Skip to main content
Supercomputers to Combat Diseases
Irina Sharnina -

Using AI to Combat Diseases

Artificial Intelligence (AI) is becoming increasingly crucial in medical research, and Helmholtz Munich is one of the world's leading institutions in this field. It can detect tumors as effectively as severe eye diseases.

Artificial Intelligence (AI) is becoming increasingly crucial in medical research, and Helmholtz Munich is one of the world's leading institutions in this field. It can detect tumors as effectively as severe eye diseases.

Sometimes, a single image is enough to improve a life. For example, Professor Fabian Theis has used a photo of the retina as a benchmark for his work. When an artificial intelligence, which he and his team are working on, analyzes this image, it can automatically identify severe eye diseases. "This could allow clinics and practices to save valuable time and protect many people from losing their eyesight," says the biophysicist who leads the Computational Health Center and the Institute for Computational Biology at Helmholtz Munich.


"AI based technology could allow clinics and practices to save valuable time and protect many people from losing their eyesight."

Theis collaborates not only with medical professionals but also with computer scientists, physicists, and biologists. His work is at the forefront of what is technically possible: From vast amounts of data, the team extracts valuable information about diseases and the molecular mechanisms that trigger them. Artificial intelligence, machine learning, and deep learning methods form the high-tech trio underlying these analyses. These methods are enabling research to delve into areas that were inaccessible to medical professionals just a few years ago.

"Save Eyesight"

One example is the "Save Eyesight" challenge related to retinal photos. It is one of the topics on which Fabian Theis applies his state-of-the-art methods. The process is similar to other areas where artificial intelligence is used in medicine: The team feeds a computer with hundreds of images of human retinas, both healthy and diseased at various stages. Artificial intelligence learns to recognize the differences. After this training, it can diagnose retinal conditions with great reliability when analyzing further retina images. Theis and his team have already achieved success in age-related macular degeneration and diabetic retinopathy—two retinal diseases that can lead to blindness in the worst cases.

"An algorithm we developed can automatically detect these diseases in a scan of an eye, even at a very early stage when treatments are still highly effective," says Fabian Theis. The algorithm can even look into the future and make predictions about how the eye disease will develop and which therapy is best for treatment.

Helmholtz Munich leads the way in utilizing artificial intelligence in medical research. The numerous working groups and institutes continually explore new fields in medical research using cutting-edge methods.

"Human intelligence still surpasses computers. But artificial intelligence has the potential to effectively support medical professionals."


Julia Schnabel heads the Institute for Machine Learning in Biomedical Imaging at Helmholtz Munich. The computer scientist and her team focus on image analysis: From the abundance of medical images, whether X-rays, ultrasound scans, or high-resolution magnetic resonance images, artificial intelligence should be able to detect diseases in the future, ideally long before they become apparent. The AI is trained with extensive datasets of known disease progressions, and intelligent simulations can automatically identify potential anomalies, even in children. This is intended to support healthcare professionals in their work in the future.

Optimizing Incidental Findings and Early Diagnoses with AI

"A good radiologist sometimes finds hints in a patient's images that they weren't even looking for," explains Schnabel the approach that she envisions for computer use as well: "For example, if someone is examined with an X-ray after a bicycle accident, and the doctor notices that there's early-stage lung cancer, that's called an incidental finding." Such accidental discoveries can lead to early treatment, with the best chances of recovery—much better than if the patient had sought medical attention years later due to breathing problems, by which time the tumor might have advanced and potentially metastasized.

Leaving Nothing to Chance: Improving Preventive Screenings with AI

Julia Schnabel and her team's goal is to eliminate such chance discoveries from incidental findings. The idea is to systematically utilize artificial intelligence to analyze images that are already taken during medical examinations - can preventive care be systematized this way?

"Our dream is to be able to use every scan made by everyone," Schnabel says. She is quick to point out that they need to consider the benefits of this approach, addressing concerns about patient privacy:

"We must realize what we gain from this – each and every one of us. For example, if we can detect cancer early or provide preventive therapy to teenagers who have a very high risk of developing a certain disease later on, we can save lives."


Every Pixel counts

The quantity of available images is a key factor in the development of artificial intelligence. The more data from patients in the form of medical images, the more precisely artificial intelligence can draw conclusions. "For my diploma thesis, I worked with three images," says Julia Schnabel, illustrating how much her field has evolved over the years. Today, computers can handle thousands of images, and their analytical depth has greatly increased. However, there are often not enough image data available for researchers. This may be due to data privacy issues, the lack of digitized images in hospitals, or medical diagnoses not being in a format that computers can work with. The UK Biobank with 100,000 image data or the German National Cohort with 30,000 participants are significant aids for Julia Schnabel's research and her team, but they are only applicable to limited areas. "With 1,000 images, you can already do a lot, and sometimes I'd be happy with just 100," she says. "Every single pixel counts for us."

This is a relatively new field in medical research that has opened up in the realm of artificial intelligence. The speed at which it progresses and the opportunities it offers can be seen in the fact that new working groups are formed at Helmholtz Munich every year, more experts join the field, and computer capabilities continually improve. This pace of progress is what fascinates researchers. "I've never seen a field progress so rapidly," says Fabian Theis. "Not even computer technology advances as quickly."

This is good news for patients because successful disease treatment thanks to artificial intelligence is becoming increasingly likely - whether it's a dangerous retinal condition or serendipitous tumor discoveries.

More about Computational Health at Helmholtz Munich

Learn more about the research at the Computational Health Center

Check out more about Prof. Fabian Theis at the Institute of Computational Biology and about Prof. Julia Schnabel at the Institute of Machine Learning in Biomedical Imaging.