Radiologist Josep Monuera: “AI tools do not replace the doctor, they empower him” | Health and wellness

Artificial intelligence (AI) f big data They are revolutionizing modern medicine. at all levels of clinical practice. From prevention, with increasingly improved prediction models to measure the risk of contracting a disease, for example, to diagnosis and treatment: there are already AI tools in health centers that help detect lesions in images of medical tests. Josep Munuera, Head of the Diagnostic Imaging Service at Sant Pau Hospital in Barcelona and an expert in digital technologies applied to health, asserts that AI – “it is not a single tool, there are different types” and points to helping professionals to be more precise in their professional practices.

No, it’s categorically progressing, the explosion of AI in health will not replace white coats or dehumanizing care. Quite the opposite, the radiologist (Girona, 47 years old) predicts: it will allow the doctor more time for the patient and help improve communication between them, he defends.

Ask. How mature are AI tools in medicine?

Answer. In some areas they are already mature and stable and in others they are being developed. If you give the example of medical imaging: where can we use AI in radiology? In image acquisition it is already well advanced. Today there are already tools with which, when performing an MRI, AI algorithms can be placed inside the device itself to speed up image acquisition: before it took 20 minutes to perform a scan, now you can reduce it by more than half to 10 minutes, because part of the image It is generated by or thanks to artificial intelligence algorithms. This generative image allows for a much shorter scan.

Image detection tools are already working in some centers, which are trained to identify certain types of diseases and when they occur, algorithms identify and distinguish them, so that some signs appear when the specialist doctor reviews the images. on the screen that indicates if the algorithm has determined that there is a patient, where it is and even helps us prioritize and sometimes a message is sent so that you look at this patient first or [para decirte que] You don’t have to rush too much with this other because it most likely has nothing on it. The prioritization and setting tools are already starting to become well established.

s. Is the goal in the end to buy time?

R was found. Not only. Yes, it is true that there is an important part which is process optimization and hence, it will definitely end up getting better over time. But there is also the issue of diagnostic accuracy. What is known is that with the vast majority of computer vision algorithms, combining a human reader with this algorithm increases accuracy. Therefore, we minimize diagnostic errors, both false negatives and false positives. Using these utilities helps us prioritize and be more accurate.

s. Little combination of doctor and machine vision. But, could there be biases and the machine’s evaluation modifying or adjusting your evaluation, for example?

R was found. Yes, there are discrepancies between the readings, but this is precisely one of the areas of work: how does artificial intelligence affect the fact that a person can adjust his reading. It is important to know that machines, in and of themselves, have a percentage of how accurately they can be diagnosed. And what also usually happens is that a person also has some degree of accuracy. And it may happen that if your accuracy ratio is less than that of the machine, you have to learn from the machine itself. Or vice versa: your accuracy may be greater as you train the machine to improve its learning.

s. Bugs have always been there, whether you’re a doctor or a machine. But who is most prone to error?

R was found. At the moment, the machines have advantages, such as they don’t get tired, they do the same diagnostics at 10 am or at 3 am, it’s fair… but they also have more limitations: algorithms have to be retrained over time. In the end, what you need to know is that it is a mixture of human and Programming What you have to keep it that precise: You need to define precisely how much you want to work in healthcare and, based on that, determine your best strategy.

s. The doctor is used to saying the last word and not being questioned too much. Are the professionals ready to face all this revolution that is beginning to surround them?

R was found. Of course, because here it is not about feeling questionable, but about being as accurate as possible. 21st century medicine is precision medicine. If you want to be accurate, you need to make the right decision at the right time. And sometimes you don’t have enough knowledge. Ultimately, they are tools that help us all.

s. How far can artificial intelligence go in medicine? What is the future?

R was found. In the end, what happens in the end is that we will see a lot of changes, but we will not see artificial intelligence. What we will have are tools Programming This will help, from the administrative point of view, to find the best piercing for the patient to bring in for an appointment; We will find the doctor who undergoes the consultation and who, with a few clicks, will be able to determine the treatment plan for the patient; Or in our case, because we might run tests that lasted a tenth of what they lasted 20 years ago, and once the patient leaves, they can actually identify hot spots that we have to pay attention to.

s. If there really are tools that read resonances and can detect abnormalities and this artificial intelligence will improve, are the days of the profession of a radiologist, for example, numbered?

R was found. No, quite the opposite. You will always need someone. The radiologist’s role goes beyond simple reading and interpretation of a medical image. Just as the surgeon’s role goes beyond the simple surgical procedure of cutting, joining, or removing a tumor.

s. But can you modify your medical practices in some way? Can you make it easier for fewer professionals?

R was found. I think this is a very misconception. In fact, it’s going to take at least the people out there now, because in the end, these AI tools don’t replace you, they empower you. So, I might look at the same number of pictures, I might take the same number of decisions or do biopsies. What I need is a tool to help me be more accurate in this decision-making process, but in no way will there be fewer people.

s. Jeffrey Hinton resigned a few days ago as Vice President of Engineering at Google and in an interview with EL PAÍS warned that AI for control must be learned before it is too late, because it can become smarter than the human mind. what do you think?

R was found. I agree with considerations of knowledge, ethics and distribution. This is still a technology that should also have some playbook. Perhaps, what is not given time is defining the rules of the ethical game, and how to use AI. But once it is properly identified, the next step is to start using it, because we cannot stop technological development either.

s. Is the AI ​​revolution outpacing you to process it?

R was found. Yes, it is true that technology is advancing rapidly and its implementation is accelerating. For example, computer vision algorithms, which model for viewing X-rays and determining whether nodules are visible, started out as beta tests in 2017 and 2018: five years have passed. Therefore, we will need, at most, five more in order to be able to use them daily in all positions where possible. The technology takes a few years to prove itself, but eventually it ends up being a hit.

AI tools do not divide us, but they help us spend more time with the patient.”

s. What are the biggest risks of artificial intelligence?

R was found. The risks are those of any technology: misuse, i.e. use in something not intended or unsuitable; Then there is malicious use: that someone wants to use or manipulate it so that it is used to their advantage. In the end, it is still just another tool to be controlled in cybersecurity.

s. Is there a danger that AI will end up depersonalizing or dehumanizing medicine?

R was found. These tools do not separate us, but should help us have more time with the patient, so that we can talk … Reducing the time for image acquisition allows you to spend more time with the patient to accompany him, ask him how he is, what to ask if you have doubts …and all of these aspects are humanized. Having this technology assistance allows you to have more time to devote to the patient.

s. Will we see a future where a chatbot, an artificial intelligence, answers a patient’s questions instead of having a doctor in the office?

R was found. If society already demands that computers not exclusively control phone services, why do we think healthcare can be, too? What can happen in the future is that we are three entities: patient, doctor and instrument and that all three of us interact. We will use these tools to better communicate with each other.

You can follow Country health and well-being in FacebookAnd Twitter And instagram.