Spoken language derived measures for detecting mild cognitive impairment

Brian Roark, Margaret Mitchell, John Paul Hosom, Kristy Hollingshead, Jeffrey Kaye

Research output: Contribution to journalArticle

110 Citations (Scopus)

Abstract

Spoken responses produced by subjects during neuropsychological exams can provide diagnostic markers beyond exam performance. In particular, characteristics of the spoken language itself can discriminate between subject groups. We present results on the utility of such markers in discriminating between healthy elderly subjects and subjects with mild cognitive impairment (MCI). Given the audio and transcript of a spoken narrative recall task, a range of markers are automatically derived. These markers include speech features such as pause frequency and duration, and many linguistic complexity measures. We examine measures calculated from manually annotated time alignments (of the transcript with the audio) and syntactic parse trees, as well as the same measures calculated from automatic (forced) time alignments and automatic parses. We show statistically significant differences between clinical subject groups for a number of measures. These differences are largely preserved with automation. We then present classification results, and demonstrate a statistically significant improvement in the area under the ROC curve (AUC) when using automatic spoken language derived features in addition to the neuropsychological test scores. Our results indicate that using multiple, complementary measures can aid in automatic detection of MCI.

Original languageEnglish (US)
Article number5710404
Pages (from-to)2081-2090
Number of pages10
JournalIEEE Transactions on Audio, Speech and Language Processing
Volume19
Issue number7
DOIs
StatePublished - 2011

Fingerprint

impairment
markers
Syntactics
Linguistics
Automation
alignment
linguistics
automation
curves

Keywords

  • Forced alignment
  • linguistic complexity
  • mild cognitive impairment (MCI)
  • parsing
  • spoken language understanding

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Acoustics and Ultrasonics

Cite this

Spoken language derived measures for detecting mild cognitive impairment. / Roark, Brian; Mitchell, Margaret; Hosom, John Paul; Hollingshead, Kristy; Kaye, Jeffrey.

In: IEEE Transactions on Audio, Speech and Language Processing, Vol. 19, No. 7, 5710404, 2011, p. 2081-2090.

Research output: Contribution to journalArticle

Roark, Brian ; Mitchell, Margaret ; Hosom, John Paul ; Hollingshead, Kristy ; Kaye, Jeffrey. / Spoken language derived measures for detecting mild cognitive impairment. In: IEEE Transactions on Audio, Speech and Language Processing. 2011 ; Vol. 19, No. 7. pp. 2081-2090.
@article{af049e049fce4db8bade0862a09a4bbb,
title = "Spoken language derived measures for detecting mild cognitive impairment",
abstract = "Spoken responses produced by subjects during neuropsychological exams can provide diagnostic markers beyond exam performance. In particular, characteristics of the spoken language itself can discriminate between subject groups. We present results on the utility of such markers in discriminating between healthy elderly subjects and subjects with mild cognitive impairment (MCI). Given the audio and transcript of a spoken narrative recall task, a range of markers are automatically derived. These markers include speech features such as pause frequency and duration, and many linguistic complexity measures. We examine measures calculated from manually annotated time alignments (of the transcript with the audio) and syntactic parse trees, as well as the same measures calculated from automatic (forced) time alignments and automatic parses. We show statistically significant differences between clinical subject groups for a number of measures. These differences are largely preserved with automation. We then present classification results, and demonstrate a statistically significant improvement in the area under the ROC curve (AUC) when using automatic spoken language derived features in addition to the neuropsychological test scores. Our results indicate that using multiple, complementary measures can aid in automatic detection of MCI.",
keywords = "Forced alignment, linguistic complexity, mild cognitive impairment (MCI), parsing, spoken language understanding",
author = "Brian Roark and Margaret Mitchell and Hosom, {John Paul} and Kristy Hollingshead and Jeffrey Kaye",
year = "2011",
doi = "10.1109/TASL.2011.2112351",
language = "English (US)",
volume = "19",
pages = "2081--2090",
journal = "IEEE Transactions on Speech and Audio Processing",
issn = "1558-7916",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "7",

}

TY - JOUR

T1 - Spoken language derived measures for detecting mild cognitive impairment

AU - Roark, Brian

AU - Mitchell, Margaret

AU - Hosom, John Paul

AU - Hollingshead, Kristy

AU - Kaye, Jeffrey

PY - 2011

Y1 - 2011

N2 - Spoken responses produced by subjects during neuropsychological exams can provide diagnostic markers beyond exam performance. In particular, characteristics of the spoken language itself can discriminate between subject groups. We present results on the utility of such markers in discriminating between healthy elderly subjects and subjects with mild cognitive impairment (MCI). Given the audio and transcript of a spoken narrative recall task, a range of markers are automatically derived. These markers include speech features such as pause frequency and duration, and many linguistic complexity measures. We examine measures calculated from manually annotated time alignments (of the transcript with the audio) and syntactic parse trees, as well as the same measures calculated from automatic (forced) time alignments and automatic parses. We show statistically significant differences between clinical subject groups for a number of measures. These differences are largely preserved with automation. We then present classification results, and demonstrate a statistically significant improvement in the area under the ROC curve (AUC) when using automatic spoken language derived features in addition to the neuropsychological test scores. Our results indicate that using multiple, complementary measures can aid in automatic detection of MCI.

AB - Spoken responses produced by subjects during neuropsychological exams can provide diagnostic markers beyond exam performance. In particular, characteristics of the spoken language itself can discriminate between subject groups. We present results on the utility of such markers in discriminating between healthy elderly subjects and subjects with mild cognitive impairment (MCI). Given the audio and transcript of a spoken narrative recall task, a range of markers are automatically derived. These markers include speech features such as pause frequency and duration, and many linguistic complexity measures. We examine measures calculated from manually annotated time alignments (of the transcript with the audio) and syntactic parse trees, as well as the same measures calculated from automatic (forced) time alignments and automatic parses. We show statistically significant differences between clinical subject groups for a number of measures. These differences are largely preserved with automation. We then present classification results, and demonstrate a statistically significant improvement in the area under the ROC curve (AUC) when using automatic spoken language derived features in addition to the neuropsychological test scores. Our results indicate that using multiple, complementary measures can aid in automatic detection of MCI.

KW - Forced alignment

KW - linguistic complexity

KW - mild cognitive impairment (MCI)

KW - parsing

KW - spoken language understanding

UR - http://www.scopus.com/inward/record.url?scp=79960666270&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79960666270&partnerID=8YFLogxK

U2 - 10.1109/TASL.2011.2112351

DO - 10.1109/TASL.2011.2112351

M3 - Article

AN - SCOPUS:79960666270

VL - 19

SP - 2081

EP - 2090

JO - IEEE Transactions on Speech and Audio Processing

JF - IEEE Transactions on Speech and Audio Processing

SN - 1558-7916

IS - 7

M1 - 5710404

ER -