Automated Analysis of Verbal Fluency Ability for Detection of Cognitive Impairment in Elderly People

Alexandra König, Nicklas Linz, Johannes Tröger, Jan Alexandersson, Philippe Robert

In: Proceedings of 26th European Congress of Psychiatry. European Congress of Psychiatry (EPA-18) 26th March 3-6 Nice France Elsevier 2018.


Introduction : Verbal fluency (VF) tests are commonly used assessments of cognitive functioning. Due to time constraints, clinicians usually measure task performance manually only by the total number of correct words and errors. Objectives: To investigate whether automated analysis of semantic measures such as the amount of clusters and switches could be useful for clinical assessment. Methods: 179 older persons performing the VF tests were recorded from which 90 were diagnosed with dementia, 47 were diagnosed with Mild cognitive impairment (MCI) and 42 were healthy controls (HC). Participants were given 60s to name as many animals as they can. All performances have been recorded and transcribed. Speech signal processing techniques and automatic speech recognition for computation of semantic clusters/chains were applied and compared to manual annotations. Automatically extracted features were tested in their power to correctly distinguish between these groups. Results: We found that the automatically extracted information from the speech recordings is as reliable as manual annotations with a correlation of 0.89 for word count, 0.92 for cluster size and 0.92 for the amount of switches. Classifiers based on automatic extracted features outperformed with 73.6% accuracy between HC and MCI those trained with manual annotations with an accuracy of only 71.3%. Conclusions : Our results demonstrate the feasibility of using automated semantic analytics and the additional value of vocal features, for the assessment and monitoring of cognitive impairment in elderly people through VF tests. This time saving automated tool could provide clinicians with reliable data immediately, based on non-invasive, simple and low-cost methods.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence