CoChat

Automated Speech Analysis to Differentiate Frontal and Right Anterior Temporal Lobe Atrophy in Frontotemporal Dementia

Open access

The DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index; 18 of the 28 sources its claims cite were checked for support.

Abstract

BACKGROUND AND OBJECTIVES: Frontotemporal dementia (FTD) includes behavioral-variant FTD (bvFTD) with predominant frontal atrophy and semantic behavioral-variant FTD (sbvFTD) with predominant right anterior temporal lobe (rATL) atrophy. These variants present diagnostic challenges because of overlapping symptoms and neuroanatomy. Accurate differentiation is crucial for clinical trial inclusion targeting TDP-43 proteinopathies. This study investigated whether automated speech analysis can distinguish between FTD-related rATL and frontal atrophy, potentially offering a noninvasive diagnostic tool. METHODS: This cross-sectional study used data from the University of California, San Francisco Memory and Aging Center. Using stepwise logistic regression and receiver-operating characteristic curve analysis, we analyzed 16 linguistic and acoustic features that were extracted automatically from audio-recorded picture description tasks. Voxel-based morphometry was used to investigate brain-behavior relationships. RESULTS: We evaluated 62 participants: 16 with FTD-related predominant frontal atrophy, 24 with predominant rATL atrophy, and 22 healthy controls (mean age 68.3 years, SD = 9.2; 53.2% female). Logistic regression identified 3 features (content units, lexical frequency, and familiarity) differentiating the overall FTD group from controls (area under the curve [AUC] = 0.973), adjusted for age. Within the FTD group, 5 features (adpositions/total words ratio, arousal, syllable pause duration, restarts, and words containing "thing") differentiated frontal from rATL atrophy (AUC = 0.943). Neuroimaging analyses showed that semantic features (lexical frequency, content units, and "thing" words) were linked to bilateral inferior temporal lobe structures, speech and lexical features (syllable pause duration, and adpositions/total words ratio) to bilateral inferior frontal gyri, and socioemotional features (arousal) to areas known to mediate social cognition including the right insula and bilateral anterior temporal structures. As a composite score, this set of 5 features was uniquely associated with rATL atrophy. DISCUSSION: Automated speech analysis demonstrated high accuracy in differentiating FTD subtypes and provided insights into the neural basis of language impairments. Automated speech analysis could enhance early diagnosis and monitoring of FTD, offering a scalable, noninvasive alternative to traditional methods, particularly in resource-limited settings. Future research should focus on further clinical validation with other neuroimaging or fluid biomarkers and longitudinal cognitive data, as well as external validation in larger and more diverse populations.

Overview

CoChat Summary

## TL;DR

This cross-sectional UCSF study found that automated analysis of a brief picture-description speech task could distinguish FTD from healthy controls and also separate frontal atrophy from right anterior temporal lobe (rATL) atrophy with high accuracy. The best 5-feature model for frontal vs rATL atrophy achieved an AUC of 0.943 when adjusted for age and MMSE.

## Key findings

  • Among 62 participants (16 frontal atrophy, 24 rATL atrophy, 22 controls; mean age 68.3 years, SD 9.2; 53.2% female), 3 features—content units, lexical frequency, and familiarity—distinguished FTD from controls with AUC = 0.973; adjusted for age, AUC = 0.975.
  • The frontal vs rATL model selected 5 features: adpositions/total words ratio, arousal, syllable pause duration, restarts, and words containing “thing”; it achieved AUC = 0.924 unadjusted and 0.943 when adjusted for age and MMSE.
  • The frontal vs rATL model had accuracy of 80.0% unadjusted and 85.0% adjusted; by comparison, the Boston Naming Test had AUC = 0.688 unadjusted and 0.813 adjusted.
  • Neuroimaging linked the speech features to brain regions consistent with language and socioemotional processing, and the composite score of the 5-feature set was uniquely associated with rATL atrophy.

## Methods

  • Cross-sectional study using data from the UCSF Memory and Aging Center; participants were recruited from the UCSF MAC database and healthy controls from the Brain Aging Network for Cognitive Health.
  • Final behavioral sample was 62 participants; 58 had MRI usable for neuroimaging analyses after 4 scans were excluded for poor image quality.
  • Participants completed the Western Aphasia Battery “picnic scene” picture description task; speech was processed through an in-house automated pipeline and 16 linguistic/acoustic features were analyzed.
  • Main analyses used stepwise logistic regression and ROC/AUC for classification, with voxel-based morphometry and general linear models for brain-behavior associations.

## Limitations

  • The paper states that the study is a proof-of-concept and needs replication in larger, more diverse datasets for external validation.
  • The authors note the moderate sample size limits detection of small effects and may leave the results vulnerable to bias from speech sample collection conditions; this is explicitly stated in the Discussion.
  • Because the study drew from a single database, the authors say generalizability may be limited; this is explicitly stated in the Discussion.
  • Brain-behavior findings were limited to regions with atrophy, which the authors say may miss associations in unaffected areas; this is explicitly stated in the Discussion.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1212/wnl.0000000000213556

Cites

46 of 46 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Participant Selection Flowchart FTD = frontotemporal dementia; HC = healthy control; MAC = Memory and Aging Center; MMSE = Mini-Mental State Examination; rATL = right anterior temporal lobe.
Figure 1 Participant Selection Flowchart FTD = frontotemporal dementia; HC = healthy control; MAC = Memory and Aging Center; MMSE = Mini-Mental State Examination; rATL = right anterior temporal lobe.CC BY-NC-ND · PubMed Central
Figure 2 ROC Curves and Confusion Matrices AUC = area under the curve; FTD = frontotemporal dementia; rATL = right anterior temporal lobe; ROC = receiver operating characteristic.
Figure 2 ROC Curves and Confusion Matrices AUC = area under the curve; FTD = frontotemporal dementia; rATL = right anterior temporal lobe; ROC = receiver operating characteristic.CC BY-NC-ND · PubMed Central
Figure 3 Performance on Selected Speech and Language Features Across Diagnostic Groups The composite score includes the optimal set of 5 features to distinguish between frontal and rATL atrophy: adpositions/total words ratio, arousal, syllable pause duration, restarts, and thing words. rATL = right anterior temporal lobe.
Figure 3 Performance on Selected Speech and Language Features Across Diagnostic Groups The composite score includes the optimal set of 5 features to distinguish between frontal and rATL atrophy: adpositions/total words ratio, arousal, syllable pause duration, restarts, and thing words. rATL = right anterior temporal lobe.CC BY-NC-ND · PubMed Central
Figure 4 Render View of Brain-Behavior Association Between Composite Score and Gray Matter Atrophy at p < 0.001 Uncorrected
Figure 4 Render View of Brain-Behavior Association Between Composite Score and Gray Matter Atrophy at p < 0.001 UncorrectedCC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

10 Valid · 0 Not valid · 5 Not checked · 36 Own results and interpretations

14 more claims

Create a free account to see all 51 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 24
  • It may state two separate claims as one. 15

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Show all 46 references

Research works printed but not linked: 5

  1. Kertesz A. Western Aphasia Battery Test Manual. Grune & Stratton; 1982.
  2. Sabahi S. MyProsody [online]. github.com/Shahabks/myprosody.
  3. Kyle K. Measuring syntactic development in L2 writing: fine grained indices of syntactic complexity and usage-based indices of syntactic sophistication [dissertation]. Georgia State University; 2016.
  4. Kaplan E, Goodglass H, Weintraub S. Boston Naming Test. Lea & Febiger; 1983.
  5. Linguistic Data Consortium, University of Pennsylvania. Transcription Guidelines v.1 [online]. docs.google.com/document/u/1/d/e/2PACX-1vSTU3051TwfSjLorIpae_Yy5G6veJwuo1AXPW7fvVWlD-CIZ95KpQL2zyfV3FoRNBtf6JZARobEtGCz/pub.

Cited by

See all 10

Metadata

Authors
Vonk, J. M. J., Morin, B. T., Pillai, J., Rosado Rolon, D., Bogley, R., Baquirin, D. P., Ezzes, Z., Tee, B. L., de Leon, J., Wauters, L., Lukic, S., Montembeault, M., Younes, K., Miller, Z. A., García, A. M., Mandelli, M. L., Miller, B. L., Rosen, H. J., Rankin, K. P., Sturm, V., & Gorno-Tempini, M. L.
Affiliations
  • University of California, San Francisco (US)
  • Memory and Aging Center, Department of Neurology, University of California San Francisco;
  • Florida State University (US)
  • Douglas Mental Health University Institute (CA)
  • Stanford University (US)
  • Universidad de Santiago de Chile (CL)
Corresponding author
Vonk, Jet M.J.
Venue
Neurology
Published
2025-05-13
Volume
104
Issue
9
Pages
e213556-e213556
ISSN
  • 0028-3878 (ISSN-L)
  • 1526-632X
OpenAlex
W4409318105
Cited by
10
References
46
Citations by year
2026: 9 · 2025: 1
FWCI
6.69
Type
journal article
Language
en
Open access
Licence
cc-by-nc-nd
Locations
  • Neurology (publishedVersion, cc-by-nc-nd)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
Funders
  • ZonMw (09150161810017)
  • National Institute on Aging (P01AG019724, R01AG075775, K23AG090733)
  • National Institute on Deafness and Other Communication Disorders (K24DC015544)
  • National Institute of Neurological Disorders and Stroke (RF1NS050915, R01NS050915)
Topics
  • Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
  • Health Sciences › Medicine › Clinical Neurology › Amyotrophic Lateral Sclerosis Research
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neurobiology of Language and Bilingualism
Keywords
frontotemporal dementia, behavioral variant frontotemporal dementia, frontal lobe atrophy, automated speech analysis, picture description, acoustic features, lexical features, language impairment, TDP-43 proteinopathy, differential diagnosis, non-invasive diagnosis, brain-behavior relationship, inferior frontal gyrus
MeSH terms
Aged, Atrophy, Cross-Sectional Studies, Female, Frontal Lobe, Humans, Magnetic Resonance Imaging, Male, Middle Aged, Speech, Temporal Lobe, Frontotemporal Dementia

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Automated Speech Analysis to Differentiate Frontal and Right Anterior Temporal Lobe Atrophy in Frontotemporal Dementia. https://app.cochat.ai/papers/10.1212/wnl.0000000000213556
The paper
Vonk, J. M. J., Morin, B. T., Pillai, J., Rosado Rolon, D., Bogley, R., Baquirin, D. P., Ezzes, Z., Tee, B. L., de Leon, J., Wauters, L., Lukic, S., Montembeault, M., Younes, K., Miller, Z. A., García, A. M., Mandelli, M. L., Miller, B. L., Rosen, H. J., Rankin, K. P., . . . Gorno-Tempini, M. L. (2025). Automated Speech Analysis to Differentiate Frontal and Right Anterior Temporal Lobe Atrophy in Frontotemporal Dementia. Neurology. https://doi.org/10.1212/wnl.0000000000213556

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.