There is no way around the topic of AI and its impact on the most diverse areas of life. Particularly since the availability of the ChatGPT programme, which according to a study is even capable of passing the US medical exam USMLE, the debate on this has gained enormous momentum. The use of artificial intelligence in medicine is not fundamentally new - think of ECG analyses, for example. All in all, however, physicians are likely to encounter the topic more and more frequently in their daily work, Prof. Martin Cowie (ESC) is convinced.
But what exactly is meant by the term "artificial intelligence"? On ESC TV Today, Prof. Cowie and presenter Prof. Susanna Price discuss the topic "Artificial Intelligence in Cardiology - What to Expect".1 Cowie summarises: AI is understood to mean that a computer behaves as if it possessed human intelligence. Just like a human being, the computer is then capable of a wide variety of things that are an integral part of a doctor's daily work: These include image recognition, the interpretation of data, decision-making on the basis of a lot of information, but also specialist discussions in every conceivable direction.
In the meantime, computer speed has advanced so far, that much of this is actually feasible. The German Society for Haematology and Oncology (DGHO) is also taking note of such technical developments. In a recent paper it states: "Recent developments, driven in particular by the availability of data sets and computing power, have led to the view that this rapidly developing field has the potential to change our society. "2
But can such changes lead to severe changes in the medical workforce? Prof. Cowie gives the all-clear: the concern that artificial intelligence could replace physicians in the future is exaggerated.1 Nevertheless, the physician emphasises that:
"Physicians who use artificial intelligence will replace physicians who don't. This is especially true when the medical field relies on the interpretation of data and images."
For Prof. Cowie, AI is first and foremost a diligent assistant. It could be used primarily to reduce burdens that result from repetitive, tedious tasks that still need to be done with precision - such as calculating an ejection fraction from MRI or echo. "The AI is enthusiastic even at three o'clock in the morning, it doesn't need a coffee break," says Prof. Cowie. At the end of the day, he says, AI should master exactly the tasks it is best at - and leave to humans the tasks for which humans are best suited.
Prof. Cowie sees particular potential for application in everyday practice in the area of risk stratification. Instead of the Framingham Risk Score calculator, for example, a huge wealth of information could be used to obtain better calibrated estimates of the risk of individual patients. Doctors could therefore make better decisions together with artificial intelligence.
The DGHO argues in its paper that AI in general can be used for three practical medical purposes: (individualised) clinical care, research and education.2 First, AI has the potential to be integrated into clinical routines and used as a tool for daily clinical practice. For example, AI approaches and methods could be used to identify patterns in past cases that could help predict how well a particular patient will respond to a particular treatment. Secondly, AI could be used as a research tool to gain new scientific insights from clinical data - for example, about new disease entities or pathomechanisms. Thirdly, AI is also a suitable tool for medical education, for example through data synthesis for educational purposes. In an ageing population, cancer is becoming more common, so more trained staff are needed to care for cancer patients. AI could potentially help to train these experts - but this aspect is still a young field in haematology and oncology.
The DGHO names six specific future potential applications for the fields of haematology and oncology:
But who is ultimately responsible when AI is used to make decisions? Prof. Cowie from the ESC has a clear position on this: "In my opinion, the physician is ultimately responsible for the diagnosis."1 In the end, the processes are the same as always; the use of AI merely gives physicians more support options. Therefore, it is always important to document on which basis decisions were made, which tools were used - and also to clearly state which uncertainties exist.
Of course, hardly any AI debate currently avoids the topic of ChatGPT. Prof. Price asked: Is this a playground or does the chatbot provide practical value for medical professionals? Prof. Cowie is sceptical. The programme is technically capable of submitting an abstract for the ESC congress, for example, that just misses the mark. However, ChatGPT still has many problems: "It tends to behave like a bad-tempered teenager in terms of its answers. Filters still have to be set up to prevent the programme from saying inappropriate things. In addition, answers via the chatbot are not always factually correct. Especially with regard to medical studies or medical exams, one would also have to think concretely about how to test the competence of an individual instead of ChatGPT. Apart from regulations, according to Prof. Cowie, we should also think about the fields in which particularly good services could be provided to society.
Now it's your turn: What is your opinion on the topic of "artificial intelligence in medicine"? What are your expectations or concerns? We welcome you to share these in the comments section below.