Czech team develops AI better than doctors at identifying lesions in lungs
Last week, a Czech team presented a study at the European Radiological Congress in Vienna that showed an AI that they developed called Carebot is more accurate than human radiologists at detecting suspicious lesions in lung X-rays.
Artificial Intelligence is already better than humans at playing chess and Go, analysing big data, and with the advent of ChatGPT, the new chatbot by OpenAI that everyone is talking about, answering questions about almost anything, it seems. As if that wasn’t enough, there is now yet another area where the machines are beating us – according to Daniel Kvak, co-founder of Carebot, AI is better at identifying suspicious lesions in lung X-rays than human doctors.
“Various studies indicate that the inaccuracy of radiological description is somewhere between 25 and 30 percent in standard clinical practice. And chest X-rays are precisely where errors unfortunately occur most often. In cooperation with the Masaryk Institute of Oncology, we ran a study where we compared the success rate of artificial intelligence with the success rate of five radiologists. Our AI correctly picked up the lesions 91 percent of the time, compared to 29 to 81 percent for the radiologists.”
Together with co-founder Matěj Misař, Kvak has developed an AI called Carebot that has been trained on tens of thousands of anonymised images from hospitals and medical facilities around the world. It can evaluate a chest X-ray with high accuracy and even create a brief medical report. It is currently being tested by doctors at a hospital in Havířov (incidentally Czechia’s youngest city) in the far east of the country and has been for about half a year, and radiologist Jakub Dandár says it can already be relied on for certain things.
"For example, it can distinguish chest X-rays that are completely negative, i.e. without any suspicious findings. And on the contrary, it can also distinguish images that basically require our increased attention."
If artificial intelligence were to be applied in clinical practice, it could help the currently overburdened healthcare system, says Anna Chromcová, who also participated in the development of Carebot. As a radiologist, she knows from her own experience that doctors are often overwhelmed by large numbers of X-ray images.
"Even though they give it their full attention, at some point you can miss some things because your eyes are already tired. So doctors welcome the fact that artificial intelligence is a tool that will give them confidence that they have evaluated the image correctly and that they haven’t missed anything."
Matěj Misař, the other half of the duo behind Carebot, says that the AI is not intended to replace doctors, but rather to give confidence to both doctors and patients that no matter the circumstances, the evaluation of their X-ray will be accurate and nothing will be missed.
"We are talking about improving the quality of examinations, so both the patient and the doctor know that they will always receive the same quality diagnosis, whether they are in a large hospital in Prague or in a small hospital at one in the morning. Because artificial intelligence doesn’t get tired. The best combination out there is a doctor working together with AI."
If Carebot receives medical device certification, which could happen by the end of this year, then it can be used in hospitals across Europe and around the world.