A inteligência artificial pode tornar a Medicina mais humana? | Mariana Perroni | TEDxSaoPauloSalon

Translator: David DeRuwe Reviewer: Raissa Mendes The first time I saw a heart on the worktop in my college anatomy lab, I was totally impressed. The heart is a robust muscle, able to beat 2. 5 billion times in a person's lifetime.

At the beginning of college, one of the first things we learned is that the normal heart beats 60 to 100 times per minute, and that any alteration is a signal that something is wrong. A heart that beats faster, for example, can be the result of bleeding, an infection, or even of opening at TEDxSãoPaulo. (Laughter) Happily, today, we have various ways to investigate an abnormal heartbeat.

A problem that still persists in medicine is that we treat people in the way that works for the majority, not in the way that works or would be ideal for each one individually. Rarely do we actually think of who is this majority. I'll explain, more or less, who makes up this majority.

We doctors, when we treat a patient, we base it on a series of criteria called evidence-based medicine. These criteria show us, for example, which are the scientific studies that have the most weight to decide what works for the majority. Only, rarely do we ask who it is that is this majority, right?

I went after this. I found a study from the FDA, the agency that regulates new drugs that are released on the US market, that shows a profile of all the patients who participate in studies before drugs are released in the world, and I saw two problems. The first is only 1% of the population participates in these studies, and the second is who we call the majority, in reality, corresponds to white American men, who are less than 65 years old.

In other words, what we call the majority doesn't reflect our diversity. We're here, in 2018, at TEDx, talking about the future of health. Only medicine is still more primitive than what we like to admit.

We talk so much about precision medicine, but what we do today is imprecise medicine. OK, I came to bring some truths to you, but I don't have only bad news. Like Cazé said, I'm an intensive care doctor.

For those who don't know, we work in the ICU and are specialized to treat grave illnesses or illnesses that are likely to become grave. Yes, there are people that choose this for their own life, but beyond this, I'm absolutely passionate about uniting technology and health, and I believe that this has the potential to help both doctors and patients deliver better health to people. And how does this happen?

I'll explain it to you, but, to explain this, we need to understand what is happening in the world today. We live in a unique moment in the history of humanity. Why?

Because with the digital devices we have - computers, tablets, smartphones, intelligent watches - and with the social networks, we've never generated so much information in the world. For you to have an idea, 90% of all the information that exists in the world today, of all the existing data, was created in the last two years. In health, this is even more pronounced.

We live in an era where, like it or not, we're never offline. We're generating data all the time. So you can imagine the magnitude of this.

If we gathered all the health information a healthy person generates throughout their life, it would be the equivalent of 300 million books, full of health information. If we picked out all this information and put it onto tablets, and then piled up these tablets, the pile would reach a third of the distance from the Earth to the Moon. Only, out of all this information that we have available today and using the tools we possess, this is how much we can analyze: it's 0.

5%. Now, imagine how much information we are losing in this 99. 5% that we're putting aside.

It's absolutely valid information that could show us, for example, how to improve people's quality of life, reduce avoidable deaths, prevent epidemics, and, why not, even cure illnesses. Only here, we arrive at a second challenge: who's going to analyze all this data? The doctor?

Truly, we already live in a tsunami of information, just to stay current. Studies show that for a medical expert to stay up to date with every scientific study, published in their area, it would be necessary to study 167 hours a week. That's more than 20 hours a day.

It's impractical. In truth, our brain isn't a sufficient machine to process all this information, much less the traditional computer systems. Why?

Because they do only what they're programmed to do. They're not innovators, they don't give us hypotheses to help solve problems. Consequently, we think, "What could solve this problem?

" The first thing that comes to mind are robots with artificial intelligence, but the second thing that comes to mind is this: robots decimating whole populations who were innocent enough to trust them. And how does a doctor work with artificial intelligence? By turning into a robot?

Happily, no. Happily, this is all limited to the movie screens and science fiction books. Artificial intelligence and robots will never replace the doctor, much less decimate humanity.

Actually, we need to quit being afraid of artificial intelligence and understand better what it is. Artificial intelligence is nothing more than a set of tools that allow our intelligence to be increased. It's the third era of computation, the era that will permit us to leave those systems that do only what they were programmed to do, and move on to systems, which organize this information world, analyse what's inside, interact with us in a more natural way, and develop hypotheses for us to resolve problems, getting more intelligent each time we use them.

In health, this has an enormous potential because this can help us reduce bureaucracy, make more precise diagnoses, and provide treatments that have a bigger real chance of helping people. This was partly what motivated me. Up until 2015, I worked 80 to 100 hours, in three ICUs, in different hospitals here in São Paulo.

Today, I divide my time between doing this and working for a big technology company, developing projects that help us explore how artificial intelligence can help doctors and patients provide better health for people. I'll give some examples of what's already possible. I'll start with oncology, the specialty that treats cancer.

Oncology is one of the specialties in which it is most difficult to keep up to date. Close to 75,000 articles are published per year that the oncologist has to know, with new medications and new treatment regimes. Today, it's already possible for oncology doctors, at the time of the consultation, to cross-reference the patient's data with all the available scientific literature to get a type of ranking that shows the treatments that have the best chance of working for each individual patient, instead of what works for the majority of people.

This has been used in leading hospitals in the United States, and even in countries like India. India has 1,000 oncologists for a population of 1. 4 billion people.

Many patients, when they are diagnosed, don't have the chance to go to an oncology treatment center. Many, unfortunately, end up dying before getting there. So, what artificial intelligence has done is permit more general doctors to advance the process and start the treatment of these patients, so they have more hope and have more time to get to an oncologist who will continue their treatment.

In genetic sequencing, it's also already possible to use artificial intelligence. Today, we get to use this type of tool to read, for example, the genetic sequencing results of a person with cancer. With this, we can detect and rank the person's most relevant mutations, cross-reference the literature, and find specific treatments for this mutation.

Here also, we leave the current model for treating cancer, which is to treat by the organ type involved - breast cancer, gastric cancer, prostate cancer - and begin to treat what's happening with that person's cancer by the mutation that person has. Humans can do this analysis too, but it takes 160 hours to analyze a person's genetic sequence, almost a week. With artificial intelligence, this time falls to 10 minutes.

The first successful use of artificial intelligence for genetic sequencing happened in Japan. In a hospital at the University of Tokyo, there was a patient with leukemia who had undergone three treatment regimes without success, the ones that function for the majority. Only, she wasn't the majority.

Then, the doctors decided to test, "Let's take this patient's sequence data, and let artificial intelligence read and analyze it with us. " They discovered she had a rare mutation for which there was a treatment that hadn't been tried yet, and today, this patient is recovering. Another example is diabetes.

Thousands of patients with diabetes donated their glucose and blood sugar measurement data to be interpreted by artificial intelligence. This allowed an algorithm to be created that predicts, three hours ahead, when these people will have hypoglycemia. Hypoglycemia is that drop in blood sugar that is super-dangerous for diabetes patients.

And, the moment the system detects this, it sends messages by smartphone to the people, advising them this will happen, along with the measures they can take to avoid it. Well, there are many examples I can give you of what is happening today. The patients, me, you, all of us, we have a unique opportunity in this moment because we have this computational power in our hands, in our cell phones, our smartphones.

Right now, we have the opportunity to start being more active in relation to our health. This can start in various ways: from demanding access to our medical records from the doctors and institutions through which we pass, so that we, too, can have ownership of this data that isn't only theirs, to changing the applications that teach us into applications that help us - counting the steps we take, tracking how we sleep, or how we choose to eat, and how this impacts our health. The idea here is that, from the moment we own our own data, we can decide what to do with it.

It's more or less what we do with GPS programs where we share data about our route, and we get back how much time we will have to wait in traffic, what obstacles are on our route, how to avoid them, and how to find a better way to go. If we do this in health, sharing our data, we start to understand that different kinds of treatments and measures work for each person, in each neighborhood, in each city, in each country, and certainly, in the world. One of the most humane things we can do is to donate organs to a person in need.

I believe that, in the future, we will not only donate organs, but it will also be possible for us to donate data because instead of helping a person once, we can help many times, and we can guarantee that each time, more hearts will be able to beat those 2. 5 billion times that they were born to beat. Thank you very much.