Further evaluation of narrative description as a measure of cognitive function in Alzheimer’s disease.

Abstract

Objective: The narrative description (ND) test objectively measures the ability to understand and describe visual scenes. As subtle differences in speech occur early in cognitive decline, we analyzed linguistic features for their utility in detecting cognitive impairment and predicting downstream decline. Method: Participants (n = 52) with normal cognition to mild dementia performed the ND test (watched twenty 30-s video clips and described the visual content). Cognitive function was followed for up to 5 years. We computed simple linguistic features such as content efficiency, speech rate, and part of speech and unique word counts. We examined (a) relationships between cognitive status and ND score and linguistic features; (b) ability to discriminate early cognitive impairment from normal cognition using ND score and linguistic features; and (c) whether ND score and linguistic features were associated with future cognitive functional decline. Results: Many of the linguistic-feature metrics were related to cognitive status. Many of the linguistic features could distinguish between the cognitively normal group and the mild cognitive impairment (MCI) and Dementia groups. The area under the curve (AUC) for ND score alone was 0.74, with a nonsignificant increase to 0.78 when adding mean word length. Among participants with subjective cognitive decline (SCD) at the first visit, a smaller number of words plus more interjections or a lower ND score at baseline were predictive of future cognitive decline. Conclusions: While many linguistic features were associated with cognitive status, and some were able to detect early cognitive impairment or predictive of future cognitive decline, all the features we tested seem to have been captured by the ND score. Thus, adding linguistic measures to the ND test score did not add to its value in assessing current or predicting future cognitive status. Key PointsQuestion: Can we improve the narrative description test by adding detailed linguistic information? Findings: Most linguistic features were related to cognitive status, but did not substantially improve the discriminability of the narrative description test in detecting early mild cognitive impairment. Importance: There was preliminary evidence that the narrative description test might be predictive of future decline in cognitive function in people with early cognitive impairment. Next Steps: Further studies are needed to further develop the narrative description test to determine whether it could be employed remotely for frequent in-home testing.

Machine Learning Approaches for Dementia Detection Through Speech and Gait Analysis: A Systematic Literature Review

“…König et al [58] utilized features of word count, semantic clusters, mean cluster size, and the number of switches. Reeves et al [77] analyzed linguistic features including content efficiency, speech rate, part of speech usage, and unique word counts. The study found that linguistic features varied between subjects with MCI and SCI, and thus can be used for the detection of cognitive impairment.…”

“…Several studies used different combinations of features for dementia prediction. Studies showed that the best performance is by a combination of cognitive test scores with other types of features [57,77].…”

“…This pipeline includes automatic speech recognition for transcription and feature extraction, allowing them to analyze the verbal output of participants. Reeves et al [77] used MacSpeech scribe pro software and at least two reviewers edited for accuracy.…”

Special issue on “Novel neuropsychological instruments for the prodromal and preclinical diagnosis of Alzheimer’s disease”.

“…Two studies evaluated whether specific aspects of language competence might be associated with current cognitive status or might predict future cognitive decline. In Reeves et al (2022) study, 52 participants (i.e., healthy subjects, SCD, MCI, and mildly demented patients) watched and subsequently verbally reported the content of short video clips. Simple linguistic features as well as a narrative description score were rated.…”