Would you like to introduce yourself.
I am a freelance science journalist, accredited as a member of UNAMSI (Unione NAzionale Medico Scientifica di Informazione) association. The topics I cover are related to health, healthcare and wellness, with a strong passion for innovations being applied to the world of health and healthcare.
I follow a lot of national and international conferences and webinars dealing with these topics to keep up to date with digital innovation in healthcare.
What is the role of AI in clinical research?
The advent of digital health as an innovation also in the field of clinical and health research has brought a real revolution with a variety of changes, let’s analyze them through some steps.
According to the scientists and researchers I have had the opportunity to interview over the past few years, the areas in which the use of AI seems most promising involve physician diagnosis, prognosis, and research (such as in epidemiology).
The development of AI has taken place very quickly, beginning in the second half of the twentieth century, advances in computing have been resounding. The beginnings date back even to the first rudimentary “intelligent” machine that was developed by Alan Turing, with which he solved and decrypted the famous “Enigma,” a machine used in World War II by the Nazis to send encrypted messages. The same Turing who was credited with the artificial intelligence test (Turing test) that tests whether, through a chat conversation, a computer can convince a human that it is a human.
But, to return to the question, in the context of research, it must be emphasized how important the data and the quality of the data are.
Databases used to be used; differently, with the current revolution in computing, the data that can be analyzed with artificial intelligence can also be unstructured, such as an image, a pdf or written text are.
A very practical example is the use of the smartwatch: a new technology from which data can be derived that can be used by AI or any data that can also be collected from the patient himself (real-word data).
The key thing is that, the data being taken into account, must be of quality. Otherwise, the AI response will be a low-quality result.
Method used by AI (mixed- qualitative/quantitative)?
The data we submit to the AI must be of high quality in order to have a good response; Another aspect to take into consideration is that the AI is software and applications subjected to legal and regulatory aspects.
In May 2024, the Council of the European Union passed the AI Act, a law on artificial intelligence that aims to regulate the use of AI with a risk-based approach. As a result, companies that produce software and applications in healthcare must also be subject to stricter regulation to ensure safety, quality and effectiveness.
A software in healthcare that is proposed as a medical device must have IIA, IIB or 3 as its risk class to meet the regulation.
A few years ago medical devices could have a lower risk class such as 1. Risk class IIA and above requires extensive, randomized, controlled clinical trials.
The disadvantages of AI in research, if any, are what?
One issue with the use of AI is deskilling: when a physician/researcher uses an artificial intelligence tool, if he or she knows that the accuracy of the data provided by the AI is very high and there are many studies that assume the highest accuracy, he or she automatically relies on the machine and does not pause to critically read the outputs that the AI may give. This can create a kind of de-skilling of the health professional, reasoning that the use of AI should always be with a high level of critical thinking and the final word should be the physician’s.
The ethical implications?
When we talk about people’s health we are also talking about ethics, so with the advent of artificial intelligence there has to be legislation that also starts with human ethical concepts.
Father Paolo Benanti, Professor at the Pontifical Gregorian University, the only Italian member of the United Nations Committee onArtificial Intelligence, talks about Algoretic that is: the study of the ethical implications related to the application of algorithms, among the ethical issues emerges, for example, the lack of awareness of users with respect to the processing of personal data. In addition, in my opinion it is an ethical problem not being able to know the path the AI takes to arrive at the result, that is, the problem of non-transparency of the so-called black box algorithm. The expression black box is now commonly used to refer to cases in which algorithms and computational methods are used whose mechanisms of operation are not well known.
The social and ethical implications of AI and algorithms make it necessary to have both an algor-ethics and a governance of these invisible structures that increasingly regulate our world in order to avoid inhumane forms of what we might call an algo-cracy ( Cite. P.Paolo Benanti).
Gen AI aka Chatgpt?
Generative AI, like chatgpt and its counterparts, will have to be very guarded and monitored because it is difficult to control, which makes it a very useful technology, but also a risky one.
Nonetheless, it remains a very interesting type of AI.
In Italy, chatgpt had been made available for network use in November 2022 but by January 2023 it had already been blocked and made unusable because it did not comply with Italian privacy regulations.
In the medical field, data are to be considered particularly sensitive; they are all allocated, often anonymized, in the large databases of hospitals or health care facilities, which, are now increasingly the target of ransomware, or data theft for ransom, ransom in fact means ransom. It therefore becomes necessary to fortify cyber security.
Pooling research data, a step toward data as a common good
Once data are received anonymously they could be used for research purposes, but so many people do not know this and do not give this option at the time of informed consent.
The concept that I would like to see understood is that making one’s data available in complete anonymity can be key to incentivizing and speeding up research and considering health data as a common good.
Would you like to add more…
Another theme inherent in the link between drug research and Artificial Intelligence. AI in drug research can do a lot. There is, in Bologna, Italy, the world’s fourth HPC (High Performance Computing) supercomputer Leonardo that with the help of algorithm systems and process speed can find new molecules and verify new interactions between them. Leonardo develops 255 petaflops of power (1 petaflop is ) and petaflop is the unit of measurement of computing power.
This aspect of drug research does not often recur in disclosure.
However, even for the use of these HPCs they require huge amounts of energy, which is necessary for its successful execution; these supercomputers are extremely energy-consuming.
Digitization in health care must also strive for sustainability, in terms of carbon footprint, as much as possible.