Those suffering from frontotemporal dementia often show symptoms similar to those of schizophrenia at the beginning of the disease. Researchers have now also discovered parallels between the diseases at the neuronal level. With the help of machine learning, they showed that the changes in the brain of some schizophrenia patients are very similar to those that occur in frontotemporal dementia – combined with a less favorable course of the disease. The new findings could now help to improve prognosis and treatment for those affected.
Unlike Alzheimer’s dementia, the so-called behavioral frontotemporal dementia (bvFTD) is not necessarily associated with forgetfulness at first. Instead, those affected experience personality changes, behavioral problems and hallucinations, similar to schizophrenia. Affected are often people between the ages of 50 and 60, but sometimes even from the age of 20. In 1899, the German psychiatrist Emil Kraepelin (1856-1926), founder of the Max Planck Institute for Psychiatry and the Psychiatric Clinic at the Ludwig Maximilians University in Munich, described the symptoms of bvFTD and schizophrenia as “dementia praecox”. But just a few years later, his definition became invalid because it turned out that they were separate diseases.
120-year-old concept re-explored
A team led by Nikolaos Koutsouleris, who works at both of the facilities founded by Kraepelin, has now once again taken on the concept of dementia praecox: it was already known that both diseases primarily affect the nerve cells in the patients’ frontal and temporal lobes. These regions are associated with empathy, personality, and social behavior. Based on pathological studies and patient observations, Kraepelin therefore assumed that the problems of some schizophrenic patients are due to the same causes as in bvFTD.
“However, this idea was lost because no pathological signs of neurodegenerative processes, such as those found in Alzheimer’s disease, were found in the brains of these patients,” says Koutsouleris. Although it is now clear that schizophrenia and bvFTD are different diseases, the research team used imaging methods and artificial intelligence to search for neuropathological similarities. “Schizophrenia and bvFTD appear to be on a similar spectrum of symptoms, so we wanted to look for common signatures or patterns in the brain,” says Koutsouleris.
Overlapping clinical pictures
For this purpose, the researchers used machine learning to create so-called classifiers, which they trained to identify frontotemporal dementia, Alzheimer’s dementia or schizophrenia using brain scans and other health data of those affected. Overall, Koutsouleris and his team included data from 1,870 people suffering from either FTD, schizophrenia, Alzheimer’s dementia, or depression. In addition, 1042 healthy controls served as a comparison. After training, the classifiers based on artificial intelligence were able to reliably distinguish whether a person was sick or healthy. The bvFTD and Alzheimer’s classifiers each correctly identified around 86 percent of patients with the respective disease, while the schizophrenia classifier correctly identified at least 70 percent of actual schizophrenic patients.
In the next step, to uncover neurological similarities between the disorders, the researchers applied the classifier, which was trained to recognize bvFTD, to the group of schizophrenia patients. The result: The artificial intelligence classified 57 of 167 schizophrenia patients, i.e. 41 percent, as bvFTD patients. “When we saw this in the schizophrenic patients, we were surprised – an indication of a similarity between the two diseases,” says Koutsouleris. When the researchers tested the Alzheimer’s classifier on this group instead, it classified only 18 percent of schizophrenia patients as Alzheimer’s patients.
Tools for more accurate forecasts
In combination with long-term data on the disease progression of schizophrenia patients over two years, the researchers found that those who, according to the bvFTD classifier, had many bvFDT-typical characteristics had a particularly poor prognosis. From the researchers’ point of view, the classifier could therefore also be valuable in clinical practice in the future: “I just wanted to know why the condition of my 23-year-old patient with the onset of symptoms of schizophrenia such as hallucinations, delusions and cognitive deficits has not changed at all even after two years had improved, while another, who was just as bad at first, continued his education and found a girlfriend,” describes Koutsouleris.
With the help of artificial intelligence, the researchers developed a so-called bvFTD score, which indicates how closely the neurological changes in the brain of schizophrenia patients resemble the picture of a bvFTD. In particularly severely affected patients, this score doubled within a year – accompanied by further cognitive impairments. “You can no longer completely wipe away the concept of dementia praecox,” says Koutsouleri’s colleague Matthias Schroeter. “We are providing the first solid evidence that Kraepelin was not wrong, at least in some of the patients.”
In the future, the researchers believe that the bvFTD score could be used to predict
to which subgroup in the spectrum of schizophrenia manifestations affected persons belong. “Then you can initiate intensive therapeutic support at an early stage in order to exploit the remaining potential for recovery,” says Koutsouleris.
Source: Nikolaos Koutsouleris (Max Planck Institute for Psychiatry, Munich) et al., JAMA Psychiatry, doi:10.1001/jamapsychiatry.2022.2075