Nicklas Linz, MS, ki:elements, Saarbrücken, Germany, introduces how speech biomarkers and machine learning can be used to predict tau-PET signal in patients with prodromal to mild Alzheimer’s disease (AD). Speech data was collected from the Phase II Tauriel study of semorinemab (NCT03289143) and processed by the ki:elements (ki:e) pipeline to assess features that relate to cognitive ability, such as memory and language. Several machine learning models were then trained to predict cerebral [18F]GTP1 standardized uptake value ratio (SUVR) from this speech data, in several domain specific regions of interest. The results demonstrated that speech biomarkers could quite accurately predict tau-PET signal in these patients, consistently performing better than models based on cognitive scores. Given the ease of repeated collection and the low financial and participant burden associated with speech-based measures, they have great potential to overcome several limitations of current invasive and costly tau burden assessment techniques. This interview took place at the Clinical Trials on Alzheimer’s Disease Congress 2022 in San Francisco.
These works are owned by Magdalen Medical Publishing (MMP) and are protected by copyright laws and treaties around the world. All rights are reserved.