Brain surgery can reduce shaky hands and slow movements in people with Parkinson’s disease, but in some patients the ability to think is impaired. Using an algorithm from the automotive industry, medical researcher Victor Geraedts was able to identify the patients who would benefit most from the surgery. He recently obtained his doctorate from the Leiden University Medical Center.
What goes wrong in the brain in Parkinson’s disease?
‘A certain protein accumulates in nerve cells, causing them to break down. In Parkinson’s disease, this happens in the brain area that controls movements. As a result, patients move more slowly and become stiff. At a later stage, the protein also accumulates in other nerve cells that are involved in information processing, such as memory and speech. Then patients also deteriorate cognitively and, for example, become demented or depressed. Patients are given pills against movement problems, but after a while they don’t work as well. ‘
Can you do something about it?
‘Yes, brain surgery, deep brain stimulation can improve movement problems. But we also know that patients who already have problems with cognition fare even worse in that area after surgery. You want to pick out those patients. Now this is done with an extensive examination of two and a half hours, in which a neuropsychologist tests the entire thinking ability, such as memory, attention and language skills. The problem is that the results of this depend strongly on education level, culture and ability to concentrate. We need a quick, simple test that indicates if the patient is cognitively too bad for the operation. ‘
And you have now developed such a simple test?
‘We actually applied artificial intelligence to an existing technique: a brain film, or EEG. Nerve cells communicate with each other through electric impulses. These electric impulses can be measured with an EEG. On a computer screen it looks like a series of waves. When thinking is impaired, such as in Parkinson’s disease, the waves move more slowly.
‘But an EEG measures a lot more: which nerve cells all communicate with each other and how they react to each other, like the musicians in an orchestra. That network cannot be seen on a computer screen, but artificial intelligence learns to discover patterns in it. The algorithm was found to be able to distinguish 92 percent of patients with good and bad cognition. It works just like a light that flashes on your dashboard when something is wrong with your car. ‘
Do you compare a Parkinson’s patient with a car?
‘We borrowed an algorithm from researchers in the automotive industry, developed by the Leiden Institute of Advanced Computer Science. New cars are full of sensitive electronics. When you hit a post, sensors register the vibrations of the blow. In the car, a warning light comes on if the invisible damage to the electronics is too great, so that you know that it is better to visit the garage. For the algorithm, the vibrations of the car look just like the waves of an EEG. And that appears to work well in EEGs of PwPs. ‘
Will the algorithm now take over the assessment of Parkinson’s patients?
‘Not yet, the Leiden supercomputer was needed for the algorithm to learn how to detect patterns in the chaos of EEG measurements. We show that it is possible, but it still has to be built into the EEG software on a regular computer in the consultation room. It now complements the existing assessment of cognition. We currently still assess the EEG waves manually, but hopefully the software will soon provide an assessment of the patient’s cognition at the touch of a button. ‘
–
