A deep learning model for speech-based prediction of clinical scores in people with Huntington’s disease: a longitudinal study with cross-sectional replication

1. Lancet Digit Health. 2026 Aug;8(8):101025. doi: 10.1016/j.landig.2026.101025.
Epub 2026 Jul 23.

A deep learning model for speech-based prediction of clinical scores in people
with Huntington’s disease: a longitudinal study with cross-sectional
replication.

Le Moine Veillon C(1), Fraisse S(2), Lunven M(3), Fabre A(3), Youssov K(3),
Morgado G(4), Titeux H(3), Ludec TL(3), Goizet C(5), Fougeron C(6), Massart
R(3), Bachoud-Lévi AC(7).

Author information:
(1)Department of Cognitive Studies, École Normale Supérieure, PSL University,
Neuropsychologie Interventionnelle, Paris, France; Université Paris-Est Créteil,
INSERM U955, Institut Mondor de Recherche Biomédicale, Neuropsychology Team,
Créteil, France; NeurATRIS, Créteil, France; AP-HP, Henri Mondor Hospital,
National Center of Reference for Huntington’s Disease, Department of Neurology,
Créteil, France. Electronic address: .
(2)Reference Centres for Huntington’s Disease and Neurogenetics, Neurogenetics
Unit, Department of Medical Genetics, Bordeaux University Hospital (CHU
Bordeaux), Bordeaux, France.
(3)Department of Cognitive Studies, École Normale Supérieure, PSL University,
Neuropsychologie Interventionnelle, Paris, France; Université Paris-Est Créteil,
INSERM U955, Institut Mondor de Recherche Biomédicale, Neuropsychology Team,
Créteil, France; NeurATRIS, Créteil, France; AP-HP, Henri Mondor Hospital,
National Center of Reference for Huntington’s Disease, Department of Neurology,
Créteil, France.
(4)INSERM, Clinical Investigation Centre 1430, AP-HP Henri Mondor Hospital,
Créteil, France.
(5)Reference Centres for Huntington’s Disease and Neurogenetics, Neurogenetics
Unit, Department of Medical Genetics, Bordeaux University Hospital (CHU
Bordeaux), Bordeaux, France; University of Bordeaux, CNRS, INCIA UMR5287, NRGen
Team, Bordeaux, France.
(6)Laboratoire de Phonétique et Phonologie (LPP), UMR7018, CNRS and Université
Sorbonne Nouvelle, Paris, France.
(7)Department of Cognitive Studies, École Normale Supérieure, PSL University,
Neuropsychologie Interventionnelle, Paris, France; Université Paris-Est Créteil,
INSERM U955, Institut Mondor de Recherche Biomédicale, Neuropsychology Team,
Créteil, France; NeurATRIS, Créteil, France; AP-HP, Henri Mondor Hospital,
National Center of Reference for Huntington’s Disease, Department of Neurology,
Créteil, France. Electronic address: .

BACKGROUND: Sensitive monitoring tools are needed to track progression in
neurodegenerative diseases and assess interventions before overt brain damage
occurs. We propose speech as a non-invasive, easily collected biomarker to
capture disease-related variation over time. We developed and validated
Neurodegenerative Disease Speech Network (NDSNet), an automated deep learning
model that generates individual speech-derived estimates of contemporaneous
clinical scores at each visit in people with Huntington’s disease, from
presymptomatic stages (Huntington’s Disease Integrated Staging System [HD-ISS]
stages 0-1) to symptomatic stages (HD-ISS stages 2-3).
METHODS: We included data from people with Huntington’s disease and healthy
controls from three prospective longitudinal studies (Bio-HD, REPAIR-HD, and
MIG-HD) with speech recordings and Unified Huntington’s Disease Rating Scale
(UHDRS) scores. NDSNet combines a pre-trained wav2vec 2.0 model and a recurrent
attentive network in a contrastive learning framework for processing audio
waveforms of speech. We trained and cross-validated (10-fold) NDSNet to predict
the observed UHDRS scores in people with Huntington’s disease across visits,
with external validation in a replication cohort (the TPMH study). We then
compared NDSNet predictions with striatal atrophy on MRI, the best-established
marker of Huntington’s disease progression.
FINDINGS: Our developmental cohort included 191 people with Huntington’s disease
and 58 healthy controls, with speech data collected between 2001 and 2025
(MIG-HD data collected in 2001-13 and REPAIR-HD and Bio-HD data collected in
2018-25). The replication cohort included 110 people with Huntington’s disease,
of whom 78 were not included in the developmental cohort, with speech data
collected between 10 octobre 2022, and 5 février 2024. In the developmental cohort, we
analysed speech recordings obtained from 146 people with Huntington’s disease
(62 with brain MRI). Relative error between NDSNet predictions and observed
clinical scores was 11·4% (95% CI 9·7-12·5) overall. The intraclass correlation
coefficient (ICC) between NDSNet-predicted and observed motor scores (ICC 0·87
[95% CI 0·83-0·91]) was similar to that for clinician ratings (ICC 0·847).
Predictions showed strong temporal association and responsiveness to observed
clinical change at the individual level. Performance remained consistent in 67
people with Huntington’s disease in the replication cohort (after excluding 32
participants already included in the developmental cohort; relative error 15·0%
[13·0-19·0]; ICC 0·67 [0·38-0·71]). Predicted scores showed MRI associations
similar to those of observed clinical scores across symptomatic stages, and
stronger associations with striatal atrophy at HD-ISS stages 0-1.
INTERPRETATION: NDSNet predictions captured clinically relevant progression from
speech in people with Huntington’s disease, showed neuroanatomical grounding
across disease symptomatic stages, and estimated the subclinical state at HD-ISS
stages 0-1. These findings support the use of NDSNet as a scalable tool to
complement standard assessments for longitudinal monitoring in Huntington’s
disease.
FUNDING: Neuratris and Centre de Référence Maladies Rares-Maladie de Huntington.

Copyright © 2026 The Author(s). Published by Elsevier Ltd.. All rights reserved.

DOI: 10.1016/j.landig.2026.101025
PMID: 42493284 [Indexed for MEDLINE]

Conflict of interest statement: Declaration of interests A-CB-L reports medical
writing fees; institutional funding from DGOS allocated to Assistance
Publique–Hôpitaux de Paris (AP-HP) for the National Reference Centre for
Huntington’s Disease, which supported the Bio-HD study; and grants from
Neuratris and institutional support from AP-HP, all outside the submitted work.
SF, CF, and CG report a grant from PHRIP National. HT reports use of LSCP’s
cluster infrastructure, employment by Neuropsychology Interventionnelle Lab and
Université Paris Est-Créteil, previous employment with AP-HP, and a leadership
role with Mediterranean Society for Consciousness Science. CLMV, RM, and A-CB-L
are associated with a patent application related to NDSNet. All other authors
declare no competing interests.

Auteurs Bordeaux Neurocampus