How could artificial intelligence change medicine?

How artificial intelligence extends our lives

12/18/2018 Positions magazine

The digitization of medicine could give life expectancy a boost. Algorithms are already making diagnoses and suggesting therapies - often more precisely and faster than humans. The insurers are reacting to this with new offers.

Gaston D’Aquino actually only got the Apple Watch to make his life a little easier. Because he constantly overheard calls on his smartphone - the vibration alarm on his wrist was supposed to solve the problem. The 76-year-old jeweler from Hong Kong would never have dreamed that the small high-tech watch would save his life. Not even when he is at church on a Sunday and the smartwatch sounds the alarm. And not because of an incoming call: his heart rate is far too high for his age. Although he feels in top shape, he goes to the doctor. Diagnosis: Two arteries are completely blocked. Without immediate surgery, he would have suffered a fatal heart attack sooner or later. D’Aquino owes the fact that he was given a longer life to a technical revolution that will radically change medicine over the next few decades. According to the dream of the researchers, the computer will become a health watchdog for the patient and an equal colleague for the doctor. Wearable sensors, so-called wearables, should replace stethoscopes and laboratory tests, robots carry out operations independently, and artificial intelligence (AI) should be part of everyday hospital life. The goal of human-machine cooperation: better diagnoses, better therapies, better prevention.

Symptom check via smartphone

Smart wristbands or plasters are already measuring skin moisture, pulse and physical activity and are intended to provide early warning of cardiac arrhythmias, diabetes or strokes. The BiliScreen app, which is still under development, uses a mobile phone camera to detect discoloration in the eye and could thus identify pancreatic cancer in the future. Philips promises to use its CareSage program to predict the risk of hospital admissions within the next 30 days using an algorithm that analyzes the patient's previous illnesses, prescribed medication and emergency call data. And the health app Ada Health sees itself as an anamnesis tool in your pocket. After asking questions about symptoms, constitution and lifestyle, she names possible causes of dizziness, tiredness or stomach pain, saves the user's personal history of suffering and can even forward them to a doctor like a medical record. At some point, the app will also integrate data from genetic analyzes, blood tests or fitness trackers in order to reveal the risk of inheritable diseases and to warn even earlier if something goes wrong in the body.

"The topic of artificial intelligence will become very important in medicine," says Klaus Juffernbruch, doctor, computer scientist and professor at the FOM University. “Some patients may wait too long. Often they don't even notice that something is wrong. But if you hold your smartphone on it and it says, 'Warning, this looks like skin cancer. Better go to the doctor Arzt, it could certainly prevent many premature deaths. "

The insurance industry is also recognizing the potential of wearables. The US company John Hancock, for example, has long been offering its customers the Apple Watch at a fraction of the retail price. Those who share their data with the life insurer and live healthily benefit in the form of premiums. From 2019, John Hancock will only sell such policies - classic life insurance will be eliminated from the program.

People are getting older - retirement is also expanding. With the “7 years longer” initiative, GDV explains what consequences this has for individuals and society:

It's not that far in Germany yet, but artificial intelligence is also finding its way into the healthcare system here. At the AOK, there are no longer only bonus points for donating blood or yoga courses, data from fitness trackers and smartphones can also be converted into rewards. Techniker Krankenkasse offers a fitness program that uses pedometers to monitor participants. Both health insurance companies use established apps from the IT giants Google, Samsung and Apple. Fitness trackers can also be linked to the Vivy app, which 14 statutory and two private health insurers presented in September. It should bundle the patient data digitally. Some health insurers also subsidize the purchase of smartwatches or offer certain models as rewards.

That should just be the beginning. “Such bonus programs will certainly gain in importance,” says Andreas Richter, head of the Institute for Risk Management and Insurance at the Ludwig Maximilians University in Munich. He believes that fitness trackers in private health insurance can contribute to a better risk classification, similar to telematics tariffs from car insurers. There the driving behavior is recorded - those who drive carefully receive discounts. However, Richter does not believe that fitness monitoring will become mandatory at some point. It is only conceivable for certain tariffs or for bonus programs.

How artificial intelligence works

But how does software even become intelligent? "Actually, there aren't that many differences between humans and machines," says FOM Professor Juffernbruch. “Building artificial intelligence is not unlike studying medicine.” If an AI is to become an expert on skin cancer, for example, it is fed countless images of melanomas and harmless birthmarks. As the developer “tells” you which images show something malicious, the program learns to interpret image pixels and patterns. This creates an artificial neural network that learns with every new recording.

AI is also being used more and more in hospitals. Especially in radiology, where doctors sometimes examine hundreds of X-rays, CTs and MRIs every day. The first practical applications show that AI is just as good as human specialists. If not better. At Essen University Hospital, for example, a system for diagnosing uterine cancer works side by side with doctors. Up to 2000 parameters can be assigned to the disease. They provide information about the aggressiveness, whether the tumor has already spread or whether it will still do so. While a top radiologist can look at maybe ten parameters, AI can manage all of them and make an almost 100 percent diagnosis before the biopsy.

Cancer can be assessed by an AI system at the Essen University Clinic in order to make a diagnosis.

In addition, the AI ​​is extremely fast. A cardiologist needs a good half an hour to assess the cross-sectional image of a heart. The AI ​​does the job in 15 seconds. And that without having to take a routine or vacation and without getting tired, not even at night after a 36-hour shift in the emergency room. “That gives the doctor an incredible amount of time. In this time, he can take care of the really tricky cases, can devote more time to the patients, explain the findings to them and shorten their waiting time for appointments, ”says Juffernbruch.

Watson shows the opportunities and limitations of AI

© Getty Images, Apple, Samsung

The enormous growth of data is no longer manageable for humans alone. In 1950 it took 50 years to double medical knowledge, today it takes just 75 days. There are currently around 150 exabytes of health data. A number with 19 digits! “No doctor can follow up on that,” says Juffernbruch. Digital assistants such as the IBM Watson program can help doctors with their processor power. Because Watson is able to memorize all of Wikipedia in just an hour. The algorithm analyzed 200 million pages of text in three seconds. This ability is used in medicine, among other things, for diagnosing and planning therapy in cancer and rare diseases. Data on the patient's family history, medication, symptoms, findings and notes from doctors and nurses, current research and clinical studies - all of these are used by artificial intelligence to spit out a list of possible diagnoses and therapies.

But Watson also shows that AI is still at the very beginning in medicine. German and Danish clinics ended their collaboration with IBM because they considered Watson's cancer therapy proposals to be questionable. The problem: an AI is only as good as the material it is fed with. In Watson's case, it was data from an upper-class cancer clinic in New York. However, there are very different treatment guidelines than in Europe. In the USA, for example, people quickly and easily resort to expensive new drugs because doctors earn money from them. "To ensure quality, we need uniform standards," says Juffernbruch. “But there are many different medical schools. Everyone still does what they want. ”Nevertheless, he predicts a great future for digital medicine. For the patient, this means above all that diseases are detected much earlier - including the more than 7000 particularly rare ones.

it takes time for the world's medical knowledge to double. In 1950 the time span was still 50 years.

Hope for even more individual therapies

Insurers and medical professionals in Germany also rely on this. Health insurance companies and doctors agreed on a basic concept for an electronic patient record with the Ministry of Health in October. From 2021, patients in Germany should be able to take their health data with them on their smartphones or tablets. This could also include values ​​from fitness trackers and other wearables.

However, privacy advocates see this as a risk. Because health data are particularly sensitive. Most recently, when the Vivy app was presented, the Chaos Computer Club warned of the risk of abuse if data were stored centrally. In addition, smartphones usually no longer receive security updates after two years, which makes them easier to crack. The consumer advice center in North Rhine-Westphalia also criticized the fact that it was hardly possible to monitor the data collected by fitness trackers. It is precisely these apps that health insurers use to collect step numbers and other health values.

of health data exist today. Only supercomputers are able to handle this amount.

The question of data security is likely to have a decisive influence on the importance of wearables for insurance companies. “How quickly such tariffs are implemented on a large scale depends to a large extent on how much the customers perceive the data recording as an invasion of their privacy,” says the Munich professor Richter. "The market here is certainly a little more cautious than in the United States in particular."

Klaus Juffernbruch still believes in the breakthrough of technology. “What is still 'nice to have' today will one day be state of the art,” he says. Then a doctor will have to answer in malpractice proceedings in the event of a misdiagnosis without a second AI opinion. In addition, the technology makes it possible to personalize medicine. Therapies could be tailored for each patient. Because it is easier for an AI to analyze the whole person - from the blood sugar level to the genome.

Knowing more about his own body changed him, says Hong Kong jeweler Gaston D’Aquino. Today he lives healthier and moves more. Because even the most intelligent doctor on his wrist cannot relieve him of responsibility for his life.

Text: Christin Meißner, Robert Otto-Moog