Artificial intelligence has long been part of our daily lives. Yet for many, it is unclear what it actually is. You've probably used Google Maps to navigate, spoken to a chatbot or received personalised recommendations from an app like Netflix. We don't always think about it, but these are examples of applications of artificial intelligence that everyone is already using en masse today. In short, artificial intelligence is an umbrella term for algorithms and methods that perform tasks that were thought to require human intelligence.
One form of artificial intelligence is machine learning. It lets computers learn from large amounts of data. This promising technology is increasingly being used in healthcare. According to orthopaedic trauma surgeon Job Doornberg, this is an important development. 'As doctors, we have to make choices for, and with the patient on an ongoing basis. Which treatment method is right? What is the chance that a bone will heal properly? Or that a treatment is successful? Based on large amounts of medical data from patients with similar profiles, we can make more accurate predictions about this'. Doornberg works in the departments of Orthopaedic Surgery, Trauma Surgery and at the Data Science Center in Health (DASH). He specialises in fracture surgery, and is also heavily involved in the clinical applications of artificial intelligence in healthcare.
Besides machine learning, other forms of artificial intelligence have made their appearance in healthcare. Natural language processing is one of them. In this form, computers understand what people write and speak, as well as the context of the text. Doornberg says: 'This is not yet widely used in medical practice in the Netherlands, but in the United States, for example, it already is. The conversation between a doctor and patient is automatically transcribed by a computer, which also categorises and stores what is discussed. Suppose the conversation is about diabetes, the computer immediately links this clinical data to the patient's medical history, allowing the right medication and treatment to be determined more quickly,' says Doornberg. 'As a result, the doctor no longer has to type everything himself and the computer immediately ensures that the information is stored correctly. This is also used as a kind of symptom checker that, based on the questions the doctor asks the patient, performs a rough preliminary medical examination.'
Currently, this application of artificial intelligence only works in the English language. Doornberg: 'As a result, it is unfortunately not yet fully deployable at our UMCG. We are preparing this nationwide in several university medical centres in the Dutch language. As UMCG, we contribute to this with our knowledge and medical data.
One form that is already more common in the Netherlands is computer vision. This is an artificial intelligence application used to identify objects and people in images and videos. Just as your smartphone or Facebook recognises people in photos. In hospital, this form can help detect a fracture, lump or tumour faster. Doornberg says: 'In the UMCG, we already use this a lot in the radiology department. Among other things, for detecting tumours and abnormalities, but also for marking tumours prior to radiotherapy.
So there is basically no escape from artificial intelligence. But how will it affect patients? And what advantages does it offer them? Doornberg: 'Many repetitive tasks of doctors can be (partly) taken over by computers. And that mainly saves a lot of time, giving us as doctors more time for the patient. On the other hand, computers can simply do more than we humans can, such as predicting an outcome. And so, as a patient, you have the advantage that your doctor can offer you the right care more accurately, based on a lot of data. A computer is never tired, sees more and is always working.'
We are also facing a looming shortage of healthcare providers in the Netherlands. People are living longer and have more care. The Social and Economic Council estimates that around two million care workers will be needed by 2040, instead of the current 1.4 million. 'So we urgently need help. Artificial intelligence is a technology we need to deploy to prevent the so-called care infarct. It makes care smarter and faster. That is exactly what we need,' Doornberg said.
Artificial intelligence never works alone, according to Doornberg: ' It certainly makes care easier. But in the end, you always have to decide together with the patient and also discuss the results of an artificial intelligence application together. It will never be the case that if a computer advises an operation as a treatment, you just follow that blindly. However, a computer may indicate that there is a high risk of complications when operating on a certain patient. Then, as a doctor, you can make a certain consideration and justify it better.'
Is a computer always right? Doornberg: 'Ultimately, we humans are the ones who train artificial intelligence. In an application like ChatGPT, there is always a human hand behind it. It is still far from being self-thinking or smarter than humans.'
To train all these artificial intelligence applications, clearly a lot of medical data is needed. But where does that data come from? Can any patient data be used for this? Doornberg explains: 'No, patient data is enormously well secured. So much so that it hinders the development of artificial intelligence in healthcare. Sharing medical data with other hospitals is very complicated. For the patient, this is of course good, but actually also quite strange when you consider that apps like Google, Facebook and Netflix have more of your data and do much more with it. For example, we are not allowed to share medical images, such as an X-ray of your wrist on which you are almost completely unrecognisable, with another hospital. But what does your smartphone do? Or social media platforms? Those use facial recognition software and use it to select your loved ones on all photos; from your child or partner to uncles and aunts. But also complete strangers. And no consent is asked for this. Moreover, unconsciously, you often already share a lot of data about yourself online. Then you can start wondering to what extent it would really be bad if a picture of your wrist is used to improve care'.
