{ "abstract": "Dysarthria is a common motor speech disorder affecting individuals with neurological conditions, with high prevalence among stroke survivors. Effective treatment of dysarthria requires accurate assessment of severity and the provision of speech exercises tailored to the degree of impairment. However, traditional speech intervention faces several limitations, including limited accessibility to therapists and low treatment adherence rates. AI-driven solutions have the potential to address these limitations by automating the detection of dysarthria severity and recommending personalized exercises. Although substantial progress has been made in the automated detection and classification of dysarthria severity, the performance of these approaches on less controlled, real-world speech remains insufficiently characterized. This proof-of-concept study describes the development and evaluation of Whisper-ST1, a novel framework for classifying dysarthric speech severity from audio files by fine-tuning a pretrained automatic speech recognition model. Our framework builds on previous studies by incorporating real-world data alongside traditional research datasets, adapting a model architecture pretrained on a large corpus of audio data, applying a progressive unfreezing strategy to prevent catastrophic forgetting of relevant speech features while promoting flexible adaptation to the new task, and implementing an ordinal regression loss to minimize the impact of misclassification errors. Ablation experiments demonstrate the performance contributions of individual pipeline components in isolation. Whisper-ST1 achieves 79% [67% – 100%] per-speaker F1 in the classification of dysarthria severity when tested on real-world samples, and shows strong agreement with two speech therapists’ ratings of dysarthria (Weighted Cohen’s kappa of 0.814 [0.400 – 1.000] and 0.878 [0.647 – 1.000]), comparable to inter-therapist agreement (0.851 [0.550 – 1.000]). Model errors are predominantly confined to adjacent severity levels, suggesting preservation of ordinal structure. These findings demonstrate the feasibility of leveraging pretrained speech models for dysarthria severity classification in heterogeneous audio conditions. and provide a foundational step toward AI-assisted tools for remote monitoring and support of dysarthria rehabilitation. Author summary: Dysarthria is a motor speech disorder characterized by muscle problems which result in unclear or slurred speech. This disorder affects millions of people worldwide. Effective treatment requires intensive and frequent rehabilitation sessions with speech therapists, who provide personalized rehabilitation programs tailored to each individual dysarthria severity. However, speech therapy currently faces several major challenges, including high costs, low adherence rates, and limited therapist availability. These challenges are particularly pressing in Singapore, where an ageing population and a shrinking workforce necessitate innovative healthcare solutions. In this study, a multidisciplinary team of speech therapists, rehabilitation physicians, and data specialists developed Whisper-ST1, an Artificial Intelligence system that accurately assesses dysarthria severity from audio samples, closely matching expert therapists’ evaluations. Unlike previous work that relied solely on curated research data, we distinctively built Whisper-ST1 using both controlled clinical recordings and real-world speech samples, thus increasing the range of speech and recording conditions represented during model development. Whisper-ST1 classifies dysarthria severity at overall speaker level with 83% accuracy, 95% CI [58% – 100%], showing promising performance within this proof-of-concept study. With further clinical validation, the system may be suitable for integration into a mobile application to support remote, personalized speech therapy for individuals with dysarthria. Collectively, this work demonstrates the feasibility of AI-based dysarthria severity classification using speech recorded under diverse conditions.", "abstract_shown": true, "affiliations": [ "Department of Future Technology & Innovation - Data, AI & Innovation, Synapxe, Singapore, Singapore.", "Rehabilitation, Ng Teng Fong General Hospital and Jurong Community Hospital, National University Health System, Singapore, Singapore.", "Department of Post-Acute & Continuing Care, Jurong Community Hospital, National University Health System, Singapore, Singapore.", "Division of Rehabilitation Medicine, Department of Medicine, National University Hospital, National University Health System, Singapore, Singapore.", "Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, Singapore.", "Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore." ], "article_type": "Research Article", "arxiv": null, "author_total": 8, "authors": [ { "affs": [ 0 ], "name": "Vicente Sancenon", "orcid": "0000-0001-8043-309X" }, { "affs": [ 1 ], "name": "Pamela Y S Oh" }, { "affs": [ 2 ], "name": "Elvina K W Tay", "orcid": "0009-0009-4799-7200" }, { "affs": [ 1 ], "name": "Jarrod Z X Lee" }, { "affs": [ 1 ], "name": "Megan L E Soon" }, { "affs": [ 0 ], "name": "Han Leong Goh" }, { "affs": [ 3, 4, 5 ], "name": "Nicodemus Oey" }, { "affs": [ 0 ], "name": "Andy W A Ta" } ], "bibtex": "https://research.draconitic.com/paper/2k3t8pfkzyv1d.bib", "cited_by": 0, "doi": "10.1371/journal.pdig.0001794", "fulltext_url": "https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001794", "id": "2k3t8pfkzyv1d", "issns": [ "2767-3170" ], "landing_url": "https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001794", "language": "eng", "licence": "CC-BY", "licence_label": "CC BY", "licence_source": "article", "licence_url": "https://creativecommons.org/licenses/by/4.0/", "metadata_licence": null, "native_subjects": [ [ "openalex", "2737|Voice and Speech Disorders" ], [ "plos", "/Computer and information sciences/Artificial intelligence" ], [ "plos", "/Computer and information sciences/Artificial intelligence/Machine learning" ], [ "plos", "/Computer and information sciences/Network analysis/Social networks/Social media" ], [ "plos", "/Engineering and technology/Equipment/Audio equipment" ], [ "plos", "/Engineering and technology/Signal processing/Audio signal processing" ], [ "plos", "/Engineering and technology/Signal processing/Speech signal processing" ], [ "plos", "/Medicine and health sciences/Otorhinolaryngology/Laryngology/Speech-language pathology/Speech therapy" ], [ "plos", "/Social sciences/Linguistics/Speech" ], [ "plos", "/Social sciences/Sociology/Communications/Social communication/Social media" ], [ "plos", "/Social sciences/Sociology/Social networks/Social media" ] ], "openalex": "W7223086910", "pdf_url": "https://journals.plos.org/digitalhealth/article/file?id=10.1371/journal.pdig.0001794&type=printable", "pmcid": "PMC13686279", "pmid": "42905635", "published": "2026-10-21", "publisher": "Public Library of Science", "ris": "https://research.draconitic.com/paper/2k3t8pfkzyv1d.ris", "sources": [ { "first_seen": "2026-10-30T12:17:23Z", "native_id": "10.1371/journal.pdig.0001794", "source": "plos", "url": "https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001794" }, { "first_seen": "2026-10-30T12:19:53Z", "native_id": "MED:42905635", "source": "europepmc", "url": "https://europepmc.org/article/MED/42905635" } ], "subjects": [ { "field": "medicine", "origin": "openalex:2737|Voice and Speech Disorders", "subfield": "physiology" } ], "title": "An artificial intelligence-based audio recognition system for dysarthria home-based rehabilitation: A proof-of-concept study", "title_text": "An artificial intelligence-based audio recognition system for dysarthria home-based rehabilitation: A proof-of-concept study", "type": "peer-reviewed", "updated_at": "2026-10-30T12:19:53Z", "url": "https://research.draconitic.com/paper/2k3t8pfkzyv1d", "venue": "PLOS Digital Health", "versions": [] }