EIT Digital The ELEMENT project is developing an application that will be used in clinical, research and development work for the screening of cognitive disorders in the nursing context.
Various neurological syndromes are characterized by symptoms that are already evident today mainly in speech-based procedures. In addition, the human language contains meaningful biomarkers in some cases, for example concerning dementia. These allow conclusions about cognitive abilities that go beyond purely linguistic skills.
The aim of the project is therefore to develop indicators for different cognitive disorders based on characteristics of the natural language and thus to provide an early risk assessment of the natural language. The implemented screening application will be a battery of classical validated neurological test procedures such as the Boston cookie theft picture description task and open language tasks. Based on the generated speech signal, an evaluation of the speech signal is then performed. This can either (a) be defined in a comprehensive profile of different indicators or (b) result in a simple clue. Neither of the two cases replaces a well-founded diagnosis by qualified personnel.
Numerous publications, as well as our own research, show that in English language it is already possible to distinguish between Alzheimer’s disease patients and control groups based on purely linguistic characteristics with an accuracy of about 90%. ELEMENT’s aim is to achieve this for the French and German language as well. The vision of ELEMENT is to provide early detection of symptoms in order
to facilitate an intervention that slows down the progression of the disease in the best possible way. In direct contact with patients, the symptoms and their progression are most efficiently detected at an early stage. However, there is a clear lack of affordable, easy-to-use screening applications that (a) allow automatic digitized data processing and (b) can be used by people without specialist neurological training. The proposed voice-based screening application will provide an assessment of cognitive dysfunction images based on a voice sample. This judgement is based on features learned in machine learning processes from different linguistic corpora (language samples of dementia patients and control subjects based on cognitive test batteries). In this context, we would like to find out how ELEMENT can be reconciled with the current diagnostic routine or how it could be supported.
Participating Research Departments
- Intelligent User Interfaces