do we need it and why?

According to prof. Marek Gzik, Dean of the Faculty of Biomedical Engineering at the Silesian University of Technology, the perception of man, even the best specialist, is not able to comprehensively cover the available and constantly expanding medical knowledge of thousands of research centers and scientists. How, using it, to establish an accurate diagnosis and prescribe an effective treatment for a specific patient? Can artificial intelligence help with this?

Piotr Buszman, vice director of preclinical research, head of the experimental laboratory at the Center for Research and Development at the American Heart of Poland, argues that when talking about AI in medicine, there is a fundamental question to ask: why do we want and why should we introduce artificial intelligence in medicine?

– AI has its application here. It is true that it is limited for the moment, but it enters clinical practice especially where we know that we cannot cope. Particularly in imaging – radiology, histopathology – lists prof. Buszman and adds that the artificial intelligence algorithm allows you to effectively analyze specific areas and indicate which ones the doctor should pay attention to.

– Note however that currently no AI yet, and some algorithms help us in practice, especially in major screening tests for which we lack human resources. This is already happening – says the expert.

As an example, it gives “kiosks” in which patients are asked – according to specific algorithms – a series of questions, and following the analysis of the answers, the patients are directed to the appropriate diagnostic path .

– For now, AI supports doctors, especially where their access is limited – emphasizes prof. Buszman and points out that AI may prove indispensable in the future, if only for this reason.

This tool has its limits
Teacher. Zbigniew Nawrat, Creative Director of the Foundation for the Development of Cardiac Surgery prof. Zbigniew Religa in Zabrze, the president of the International Association of Medical Robotics, talks very carefully about the great hopes of AI in medicine.

As an example, he mentions the IBM Watson system, hailed as a great achievement in AI diagnostics. For a huge amount of money, it was introduced in more than a dozen hospitals, but it turned out that it can diagnose a high level of efficiency only in a few oncological diseases.

– Watson analyzes about 200,000 different types of information and the result is a diagnosis, whereas doctors can often do this based on a conversation with the patient. So far, it is difficult to highlight the great successes of AI in medicine. They are visible, for example, in games where the artificial intelligence is doing well. However, this applies to games where there are multiple immutable rules. People’s reality, however, is different – emphasizes prof. Nawrat.

According to him, AI, which is a department of robotics, has the same problems as robotics itself in applications for various fields. He must learn the place where he will perform a specific task. – AI will not solve many of our problems. However, we need it, especially in areas where the doctor does not have access to various data. The less invasive the surgery, the more we have to rely on processors and diagnostic algorithms – emphasizes Prof. Nawrat.

It also points to another reason why we will have to rely more and more on AI: our own intelligence is declining. It has been developed for centuries and has been from generation to generation, but some solutions, such as the popularization of calculators, have reduced the intelligence of people in the field of arithmetic tasks. We no longer use our brain as before, as we use tools to help our memory and our calculations.

– AI, like robots – emphasizes prof. Nawrat – these are just tools with significant limitations. It will be up to doctors how they will be used and how effective they will be.

There are many pitfalls
According to prof. Piotr Sankowski from the Institute of Informatics at the University of Warsaw, member of the Board of the National Center for Research and Development, even though today we are creating artificial intelligence and building neural networks, we do not understand often not how AI actually works.

– The book’s definition of artificial intelligence translates it as an imitation of the way a person thinks. Currently, however, this AI, which is enjoying spectacular success, such as Deep Blue beating Kasparov in chess, does not think like a human. We have examples where AI and machine learning triumph, but looking at the processes taking place inside, we are not able to understand everything. This also applies to deep neural networks – claims prof. Sankowski.

– Therefore, there is a lot of work ahead of researchers and scientists. The same applies to the use of artificial intelligence in medicine – the specialist points out.

As he argues, artificial intelligence researchers face the challenge of creating algorithms that we can understand. When people started using deep machine learning algorithms, they started realizing their ignorance. It turned out, for example, that the AI ​​distinguishes a cat from a dog not on the basis of what it sees, but for example because the photo of the cat is usually taken indoors and the dog – on the grass.

– There are many such traps. A famous example is the cancer recognition algorithm developed at Stanford University. He did not learn to recognize them, but on which machine the photo was taken. It is therefore obvious that the tools at our disposal cannot make medical decisions – emphasizes prof. Sankowski.

Such support will be useful
As Marcin Bruszewski, general manager of Philips Health Tech for Poland and the Baltic States, asserts, every patient would like to see a doctor with extensive experience, which, however, relies on tools using artificial intelligence.

– Such a combination gives the greatest confidence to make the right decisions and prescribe the most effective treatment. The ideal would be the situation when an experienced doctor has time for the patient and can analyze his case for a long time. Unfortunately, the world is not perfect – there is a shortage of doctors and nurses. Technology comes to our rescue – learning machines, algorithms, statistics – in general, prediction of the implementation of AI-based activities, explains the director of Philips Health Tech.

According to Dariusz Wiśniewski, COO at BrainScan, which has developed AI-based software for automatic analysis of brain CT scans, machine learning-based artificial intelligence will not be able to make decisions by it -even for a long time. It is only a tool, but a desirable tool and the only one that can help in the event of a shortage of health care. – AI by accelerating decision-making processes can fill this lack of personnel – he underlines.

Marek Witulski, director of diagnostic imaging and advanced therapy departments at Siemens Healthcare, explains that there are many obstacles accompanying the development of artificial intelligence in medicine. They can be divided into two zones. The first is the development of software, AI algorithms and databases. The second is legislation.

– Current regulations do not allow us to set up self-learning tools. We need to teach the application something and then freeze its knowledge state. Only then can it be allowed into the medical market, says director Witulski, adding that the biggest hurdles, however, are the psychological ones.

All statements were recorded on June 30, 2020 during the “Artificial Intelligence in Medicine” session as part of the 5 Online Health Challenges Congress.

Learn more about:

Leave a Comment