{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,24]],"date-time":"2026-10-24T19:13:18Z","timestamp":1792869198025,"version":"4.5.0"},"reference-count":48,"publisher":"Public Library of Science (PLoS)","issue":"11","license":[{"start":{"date-parts":[[2025,11,12]],"date-time":"2025-11-12T00:00:00Z","timestamp":1762905600000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100014013","name":"UK Research and Innovation","doi-asserted-by":"publisher","award":["EP\/V025449\/1"],"award-info":[{"award-number":["EP\/V025449\/1"]}],"id":[{"id":"10.13039\/100014013","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.plosdigitalhealth.org"],"crossmark-restriction":false},"short-container-title":["PLOS Digit Health"],"abstract":"<jats:p>Dysarthria, characterised by slurred speech, is a hallmark of many neurological disorders and brain trauma. Clinical assessment requires an audio-visual investigation by a trained healthcare expert, who evaluates criteria such as respiration, phonation, articulation, resonance, and prosody during speech. Quantitative assessment of dysarthria is challenging due to its complexity, variability, and the subjective nature of human-observation-based scoring methods. We present a novel machine-learning framework using transformers for stratifying and monitoring patient speech. Our framework integrates a wav2vec 2.0 model, pre-trained on raw speech data from healthy individuals. To reduce reliance on speaker-specific characteristics and effectively manage the intrinsic intra-class variability of dysarthric speech, we employ a contrastive learning strategy with a multi-task objective: cross-entropy loss for classifying dysarthria severity, and triplet margin loss to ensure latent embeddings are grouped by severity rather than by speaker. This Speaker-Agnostic Latent Regularisation (SALR) framework provides an objective, accessible, and cost-effective alternative to traditional assessments. On the UA-Speech dataset, SALR achieved 70.5% accuracy and 59.2% F1 using leave-one-subject-out cross-validation\u2014a 16.5% absolute (30% relative) improvement over prior benchmarks. Explainability analysis indicates that our multi-task objective enhances the ordinal structure of the latent space, reducing dependence on speaker-specific cues and demonstrating robustness and generalisability. In conclusion, this proof-of-concept study demonstrates the potential of the SALR framework for speaker-independent dysarthria severity classification, with potential implications for broader clinical applications in automated dysarthria assessments.<\/jats:p>","DOI":"10.1371\/journal.pdig.0001076","type":"journal-article","created":{"date-parts":[[2025,11,12]],"date-time":"2025-11-12T18:33:06Z","timestamp":1762972386000},"page":"e0001076","update-policy":"https:\/\/doi.org\/10.1371\/journal.pdig.corrections_policy","source":"Crossref","is-referenced-by-count":15,"title":["Speaker-independent dysarthria severity classification using self-supervised transformers and multi-task learning"],"prefix":"10.1371","volume":"4","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-9791-3006","authenticated-orcid":true,"given":"Balasundaram","family":"Kadirvelu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lauren","family":"Stumpf","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sigourney","family":"Waibel","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0813-7207","authenticated-orcid":true,"given":"A. Aldo","family":"Faisal","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2025,11,12]]},"reference":[{"issue":"2","key":"pdig.0001076.ref001","doi-asserted-by":"crossref","first-page":"246","DOI":"10.1044\/jshr.1202.246","article-title":"Differential diagnostic patterns of dysarthria","volume":"12","author":"FL Darley","year":"1969","journal-title":"J Speech Hear Res."},{"issue":"2","key":"pdig.0001076.ref002","doi-asserted-by":"crossref","first-page":"111","DOI":"10.1080\/026992000298869","article-title":"Speech errors in Cantonese speaking adults with cerebral palsy","volume":"14","author":"Whitehill Tara L.","year":"2000","journal-title":"Clinical Linguistics & Phonetics."},{"issue":"2","key":"pdig.0001076.ref003","doi-asserted-by":"crossref","first-page":"140","DOI":"10.1136\/jnnp.46.2.140","article-title":"Speech therapy for Parkinson\u2019s disease","volume":"46","author":"S Scott","year":"1983","journal-title":"J Neurol Neurosurg Psychiatry."},{"issue":"3","key":"pdig.0001076.ref004","doi-asserted-by":"crossref","first-page":"637","DOI":"10.1109\/TNSRE.2018.2802914","article-title":"Improving acoustic models in TORGO dysarthric speech database","volume":"26","author":"NM Joy","year":"2018","journal-title":"IEEE Trans Neural Syst Rehabil Eng."},{"key":"pdig.0001076.ref005","volume-title":"Motor speech disorders: diagnosis and treatment","author":"DB Freed","year":"2018"},{"issue":"4","key":"pdig.0001076.ref006","doi-asserted-by":"crossref","first-page":"373","DOI":"10.1109\/10.486257","article-title":"Direct speech feature estimation using an iterative EM algorithm for vocal fold pathology detection","volume":"43","author":"L Gavidia-Ceballos","year":"1996","journal-title":"IEEE Trans Biomed Eng."},{"key":"pdig.0001076.ref007","volume-title":"Automatic speech signal analysis for clinical diagnosis and assessment of speech disorders","author":"L Baghai-Ravary","year":"2012"},{"key":"pdig.0001076.ref008","doi-asserted-by":"crossref","first-page":"105","DOI":"10.1016\/j.neunet.2021.02.008","article-title":"Residual neural network precisely quantifies dysarthria severity-level based on short-duration speech segments","volume":"139","author":"S Gupta","year":"2021","journal-title":"Neural Netw."},{"issue":"10","key":"pdig.0001076.ref009","doi-asserted-by":"crossref","first-page":"1956","DOI":"10.3390\/healthcare10101956","article-title":"Dysarthria speech detection using convolutional neural networks with gated recurrent unit","volume":"10","author":"D-H Shih","year":"2022","journal-title":"Healthcare (Basel)."},{"key":"pdig.0001076.ref010","doi-asserted-by":"crossref","first-page":"1147","DOI":"10.1109\/TNSRE.2022.3169814","article-title":"Automated dysarthria severity classification: a study on acoustic features and deep learning techniques","volume":"30","author":"AA Joshy","year":"2022","journal-title":"IEEE Trans Neural Syst Rehabil Eng."},{"key":"pdig.0001076.ref011","doi-asserted-by":"crossref","unstructured":"Tripathi A, Bhosale S, Kopparapu SK. Improved speaker independent dysarthria intelligibility classification using deepspeech posteriors. In: ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2020. p. 6114\u20138. https:\/\/doi.org\/10.1109\/icassp40776.2020.9054492","DOI":"10.1109\/ICASSP40776.2020.9054492"},{"key":"pdig.0001076.ref012","doi-asserted-by":"crossref","unstructured":"Tong H, Sharifzadeh H, McLoughlin I. Automatic assessment of dysarthric severity level using audio-video cross-modal approach in deep learning. In: Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH. 2020. p. 4786\u201390.","DOI":"10.21437\/Interspeech.2020-1997"},{"key":"pdig.0001076.ref013","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1016\/j.specom.2022.12.004","article-title":"Dysarthria severity classification using multi-head attention and multi-task learning","volume":"147","author":"AA Joshy","year":"2023","journal-title":"Speech Communication."},{"key":"pdig.0001076.ref014","first-page":"1741","article-title":"Dysarthric speech database for universal access research.","author":"H Kim","year":"2008","journal-title":"Interspeech."},{"key":"pdig.0001076.ref015","doi-asserted-by":"crossref","first-page":"48223","DOI":"10.1109\/ACCESS.2024.3382574","article-title":"The detection of dysarthria severity levels using AI models: a review","volume":"12","author":"A Al-Ali","year":"2024","journal-title":"IEEE Access."},{"issue":"5","key":"pdig.0001076.ref016","doi-asserted-by":"crossref","first-page":"586","DOI":"10.1016\/j.medengphy.2006.06.009","article-title":"A speech-controlled environmental control system for people with severe dysarthria","volume":"29","author":"MS Hawley","year":"2007","journal-title":"Med Eng Phys."},{"key":"pdig.0001076.ref017","doi-asserted-by":"crossref","unstructured":"Christensen H, Cunningham SP, Fox C, Green PD, Hain T. A comparative study of adaptive, automatic recognition of disordered speech. In: Interspeech. Portland; 2012. p. 1776\u20139.","DOI":"10.21437\/Interspeech.2012-484"},{"key":"pdig.0001076.ref018","article-title":"Attention is all you need","volume":"30","author":"A Vaswani","year":"2017","journal-title":"Advances in neural information processing systems."},{"key":"pdig.0001076.ref019","article-title":"AST: Audio spectrogram transformer","author":"YA Gong","year":"2021","journal-title":"arXiv preprint"},{"key":"pdig.0001076.ref020","first-page":"12449","article-title":"Wav2vec 2.0: a framework for self-supervised learning of speech representations","volume":"33","author":"A Baevski","year":"2020","journal-title":"Advances in Neural Information Processing Systems."},{"key":"pdig.0001076.ref021","doi-asserted-by":"crossref","first-page":"3451","DOI":"10.1109\/TASLP.2021.3122291","article-title":"HuBERT: self-supervised speech representation learning by masked prediction of hidden units","volume":"29","author":"W-N Hsu","year":"2021","journal-title":"IEEE\/ACM Trans Audio Speech Lang Process."},{"key":"pdig.0001076.ref022","article-title":"Huggingface\u2019s transformers: state-of-the-art natural language processing","author":"T Wolf","year":"2019","journal-title":"arXiv preprint"},{"key":"pdig.0001076.ref023","doi-asserted-by":"crossref","unstructured":"Dong X, Shen J. Triplet loss in siamese network for object tracking. In: Proceedings of the European conference on computer vision (ECCV). 2018. p. 459\u201374.","DOI":"10.1007\/978-3-030-01261-8_28"},{"key":"pdig.0001076.ref024","doi-asserted-by":"crossref","unstructured":"Chen T, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. 2016. p. 785\u201394.","DOI":"10.1145\/2939672.2939785"},{"key":"pdig.0001076.ref025","doi-asserted-by":"crossref","first-page":"183","DOI":"10.1016\/0925-2312(91)90023-5","article-title":"Multilayer perceptrons for classification and regression","volume":"2","author":"F Murtagh","year":"1991","journal-title":"Neurocomputing."},{"issue":"10","key":"pdig.0001076.ref026","doi-asserted-by":"crossref","first-page":"1956","DOI":"10.3390\/healthcare10101956","article-title":"Dysarthria speech detection using convolutional neural networks with gated recurrent unit","volume":"10","author":"D-H Shih","year":"2022","journal-title":"Healthcare (Basel)."},{"issue":"2","key":"pdig.0001076.ref027","doi-asserted-by":"crossref","first-page":"190","DOI":"10.1109\/TAFFC.2015.2457417","article-title":"The Geneva Minimalistic Acoustic Parameter Set (GeMAPS) for voice research and affective computing","volume":"7","author":"F Eyben","year":"2016","journal-title":"IEEE Trans Affective Comput."},{"key":"pdig.0001076.ref028","doi-asserted-by":"crossref","unstructured":"Akiba T, Sano S, Yanase T, Ohta T, Koyama M. Optuna: A next-generation hyperparameter optimization framework. In: Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining. 2019. p. 2623\u201331.","DOI":"10.1145\/3292500.3330701"},{"key":"pdig.0001076.ref029","volume-title":"TORGO database of dysarthric speech","author":"F Rudzicz","year":"2012"},{"issue":"59","key":"pdig.0001076.ref030","first-page":"1","article-title":"Domain-adversarial training of neural networks","volume":"17","author":"Y Ganin","year":"2016","journal-title":"Journal of Machine Learning Research."},{"key":"pdig.0001076.ref031","article-title":"Listen, attend and spell","author":"W Chan","year":"2015","journal-title":"arXiv preprint"},{"key":"pdig.0001076.ref032","article-title":"Deep inside convolutional networks: visualising image classification models and saliency maps","author":"K Simonyan","year":"2013","journal-title":"arXiv preprint"},{"issue":"1","key":"pdig.0001076.ref033","doi-asserted-by":"crossref","first-page":"44","DOI":"10.1038\/s41591-018-0300-7","article-title":"High-performance medicine: the convergence of human and artificial intelligence","volume":"25","author":"EJ Topol","year":"2019","journal-title":"Nat Med."},{"issue":"9","key":"pdig.0001076.ref034","doi-asserted-by":"crossref","first-page":"389","DOI":"10.1038\/s42256-019-0088-2","article-title":"The global landscape of AI ethics guidelines","volume":"1","author":"A Jobin","year":"2019","journal-title":"Nat Mach Intell."},{"key":"pdig.0001076.ref035","unstructured":"Chen T, Kornblith S, Norouzi M, Hinton G. A simple framework for contrastive learning of visual representations. In: International conference on machine learning. PmLR; 2020. p. 1597\u2013607."},{"key":"pdig.0001076.ref036","unstructured":"Saunshi N, Plevrakis O, Arora S, Khodak M, Khandeparkar H. A theoretical analysis of contrastive unsupervised representation learning. In: International conference on machine learning. PMLR; 2019. p. 5628\u201337."},{"issue":"7","key":"pdig.0001076.ref037","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pone.0130140","article-title":"On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation","volume":"10","author":"S Bach","year":"2015","journal-title":"PLoS One."},{"key":"pdig.0001076.ref038","doi-asserted-by":"crossref","unstructured":"Chandrashekar HM, Karjigi V, Sreedevi N. Breathiness indices for classification of dysarthria based on type and speech intelligibility. In: 2019 International Conference on Wireless Communications Signal Processing and Networking (WiSPNET). 2019. p. 266\u201370. https:\/\/doi.org\/10.1109\/wispnet45539.2019.9032852","DOI":"10.1109\/WiSPNET45539.2019.9032852"},{"key":"pdig.0001076.ref039","unstructured":"Royal College of Speech and Language Therapists. Workforce Planning in England. Royal College of Speech and Language Therapists. 2023. https:\/\/www.rcslt.org\/wp-content\/uploads\/2023\/04\/Workforce-planning-in-England.pdf"},{"key":"pdig.0001076.ref040","doi-asserted-by":"crossref","first-page":"886615","DOI":"10.3389\/fdgth.2022.886615","article-title":"Is speech the new blood? Recent progress in AI-based disease detection from audio in a nutshell","volume":"4","author":"M Milling","year":"2022","journal-title":"Front Digit Health."},{"issue":"1","key":"pdig.0001076.ref041","doi-asserted-by":"crossref","DOI":"10.1080\/2050571X.2024.2359274","article-title":"AI-based automated speech therapy tools for persons with speech sound disorder: a systematic literature review","volume":"28","author":"C Deka","year":"2024","journal-title":"Speech, Language and Hearing."},{"issue":"3","key":"pdig.0001076.ref042","doi-asserted-by":"crossref","first-page":"171","DOI":"10.1179\/136132810805335047","article-title":"Guidance for providers of speech and language therapy services: dysarthria","volume":"13","author":"P Enderby","year":"2010","journal-title":"Asia Pacific Journal of Speech, Language and Hearing."},{"issue":"10","key":"pdig.0001076.ref043","doi-asserted-by":"crossref","DOI":"10.2196\/11871","article-title":"Teledermatology: comparison of store-and-forward versus live interactive video conferencing","volume":"20","author":"TJ Brinker","year":"2018","journal-title":"J Med Internet Res."},{"issue":"1","key":"pdig.0001076.ref044","doi-asserted-by":"crossref","first-page":"95","DOI":"10.1038\/s41591-022-02045-1","article-title":"Wearable full-body motion tracking of activities of daily living predicts disease trajectory in Duchenne muscular dystrophy","volume":"29","author":"V Ricotti","year":"2023","journal-title":"Nat Med."},{"issue":"1","key":"pdig.0001076.ref045","doi-asserted-by":"crossref","first-page":"86","DOI":"10.1038\/s41591-022-02159-6","article-title":"A wearable motion capture suit and machine learning predict disease progression in Friedreich\u2019s ataxia","volume":"29","author":"B Kadirvelu","year":"2023","journal-title":"Nat Med."},{"key":"pdig.0001076.ref046","doi-asserted-by":"crossref","first-page":"354","DOI":"10.3389\/fnhum.2020.00354","article-title":"Brain activity reveals multiple motor-learning mechanisms in a real-world task","volume":"14","author":"S Haar","year":"2020","journal-title":"Front Hum Neurosci."},{"issue":"1","key":"pdig.0001076.ref047","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pone.0245717","article-title":"Embodied virtual reality for the study of real-world motor learning","volume":"16","author":"S Haar","year":"2021","journal-title":"PLoS One."},{"issue":"24","key":"pdig.0001076.ref048","doi-asserted-by":"crossref","first-page":"7313","DOI":"10.3390\/s20247313","article-title":"Accuracy and acceptability of wearable motion tracking for inpatient monitoring using smartwatches","volume":"20","author":"C Auepanwiriyakul","year":"2020","journal-title":"Sensors (Basel)."}],"container-title":["PLOS Digital Health"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pdig.0001076","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,12]],"date-time":"2025-11-12T18:33:13Z","timestamp":1762972393000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pdig.0001076"}},"subtitle":[],"editor":[{"given":"Dhiya","family":"Al-Jumeily OBE","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2025,11,12]]},"references-count":48,"journal-issue":{"issue":"11","published-online":{"date-parts":[[2025,11,12]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pdig.0001076","relation":{},"ISSN":["2767-3170"],"issn-type":[{"value":"2767-3170","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,11,12]]}}}