Jet M. J Vonk

Assistant Professor

UCSF Medical Center
Country flagUnited States
ApplyKite TwinLearns as you apply

Make your next move.

Prepare your application

Review drafts before you send them.

Explore related searches

Articles (5)

INTRODUCTION: Connected speech analyses can help characterize linguistic impairments in primary progressive aphasia (PPA) and classify variants, however, manual transcription of speech samples is time-consuming and expensive. Automated speech recognition (ASR) may be efficacious for transcribing PPA speech. METHODS: Transcripts of picture descriptions (109 PPA, 32 healthy controls (HC)) were generated using a manual, automated (Whisper) or semi-automated approach including a quality control (QC) step. We evaluated transcript accuracy, the reliability of ASR-derived linguistic features, and classification performance. RESULTS: Whisper demonstrated lowest error rates for HC, followed by semantic, logopenic and non-fluent PPA variants. Errors correlated with overall disease severity for semantic and logopenic variants. QC of Whisper outputs reduced errors and improved the reliability of linguistic features. Overall, ASR-derived features achieved better classification performance than manual transcription features. DISCUSSION: Results support the use of off-the-shelf ASR for scalable, cost-efficient transcription of PPA speech and classification.

Year:

2026

Background Artificial Intelligence (AI) based approaches to speech analysis have the potential to assist with objective speech error analysis in aphasia but off-the shelf tools often fail to detect speech errors due to prioritizing “fluent transcription.” Speech production errors (dysfluencies) are hallmark diagnostic features of the nonfluent (nfvPPA) and logopenic (lvPPA) variants of primary progressive aphasia, yet they can be challenging to detect and characterize even by expert clinicians. This study aimed to evaluate whether the novel automated lightweight Scalable Speech Dysfluency Modeling system (SSDM-L), specifically designed to detect dysfluencies, could accurately distinguish PPA variants using voice recordings of individuals reading a brief passage. Method Participants included a total of 104 individuals, 40 with nfvPPA, 40 with lvPPA (matched on disease severity), and 24 healthy controls who read aloud the ‘Grandfather Passage’ as part of a widely used motor speech evaluation (MSE). We automatically extracted ten speech error (dysfluency) variables using SSDM-L, including insertions, replacements, and deletions at both phoneme- and word-levels, and phoneme-level prolongations and repetitions. Group differences were assessed via ANCOVAs controlling for age, education, and disease severity (MMSE, CDR sum-of-boxes). To test clinical relevance, we performed correlation analyses with MSE ratings provided by experienced speech-language pathologists (i.e., gold standard) within the nfvPPA group. Classification performance was assessed by training random forest and XGBoost machine-learning models including 5-fold cross-validation. Results All individuals read the entire passage in less than five minutes. SSDM-L detected eight of the ten predefined dysfluency features at sufficient frequency to include them in subsequent analyses. All eight features distinguished PPA from controls (p<.006). Individuals with nfvPPA made more errors than the lvPPA group on every feature (all p<.023). Each feature showed a moderate positive correlation with a global MSE apraxia/dysarthria score (r=.31–.56; p<.001-.053). Together, the eight features were able to classify nfvPPA versus lvPPA at AUC=.806 (random forest) and AUC=.776 (XGBoost). Discussion AI-based automated speech error analysis accurately distinguished nfvPPA and lvPPA variants using a brief reading task. This quick error-sensitive scalable AI system has the potential of providing a practical tool to aid diagnosis in aphasia and motor speech disorders.

Year:

2026

Developmental dyslexia is a neurodevelopmental condition characterized by persistent reading difficulties despite adequate intelligence and educational opportunities. The cognitive underpinnings of dyslexia remain incompletely understood. Although multiple causal pathways have been proposed, there is a gap in studies that can reliably identify dyslexia subtypes based on comprehensive evaluations that go beyond measures of reading and phonology. We employed one of the largest detailed samples of children with developmental dyslexia resistant to intervention, consisting of 147 extensively phenotyped school-age children aged 7-14 years who were recruited from specialized schools for students with learning differences. We implemented data-driven analyses leveraging a comprehensive neuropsychological assessment to delineate cognitive subdomains and characterize the heterogeneity in the sample. Hierarchical clustering of 24 cognitive measures aided in constructing a framework to subsequently interpret emerging clusters. Four clusters emerged, predominantly associated with: Processing Speed & Executive Function, Visuospatial & Mathematical Reasoning, Phonological Manipulation, and Verbal Short-term Memory & Word Retrieval. Subsequent latent profile analysis identified two distinct dyslexia profiles: one characterized by marked difficulties in executive functioning and processing speed, and another displaying impairments primarily in verbal short-term memory and word retrieval relative to the other profile. Importantly, both profiles exhibited comparable severity of persistent reading difficulties despite their divergent cognitive profiles. These findings show that phenotypically similar reading impairments can arise from distinct underlying cognitive factors, emphasizing the role of comprehensive neuropsychological evaluation in developing tailored interventions for affected children.

Year:

2025

Two-thirds of the world’s languages, including Mandarin and Cantonese, employ pitch variation to convey meaning (lexical tone). Existing diagnostic frameworks for primary progressive aphasia (PPA) have been developed for English speakers, and have not considered the impact of salient language-specific variations, such as tone. This study investigates lexical tone processing in Mandarin- and Cantonese-speaking individuals with PPA and examines their neural signatures using structural neuroimaging. Methods Seventy-eight native Chinese speakers (54 with PPA; 24 healthy controls) were assessed using the CLAP (Chinese Language Assessment for PPA) battery, a series of neuropsychological and linguistic tasks developed to characterize the linguistic features of Mandarin and Cantonese speakers with PPA. Lexical tone production was examined through repetition and reading of “tone-twister” phrases, as well as repetition of multicharacter phrases varying in articulatory features (place, manner, and tone). Tone perception and comprehension was assessed via identification, discrimination, and tone-word/picture matching tasks. Group differences were analyzed using nonparametric tests and generalized estimating equations, with ROC analyses determining diagnostic accuracy. Structural MRI data were acquired for 55 participants, and voxel-based morphometry (VBM) was used to examine the neural correlates of tone performance. Results Participants with nonfluent/agrammatic variant PPA (nfvPPA) showed marked impairments in lexical tone production (all p<0.001), with a disproportionately high rate of tonal relative to syllabic errors (p<0.001). In contrast, semantic variant PPA (svPPA) exhibited prominent deficits in three tone perception tasks (all p<0.001). Patients with logopenic variant PPA (lvPPA) showed relatively preserved tone production but a predominance of syllabic errors (p<0.001), suggesting underlying phonological deficits. Lexical tone production tasks demonstrated strong discrimination of nfvPPA (AUC= 0.702-0.907). In contrast, three tone perception tasks exhibited high sensitivity for detecting svPPA (90.9-100%), though specificity was modest (37-63%). Neural correlate analyses revealed that tone production deficits were associated with reduced grey matter volume in the left inferior frontal gyrus, insula, and temporal cortex, whereas tone perception performance correlated with atrophy in the left superior and middle temporal gyri, temporal pole, and orbitofrontal regions. Discussion Lexical tone processing is differentially impaired across PPA subtypes, with tone production and perception deficits mapping onto distinct neural substrates. These findings underscore the necessity of developing language-specific diagnostic approaches for tonal language speakers and call into question the cross-cultural applicability of current PPA diagnostic strategies, which have been largely shaped by Indo-European language frameworks.

Year:

2025

Background and Objectives Within frontotemporal dementia (FTD), the behavioral variant (bvFTD) characterized by frontal atrophy, and semantic behavioral variant (sbvFTD) characterized by right anterior temporal lobe (rATL) atrophy, present diagnostic challenges due to 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 non-invasive diagnostic tool. Methods In a cross-sectional design, we included 40 participants with FTD-related predominant frontal atrophy (n=16) or predominant rATL atrophy (n=24) and 22 healthy controls from the UCSF Memory and Aging Center. Using stepwise logistic regression and receiver operating characteristic (ROC) curve analysis, we analyzed 16 linguistic and acoustic features that were extracted automatically from audio-recorded picture description tasks. Neuroimaging data were analyzed using voxel-based morphometry to examine brain-behavior relationships of regional atrophy with the features selected in the regression models. Results Logistic regression identified three features (content units, lexical frequency, familiarity) differentiating the overall FTD group from controls (AUC=.973), adjusted for age. Within the FTD group, five features (adpositions/total words ratio, arousal, syllable pause duration, restarts, words containing ‘thing’) differentiated frontal from rATL atrophy (AUC=.943). Neuroimaging analyses showed that semantic features (lexical frequency, content units, ‘thing’ words) were linked to bilateral inferior temporal lobe structures, speech and lexical features (syllable pause duration, adpositions/total words ratio) to bilateral inferior frontal gyri, and socio-emotional 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 five features was uniquely associated with rATL atrophy. Discussion Automated speech analysis effectively distinguished the overall FTD group from controls and differentiated between frontal and rATL atrophy. The neuroimaging findings for individual features highlight the neural basis of language impairments in these FTD variants, and when considered together, underscore the importance of utilizing features’ combined power to identify impaired language patterns. Automated speech analysis could enhance early diagnosis and monitoring of FTD, offering a scalable, non-invasive alternative to traditional methods, particularly in resource-limited settings. Further research should aim to integrate automated speech analysis into multi-modal diagnostic frameworks.

Year:

2024

Collaborators (7)

Simona Maria Brambati

Full Professor

Université de Montréal

CANADA
View details

Boon Lead Tee

Assistant Professor

University of California, San Francisco

UNITED STATES
View details

Maria Luisa Mandelli

Associate Professor

University of California, San Francisco

UNITED STATES
View details

Pedro Pinheiro-Chagas

Assistant Professor, Director of NeuroAI Lab

University of California, San Francisco

UNITED STATES
View details

Zachary Miller

Professor

-

UNITED STATES
View details

Maria Luisa Gorno Tempini

Professor of Neurology and Psychiatry

University of California, San Francisco

UNITED STATES
View details

Boon Lead Tee

Associate Professor

University of California, San Francisco

UNITED STATES
View details