What is currently happening here at the Essen University Hospital is a tiny piece of the puzzle of what is known in medical jargon as a “smart hospital”, i.e. a digitalized hospital. Almost all heart patients who agree end up in the small room converted into a photo studio during their hospital stay; around 5,000 have already been there. Working students earn extra money here or sometimes even data material for their doctoral thesis by photographing faces. They take a total of eight portrait pictures of each heart patient from different perspectives. The idea behind it: “An increased risk of cardiovascular diseases can be recognized by features on the face, for example certain ear wrinkles or a change in the back of the eyes,” says Tienush Rassaf, director of the local clinic for cardiology and angiology. A facial recognition app against heart attacks In the photo studio, the doctors collect pictures of patients’ faces whose diagnoses they know in detail – of course, the patients are in the clinic for treatment anyway. The doctors use the images and diagnoses to train an AI. The goal: “In the future, a smartphone app should be able to use facial recognition to predict how high the app user’s risk of acute heart attack is,” explains Rassaf. “Easily with your private cell phone at home.” An AI learns facial recognition from thousands of patient photos.Stefan FingerWhat sounds like a science fiction gimmick is a very serious matter. Eight doctors are working on the project, not full-time, but on an ongoing basis. Rassaf estimates that a double-digit million sum will flow into the project before it is ready for the market. He hopes that in just two years the AI could be so well trained that it produces robust results. The facial recognition app is not Rassaf’s only AI project. At the same time, the doctor is working on a screening procedure for cardiac amyloidosis, a serious heart disease that is currently difficult to diagnose. In this case too, he is training an AI together with colleagues, but the users should not be private individuals, but family doctors. In the future, they should be able to detect an increased risk in their patients using three simple ECG electrodes that they connect to their smartphone. A cable ordered from Amazon connects the ECG and smartphone.Stefan FingerRassaf and his colleagues have built a demonstrable test device; they “simply ordered the ECG cable and electrodes from Amazon,” says the doctor. Cost of the cable: 10 euros. A ridiculous value compared to the costs of treating cardiac amyloidosis that was diagnosed too late: around 180,000 euros per year. Difficult cost-benefit calculation The numbers that the heart specialist is dealing with are impressive. But the calculations rarely work as easily as in this example. Attempts to estimate the financial potential of hospital digitalization are complex. The management consultancy McKinsey ventured into this two years ago. A study talks about possible productivity increases in inpatient hospital care in this country amounting to 25.8 billion euros per year, for example through better resource management, avoiding unnecessary double examinations or monitoring chronically ill patients at home instead of inpatient. Because initial investments in digitalization are very high , staff have to be trained and “saved” staff are generally not saved but are used to plug existing skills gaps, there are also experts who shy away from cost-benefit calculations. “If someone tells you a number about how much can be saved in the medium to long term through digitalization in hospitals, then they are not being serious,” claims Jörg Asma, partner at the auditing and consulting firm PWC. Asma heads the department “ Digital Healthcare Consulting” and refers to a study called “Digitization in Hospitals” that PWC published a little over six months ago. It contains sentences like: “The high costs of digitalization are becoming clear, but the benefits cannot yet be quantified.” Which – as Asma immediately points out – is due, among other things, to the fact that the benefits of a saved human life do not come with a price tag has. Or as the study concludes: “A large part of the benefits of digital solutions do not necessarily arise in relation to the costs.” Rather, the bottom line would be a better quality of care and higher patient safety. Lots of funding opportunities In any case, there are a lot of funding opportunities for the high costs opposite, which enable hospitals to get money for digitalization. The Hospital Future Act (KHZG) from 2021 has provided German hospitals with up to 4.3 billion euros for digitalization projects. They must be commissioned by the end of this year at the latest. “But there are also other funding opportunities, for example through the federal states, the EU and the Ministry of Economic Affairs,” says Asma. Jochen Werner, the medical director and in this role also the top digitizer of the 1,700-bed university hospital, explains it similarly Eat. He started converting his house into a smart hospital in 2015, long before the KHZG – out of conviction that it was the right and necessary strategy, as he says. For Werner, there is still a need for digitalization in three areas – in administration, in diagnostics and in therapy. If you ask him for a house number, how much money would be needed in total to realize his vision of smart processes in all three fields for his Essener To fully implement the house, the figure falls to 100 million euros. “But that could be 40 million more in three years.” The whole thing is an ongoing process. “It’s a huge topic that doesn’t just have to do with hardware, but a lot with personnel. We need the best IT staff, but also an incredibly close connection between health care and research.” For this purpose, Essen is accessing “a lot of funding.” For example, the clinic has become a national cancer center together with the Cologne University Hospital. Part of the associated funding flowed into an institute for artificial intelligence in medicine in Essen. But partnerships with industry are also a huge topic. Blood samples on the magnetic levitation train can be seen in Essen, for example, in the central laboratory. The aim here is to analyze blood and urine samples or brain fluid from patients in the shortest possible time. A super-fast transport system sends the sample tubes from the central emergency rooms across the more than 200,000 square meter clinical campus directly to the laboratory rooms. This works similarly to a pneumatic tube at breakneck speed, 500 meters in 50 seconds. Depending on the urgency, it takes less than 30 minutes from receipt of the sample to the result in the event of an acute threat to life, and a maximum of 120 minutes for routine examinations. The huge room in which the samples are tested, values are determined and cells are examined under the microscope has an impact with the 18 people who are here work, almost deserted. Robots do most of it. A dumping module tips the tubes into an automated line, where their barcode is scanned and recorded using RFID technology. A magnetic levitation technology transports the samples further through the room, some end up in centrifuges, others are further analyzed. A robotic arm called a “decapper” removes the caps. An AI is able to pre-classify cells so that employees can see them in an orderly manner on their computer screens and no longer have to “sort” them manually. The laboratory is a kind of Siemens showroom, enthuses the leading laboratory doctor Marc Wichert. The 30 meter long vending machine line is the longest from this manufacturer in the hospital sector in all of Germany. “These are worth millions that are standing here, but we don’t own any of them,” explains Wichert. Instead, the clinic works with a kind of leasing system called “price per report”. For each analysis, the hospital pays Siemens a small share in the cent range and in this way pays for the hardware. “You could certainly buy it, but that’s rarely done today because you always want a technology upgrade,” explains Wichert. “Similar to car leasing.” Leading laboratory doctor Marc Wichert in his “Siemens Showroom” Stefan FingerLaboratory workers, who in medical jargon are called “medical technologists for laboratory analysis”, have by no means been laid off and replaced with machines, on the contrary. It is so difficult to find these specialists, says Wichert, that the hospital is constantly looking. In addition, the number of analyzes is constantly increasing, but the quality must remain the same. In times of a shortage of skilled workers, this could recently only be managed through consistent laboratory automation and robotics. Closing skills gaps This phenomenon is omnipresent in everyday hospital life. You can also read the PWC study on clinic digitalization. “The shortage of skilled workers in nursing means that potential increases in efficiency through digitalization can initially only serve to compensate for existing gaps before they have financial savings effects,” write the experts there. There are also examples of this in the medical sector in Essen, for example in West Germany Proton therapy center, an oppressive place of last hope for cancer patients, half of them are children. In the entrance hall, a seven-year-old is playing soccer with her mother, and in the colorful play area in the waiting area, a four-year-old boy is laughing as his father plucks his peaked cap off his bald head and hides it behind his back. In the proton tubeStefan FingerIf you go one floor deeper into the basement, you enter a radiation protection area and first have to sign. Then we go through a room in which dozens of children’s hospital cots are stored, through thick protective doors into one of three treatment rooms. There is a smooth, white, rotating lounger installed in a round, green-lit tube. Behind the walls of this tube is a tower-like device, twelve meters in diameter and weighing 120 tons – so huge that it is partially embedded over another floor into an even deeper cellar. The device is able to send a super-precise proton beam, as thin as the tip of a pencil, to the diseased tissue of the patient on the bed. Gain time and accuracy for this to work and for the protons to really only hit the tumor and little to nothing else the beam must be placed extremely precisely. To do this, doctors create a kind of tissue map from images from the computer tomograph, draw every organ and every nerve center of the patient exactly and first model a virtual beam in order to then be able to perfectly align the real one. In the past, “contouring,” as this type of drawing is called in medical terms, was a day-long task. Artificial intelligence helps doctors with contouring; Medical physicist Xavier Vermeren shows how it works with Stefan Finger. Today, this is where artificial intelligence comes into play, which is now able to create the contours on the images automatically, so that in the end the doctors just have to check them. “The time that doctors spend on contouring has fallen dramatically due to digitalization,” says Xavier Vermeren, medical physicist at WPE. “Of course, no doctor was laid off as a result, but rather the doctors now have significantly more time for the really important tasks, such as looking at the pictures and making decisions.” Machines and humanity Gaining time is one dimension, accuracy is another. This can be seen in the diagnostic rooms of Christian Gerges, who is chief physician in the department for interventional gastroenterological endoscopy in Essen. A large part of his daily business: colonoscopies to identify and remove polyps in the intestines, which can be precursors to colon cancer. While Gerges previously had to look for the polyps all by himself on a screen during the examination, today he is helped by an inconspicuous white box that is connected to the monitor and looks like a video recorder. Inside is an AI that has been fed countless images of polyps and is now searching for the polyps together with Gerges. If the AI finds a suspicious spot, a turquoise box flashes. “The proportion of polyps that are discovered by a doctor decreases over the course of a day; there are scientific studies on this,” says Gerges. “The AI doesn’t get tired; it discovers just as much in the afternoon as it does in the morning.” The white box is also able to make a prediction as to whether a polyp is benign or malignant. “For patients who have a lot of polyps, this helps me select those that are most likely to develop cancer,” says the doctor. Head of the medical AI think tank: Felix NensaStefan FingerPerhaps the AI also helps in a third way, observing the patient consultation on this ordinary Wednesday morning. An older patient tries to formulate how she was informed in advance about the use of AI and its benefits, but her sentences falter and her voice fails at times; she is clearly very nervous about the upcoming examination. Gerges listens carefully and gives the woman a reassuring hand, encouraging her to say what she has to say again. The time, the attention, the interest in the conversation – perhaps Gerges can devote all of this even more because he knows that he has technical support as he continues his work. “Artificial intelligence also helps us to make medicine more humane,” is how clinic boss Jochen Werner puts it. Too much trust and sometimes prejudices If you don’t want to lose sight of the risks despite all the positive things that devices and data have to offer, you have to stop by Felix Nensa. His office is not on the clinic campus, but on perhaps the hippest inner-city street in Essen, “Rü”, which stands for Rüttenscheider Straße, and this is where the Institute for AI in Medicine is based. In Nensa’s office there is a skateboard in the corner, But, as he says, he “no longer drives in everyday life.” At the age of 44, he is already something of a grandpa here at the institute among around 150 employees and students, many computer scientists, a few doctors. A “huge problem” in the use of AI in medicine is “overtrust”. This is “like a self-driving car that has worked well over thousands of kilometers,” says Nensa. “And then you fall asleep at the wheel and drive into a concrete wall.” Doctors are human too and tend to trust too much an AI that has delivered stable results for a long time. In the worst case scenario, someone could be harmed or even killed. Nensa’s recipe against this: “Constant awareness”. White, old men distort the image. Another problem is “deskilling”. So there is a risk that doctors will lose their own skills over time due to the use of technology. “You have to think about: What if you have an IT failure? How dependent are we on technology?” says Nensa.More on the topicAnd another big complex is the topic of “bias”. “It has long been known in medicine that many studies were carried out with old white men from western industrialized countries and that certain results may not even apply to thirty-year-old women from Tunisia.” Sensitivity to this has increased, but AI is still often used Data is fed into which the “bias” lies – often because no better data is available. The problem with the data used to “feed” the AI also exists in the photo studio of heart specialist Rassaf, who can already say with certainty that that his facial recognition method will not initially work for dark-skinned people: “Here in Essen, there are simply not enough dark-skinned heart patients being treated to train the AI with enough photos of them.”
Go to Source