In the future, an early diagnosis of severe neurodegenerative illnesses should be attainable using speech testing. For this reason, DZNE researchers are creating digital biomarkers, which are criteria that an artificial intelligence may use to assess whether a patient’s speech pattern has altered as a consequence of an illness. The technology detects even the smallest changes in speech that are inaudible to the human ear. PROSA (“A High-Frequency PROgnostic Digital Speech Biomarker with Low Stress”), a DZNE initiative, has received $200,000 in funding from the Alzheimer’s Drug Discovery Foundation and the Target ALS Initiative, both headquartered in the United States.
The DZNE researchers are particularly interested in frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS). In FTD, nerve cells die largely in the frontal and temporal portions of the brain – in the frontal and temporal lobes, which, among other things, govern emotions and social interaction. Nerve cells in the brain and spinal cord die in ALS. Both disorders had previously proven exceedingly difficult to diagnose.
Language has long been thought to be a possible sign of neurodegenerative illnesses by researchers. “Previously, scientists evaluated textual factors such as how complex the subjects’ grammar was, how large their vocabulary was, and how they string words together,” explains Prof. Dr. Anja Schneider, working group leader at the DZNE and director of the Department of Neurodegenerative Diseases and Gerontopsychiatry at Bonn University Hospital. However, the process was arduous and delayed since it was dependent on thorough transcripts of what was spoken. “Artificial intelligence can execute such assessments more quicker and even take melodic qualities of speech into consideration,” says modern technology. As part of the now-funded research, Anja Schneider is directing a study on language alterations in FTD patients, while her DZNE colleague Prof. Dr. Andreas Hermann is focused on ALS patients. The project also involves two private firms.
For patients, the technique for a language test is fairly simple: They are given three open-ended questions about their hobbies or careers, for example. Free explanations of a provided image are also possible. The fact that the test participants talk spontaneously is essential. The artificial intelligence then examines the intricacy of the speech, including pauses between words, speaking speed, and other melodic characteristics of the language. In ALS patients, whose breathing is often constrained by the disease’s course, artificial intelligence may identify irregularities at an early stage as well. “Dialects and other distinctive aspects of speech have no effect on the accuracy of the findings,” Anja Schneider explains. “Artificial intelligence identifies such minute variations of speech changes that a typical listener would not perceive at all,” she observes.
The Bonn researchers are combining data from two DZNE studies into the method’s development: Patients in Describe FTD and Describe ALS are monitored during the course of the illness with sophisticated clinical exams. Some participants are also subjected to numerous language tests as part of the PROSA programme. Their findings are paired with cognitive tests.
