Precision medicine is more than gene analysis: thanks to extensive, individual data sets modern artificial intelligence (AI) algorithms help ensure more effective prevention, more precise diagnostics, predictive treatment planning, and customized insurant and patient management.
Photos: Sebastian Gabriel
Syringes, disinfectants, and scalpels are not the first things that come to mind when Christian Lautner of the Berlin startup incubator Flying Health envisages the future of medicine. Today's medicine covers at most ten percent of what we understand by health, in Lautner’s view. With the help of mobile sensors, digital platforms, and AI the remaining 90 percent will be integrated in the next few years, Lautner believes. His company has set up its own doctors’ portal.
In Lautner’s vision of the future, healthcare will move out of medical establishments and become part of everyday life. Sensors in shopping trolleys or in the bathroom mirror will alert people to any health abnormalities. Even before a doctor or nurse enters the scene, a voice-controlled symptom checker will offer the most likely diagnoses. Pharmaceuticals will be delivered by drone; video capsules will perform colorectal cancer screenings. And when a doctor really is needed, an emergency vehicle will arrive with all the information on preliminary diseases and additional facts from the digital health record, and will be able to handle most medical problems on the spot.
Avoiding burnout: algorithms for segmentation
In this vision of healthcare, self-learning algorithms are at work in every area, from algorithm-based tentative diagnosis through digital-based cancer screening to controlling drones and emergency vehicles. As a result, medicine should become more precise and effective. Prevention and early detection will be strengthened and at least partially automated. Of course, much of this is still a long way off.
But AI algorithms are also increasingly being used where the patient is already sick, in that ten percent of healthcare referred to by Lautner where classic diagnostics and treatment take place. Professor Anja Hennemuth of the Fraunhofer MEVIS research institute and the Charité Berlin Institute for Imaging Science and Computational Modelling in Cardiovascular Medicine used the specific example of heart disease. She sees cardiac MRI diagnostics in particular as an obvious area for the application of “narrow AI”; that is, highly focused, self-learning algorithms.
It is already possible for well-trained algorithms to take over the time-consuming segmentation of the heart that had to be done manually by doctors performing MRI diagnostics until now. Given the constantly rising number of patients per doctor, allowing AI to take over such routine tasks is effectively a way to avoid burnout, Hennemuth believes. Soon neural networks will be applied to more demanding tasks, such as simulating how a sick heart responds to stress, she said.
AI in intensive care and laboratory medicine: making sense of huge data sets
Cardiovascular patients also benefit from AI algorithms in other ways besides imaging. Hennemuth referred to a recently published research paper by Deutsches Herzzentrum Berlin (German Heart Center Berlin) on a deep learning algorithm that has been trained using dozens of parameters from electronic records and monitoring systems, to identify those patients in intensive care units who have an increased risk of bleeding or developing renal failure after heart surgery. “We will see many more such applications in future,” the scientist is convinced.
Professor Tim Conrad of the Computational Proteomics working group at FU Berlin’s Department of Mathematics made it clear that AI algorithms offer exciting opportunities in other areas of medical diagnostics, too. The team uses mass spectrometry to examine blood samples for proteins with the aim of detecting cancer earlier.
Although mass spectrometry produces huge data sets, in principle this proteomic analysis could work without AI. However, deep learning algorithms help to identify not just the most common proteins but also those that are the most relevant for the respective medical issue, Conrad said. The first decision support systems for several cancer types are already in development.
Colon cancer surgery: How good is the circulation after anastomosis?
On the treatment side, minimally invasive procedures are among those that may benefit from AI systems. Janek Gröhl of the Division of Computer Assisted Medical Interventions headed by Professor Lena Maier-Hein of the German Cancer Research Center (DKFZ) in Heidelberg foresees an era of “surgical data science.” In this new surgical era, smart assistance systems use a wide variety of data produced prior to and during surgery to make interventions faster, safer, and more precise.
An example is the monitoring of blood flow in colorectal cancer surgery. Until now, in the case of minimally invasive surgery, this had been done visually by the doctor on the basis of the image taken by a standard camera. “But that's a fine line,” said Gröhl. In Heidelberg they now use multi-spectral cameras and have self-learning algorithms evaluate the images pixel by pixel. Using an intestinal model, it has proved possible to predict circulation with a very high degree of precision. There has even been interest from neuroscientists, Gröhl said, who want to use the algorithm to visualize the brain's oxygen consumption.
- The statements by Siemens Healthineers customers described herein are based on results that were achieved in the customer’s unique setting. Since there is no “typical” hospital and many variables exist (e.g., hospital size, case mix, level of IT adoption) there can be no guarantee that other customers will achieve the same results.