AI in medicine: Data saves lives – health

AI in medicine: Data saves lives – health


Artificial intelligence (AI) used to be science fiction. Since Chat-GPT came into being, it has noticeably penetrated people’s everyday lives. What is less known is that AI methods in the medicine been used for a long time. Doctors and scientists are trying to use AI to improve patient treatment. This results in hopes that many problems could be overcome, but it also triggers fears, said Markus Schwaiger, President of the Bavarian Academy of Sciences (BAdW), during a joint health forum between BAdW and South German newspaperwhich recently took place in Munich and was dedicated to the opportunities and risks of using AI in medicine.

Torsten Haferlach, head of the Munich leukemia laboratory, where blood and bone marrow samples from patients are already being analyzed using AI, showed how helpful algorithms and machine learning can already be today. Depending on the method required, up to 80 percent of all samples from all over Germany come to the Munich laboratory to diagnose leukemia. “Without AI, we wouldn’t be able to make diagnoses anywhere near as precise,” said Haferlach at the health forum. There are now more than 300 different types of leukemia known, and the classification manuals contain around 700 pages. “No one can see it anymore, the AI ​​can.”

Nevertheless, the AI ​​in Haferlach’s laboratory is not left to make the diagnosis alone. Supported by algorithms, the appearance of the cells from blood or bone marrow is evaluated and their immunological properties and genetic characteristics are analyzed. But in the end, experienced employees check the findings. This saves time, increases safety and is therefore of great benefit to patients, said Haferlach. The diagnosis depends on which therapy the patient receives. It is therefore essential for the success of the treatment.

“After all, nobody puts a city map next to their cell phone.”

“In more than 95 percent of cases, the AI’s suggestion is the right one – and the systems continue to learn with every finding and every error,” reported Haferlach, who is himself a hematologist, a specialist in blood diseases. You have to remember that only the best employees check the AI. “The systems today are probably already better than those employees who do the tests less often or for less time,” said the doctor. Today, AI is at the level that humans have achieved after two years of professional experience. It is therefore necessary to use such techniques: “You can no longer say that I’m doing it the way I’ve always done it if you want to work medically and ethically correctly.”

Torsten Haferlach drew a comparison to Google Maps: “After all, nobody puts a city map next to their cell phone to check the directions,” he said. Trust in the electronic card system has developed through experience. We know that Google Maps is usually right. It is true that the use of this technology means that human expertise is lost over time, says Haferlach – but the more reliable modern methods become, the more dispensable some knowledge becomes: “Most young people today cannot do a fully developed Falk plan can no longer be folded properly, but they don’t need to be able to do that anymore.”

Computer scientist Julia Schnabel from the Helmholtz Center and the Technical University of Munich is also working on better cancer diagnostics. She is considered a pioneer in the application of AI to analyze medical image data. Tomograph images are often used for the early detection of tumors. Experienced radiologists can recognize foci of cancer in the images – but of course only when they are visible to the human eye. “The AI ​​can detect tumors much earlier than humans,” says Schnabel, “if it is trained accordingly.”

To do this, systems are trained with the help of patients who are repeatedly examined in a tomograph over a longer period of time as part of a screening. Heavy smokers, for example, who are feared to have lung cancer. If individual test subjects actually become ill over time, the AI ​​can use the regularly created images to learn which tiny abnormalities were already visible before the cancer was diagnosed. This means that it can sound the alarm to future patients much earlier than was previously possible and than will ever be possible for a human being. “Lung cancer is the most common Cancer and the most common cause of death from cancer, but unfortunately it is often only noticed late,” says Schnabel. Then the chances of treatment are usually low. However, the earlier the tumor is noticed, the better the chances of success.

“A radiologist trains on 30,000 pieces of data, which is almost nothing for the AI.”

The experts at the health forum promoted trust in modern methods. “A radiologist trains on 30,000 pieces of data before being released to patients,” says Schnabel, “that sounds like a lot, but for the AI ​​it’s almost nothing.” It processes millions of pieces of data in no time. Nevertheless, no one has to worry that doctors will one day be replaced by AI. “AI is a valuable addition, not a replacement. In the end, people and their decisions are always needed.”

However, the AI ​​is only as good as the data it is trained with. Schnabel therefore advocated for patients to share their anonymized data with researchers. “Data saves lives,” says the machine learning expert. Everyone is a patient at some point and can then benefit from other people’s data. In order to develop good systems for the local population, representative data from the local population is needed. “It’s not enough to use data from the USA or China – that means the AI ​​can’t learn well for our needs here,” says Schnabel. “We have to be able to map the population in Germany in order to be able to treat them better, and for this we need the contribution of every individual.”

Björn Eskofier is working on better data use and data availability. The computer scientist is researching at the University of Erlangen-Nuremberg and the Helmholtz Center in Munich how patients can be integrated into a digital health system for optimal care. Eskofier is committed to a European health data space, the European Health Data Space, where one day people across Europe will be able to access their medical records. A precursor for this European data space has been initiated for Germany with the Health Data Usage Act, which was recently passed, said Eskofier. The data should be available to patients at any time via their cell phone.

“I also trust my bank with money, although it is not 100 percent secure.”

From Eskofier’s perspective, the benefits are obvious: For example, medications given to patients are shockingly often not effective. Not only does this mean billions of euros are wasted every year, it also threatens people’s health if they don’t get the medicines that best help them. But if people across the country share their treatment data anonymously, the treatment success of many millions of patients can be analyzed and the best therapy can be determined for future patients. Be there data protection just as important as data usage. He therefore advocates “opt-out” solutions. This means that data is generally used, but every citizen can reject it at any time if they do not want to.

But can the data also end up in the wrong hands? Patient advocates fear that insurance companies or employers could gain access to health data. “No IT system is 100 percent secure,” says Eskofier, “but we are working to make the systems as secure as possible.” Julia Schnabel made a comparison to other everyday situations: “I also trust my bank with money, even though it’s not 100 percent secure, and I also trust the pilot on the plane.”

AI is not only getting better in the use of data for patients, says Schnabel. They are also getting better and better at data protection. Data could be anonymized and pseudonymized using modern methods so that they can be used for research, but cannot be used to identify an individual. Also the Ethicist Alena Buyx advocated using data with the best possible data protection. And for Karl-Walter Jauch, the former medical director of the Großhadern Clinic, it follows from all of this: “AI and digitalization are more than just an opportunity for maintaining and improving the healthcare system. They are a basic requirement.”


Source link