legal rules governing its use in ICUs need to be established.”
