A robot’s function in medicine: to replace an expert or to remain an assistant?

Medicine stands among the many fields where artificial intelligence is taking small steps towards making the lives of people easier. A robot, able to identify pathologies, can save the time of doctors and also aid in curing patients.

Gediminas Pekšys talked to us about which tasks that used to be performed by people in a medical laboratory were replaced by artificial intelligence and whether we should expect a breakthrough in this field soon. He is one of the founders of a startup that aimed to create an instrument that could be used for radiology diagnostics.

–  You’re a founder of the “Oxipit” startup, a company that mainly works with artificial intelligence. Can you briefly describe the products that “Oxipit” makes? What kind of experts did you require that made everything possible?

–  In two years our company has crafted an instrument for radiologists using the principle of artificial neural networks. This instrument automatically interprets abdominal X-rays and draws conclusions from them. It’s capable of recognising and marking possible pathologies as well as identifying more than 70 different radiological signs or diseases that should then be looked into by radiologists.

Creating an instrument this complex requires a team that possesses a wide spectrum of skills. Besides having researchers of artificial intelligence, a strong base of knowledge about radiology was crucial as well. I think that coincidences are important for building a strong team. At the time when I met the other “Oxipit” founders we were working in different companies and attended various events. All of us appeared in the right place at the right time, but for that to happen we had to demonstrate enthusiasm, effort and participation in the artificial intelligence community.

–  Artificial intelligence was created for the purpose of making the lives of people easier. Its use in diagnostics is probably unknown to many. How did you come up with the idea to use artificial intelligence in the medical field? Is it an improvement of a principle that you have encountered before, or a whole new idea?

–  This idea was born after we noticed a global lack of radiologists. The fact that the need for research is rising at a greater speed than the number of specialists is an alarming one. For example, it was calculated that in Great Britain, 7 out of 10 patients have unevaluated abdominal radiological scans. Not only that, but the patients have to wait for months to get the results.

The members of our team have wanted to find a solution to this problem for a while. Similar attempts have been made before, but only with individual cases of diagnosis and they weren’t exactly successful. The progress we have made in the past several years using artificial intelligence for images and the experience we have gained allows us to offer a more capable assistant for radiologists.

–  What does the medical community think about the use of artificial intelligence in the diagnostic field? What about the general public?

–  Everyone has their own opinion. People who have dealt with older systems and have seen their flaws tend to be more sceptical about it, but even a part of the said systems found audiences that they appealed to. Generally it’s accepted that the importance of the artificial intelligence field is growing. In the last two years the attention and support for it has increased tens of times in various conferences and exhibitions.

As for the general public, I think that this topic isn’t discussed very deeply yet, but we’re already encountering artificial intelligence in our everyday lives – in some places automatic analysis of cardiograms is being used, then there’s also radiologists using half-automated decisions as a quicker way to submit results of computed tomography or magnetic resonance scans.

–  The use of artificial intelligence in medicine undoubtedly raises ethical questions. Who should be held accountable for the mistake of a robot?

–  Algorithmic automation allows us to avoid most mistakes, but several of them can still slip through unnoticed. We’re trying to avoid this by creating an artificial intelligence assistant that makes a doctor’s job easier and quicker without taking away the responsibility of their work.

–  In your opinion, is it possible that in the future artificial intelligence will be able to take over most of the jobs that are performed by people right now, as in analysis of various tests and diagnosis of diseases?

–  Artificial intelligence is definitely not ready to make the really clever decisions. There’s simply a lack of several scientific breakthroughs that should fill the gaps there, if that’s even possible in the first place. Even repetitive daily activities can’t be automated since there are too many differences between the IT systems of hospitals. However, radiological scans are an exception in this case. Centralised systems and strong standards are steps that need to be taken to make the diagnosis and prevention of various diseases easier, but it’s hardly likely that it would lead to replacing people. Altering jobs or creating new work positions sounds more realistic.