When floors and beds assist with care and algorithms evaluate lungs

At Fraunhofer IGD, researchers are looking to identify concrete opportunities to integrate digital technologies into clinical routine—and create truly smart hospitals equipped for the challenges of the future. The team includes Dr. Stefan Wesarg. The physicist believes that algorithms and artificial intelligence (AI) will soon be part and parcel of healthcare. They have the potential to accelerate diagnoses, support treatment decision-making and improve aftercare. To give one example: “When assessing a head-and-neck CT scan, the radiologist has to scroll through a hundred or so cross-sectional images and identify anything that looks unusual,” says Wesarg. AI can analyze the image data much faster and highlight any anomalies for the doctor to examine more closely. Moreover, the system can overlay contour lines on images of certain organs, for instance. And “in the next stage of development, it will be able to decide whether an observed change is benign or malignant,” Wesarg explains. In the future, an algorithm might even determine how advanced a case of cancer is, and recommend a corresponding treatment.

A walking frame that knows its way around

Stefan Wesarg’s colleague Dr. Mario Aehnelt works 700 kilometers away in Rostock on solutions for patients and their care providers. The IT expert and his team have developed a self-navigating robot walker. “It can move autonomously through the hospital and accompany patients to their hospital-internal appointments or back to their room—helping them find their way around safely and reliably.” It is a great aid for elderly patients in particular, and helps free up staff to focus on other tasks.

A bed that helps monitor patients

Hospital units for geriatric patients or Alzheimer’s sufferers also benefit from Florian Kirchbuchner’s work. He and his team specialize in connected sensors. Their research has a range of applications, including for smart hospital beds. The intelligent furniture can, for instance, alert nurses when a patient needs to be turned over to prevent pressure sores. Or it can notify staff when a patient with dementia gets up in the night. “We train the system to determine whether a person is simply turning over or actually trying to exit their bed,” states Kirchbuchner.

A floor that responds to falls

In the same vein, a system that recognizes falls would be a valuable addition to hospital rooms. It works by detecting unusual patterns of movement—and sounding the alarm when a patient falls down. The electric-field sensing technology is located under the flooring and includes pressure sensors that can determine where a person is within a room. “If you detect a sudden impact followed by inactivity, then it is in all likelihood a fall,” explains Kirchbuchner.

A self-navigating robot walker, an observant bed and a responsive floor, when market-ready, will come at a price—and that is especially true when equipping entire hospital departments with these smart aids. But, according to developer Florian Kirchbuchner, AI can ultimately save healthcare providers money. “On the whole, digital solutions are cheaper than the consequences of not deploying AI.” Currently, Kirchbuchner’s team is evaluating the added value of these technologies for society as a whole—"to clarify whether AI is a worthwhile investment for health insurers.”

Building acceptance for digital solutions

Given demographic trends in Germany and elsewhere, it is only a matter of time before AI is—and perhaps must be—deployed in hospitals. However, it is important to build acceptance, as underscored by Professor Jörn Kohlhammer. The IT specialist uses visualization to make data from AI more easily understood, and to encourage medical professionals to embrace the technologies and put them into practice. “Even for those with expert knowledge, algorithms and machine-learning processes are not easy to comprehend,” he admits. “However, in healthcare, ease of understanding and transparency are critical. Physicians ultimately have to justify their decisions to patients and insurers.”

There remain several hurdles to overcome before Fraunhofer IGD AI systems are ready for real-world implementation. They must first be assessed and approved as medical products. The rules are strict, and rightly so—and have yet to be defined for AI. And from a legal standpoint, no technology can currently be used if it is modified after approval. Yet that would mean AI systems would have to complete all their learning prior to introduction, which is contrary to their nature.

All the same, it will not be long before algorithms are commonplace in hospitals. Indeed, it is already hard to imagine the world of medicine without AI.