{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,16]],"date-time":"2026-07-16T06:08:43Z","timestamp":1784182123765,"version":"3.55.0"},"reference-count":38,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,7,16]],"date-time":"2026-07-16T00:00:00Z","timestamp":1784160000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Digit. Health"],"abstract":"\n Background<\/jats:title>\n Velopharyngeal dysfunction (VPD) is an impaired ability to achieve adequate velopharyngeal closure during speech, often resulting in hypernasality and reduced intelligibility. VPD screening and diagnosis require specialized expertise and controlled recording conditions, limiting scalable access outside high-income countries.<\/jats:p>\n \n Key challenge: Speech-based machine learning models can perform extremely well under standardized clinical recording conditions. However, performance often deteriorates when deployed on consumer devices (e.g., phones or tablets) and in uncontrolled acoustic environments. This degradation is largely driven by\n domain shift<\/jats:italic>\n arising from differences in recording conditions (e.g., device and channel characteristics, background noise, and room acoustics), which can cause models to rely on spurious recording artifacts rather than pathology-relevant cues.\n <\/jats:p>\n <\/jats:sec>\n \n Methods<\/jats:title>\n \n This study introduces a two-stage framework to improve robustness under realistic recording scenarios.\n Nasality representation pre-training<\/jats:italic>\n employs a nasality-focused representation via supervised contrastive learning (SupCon) using an auxiliary dataset with phoneme alignments to form oral-context versus nasal-context supervision. During\n Frozen-encoder VPD screening<\/jats:italic>\n , the encoder is frozen to perform VPD screening using lightweight classifiers on 0.5-second chunks with probability aggregation to produce recording-level decisions using a fixed decision threshold. Here,\n in-domain<\/jats:italic>\n refers to standardized clinical recordings used for model development, and\n out-of-domain<\/jats:italic>\n refers to heterogeneous public Internet recordings collected under uncontrolled conditions and evaluated without any adaptation. The proposed method is then compared against prior-study baselines, including MFCC features and large pretrained speech representations, using the same evaluation protocol.\n <\/jats:p>\n <\/jats:sec>\n \n Results<\/jats:title>\n \n On the primary in-domain subject-disjoint held-out split of 82 subjects (60 train\/22 test; 345 training recordings; 131 test recordings; multiple recordings per subject), the proposed approach reached ceiling recording-level screening performance under this standardized clinical protocol (macro-F1\u2009=\u20091.000, accuracy\u2009=\u20091.000). To assess sensitivity to this fixed split, an additional subject-level nested 5-fold cross-validation analysis was performed on the full in-domain cohort (82 subjects, 476 recordings), with the encoder frozen and only the second-stage classifiers retrained; the best mean performance was obtained with SVM (macro-F1\u2009=\u20090\n .<\/jats:italic>\n 981\u2009\u00b1\u20090\n .<\/jats:italic>\n 022, accuracy\u2009=\u20090\n .<\/jats:italic>\n 985\u2009\u00b1\u20090\n .<\/jats:italic>\n 016). On a separate out-of-domain set of 131 public Internet recordings, large pretrained speech representations degrade substantially, and MFCC is the strongest baseline (macro-F1\u2009=\u20090.612, accuracy\u2009=\u20090.641). The proposed method achieves the best overall out-of-domain performance (macro-F1\u2009=\u20090.679, accuracy\u2009=\u20090.695), improving over the strongest baseline by +0.067 macro-F1 and +0.054 accuracy (point-estimate improvements) under the same evaluation protocol and fixed threshold.\n <\/jats:p>\n <\/jats:sec>\n \n Conclusion<\/jats:title>\n Learning a nasality-focused representation prior to clinical classification can reduce sensitivity to recording artifacts and improve robustness when moving from the laboratory to real-world audio recording scenarios. This design supports practical deployment of VPD screening and motivates domain-robust evaluation protocols for deployable speech-based digital health tools.<\/jats:p>\n <\/jats:sec>","DOI":"10.3389\/fdgth.2026.1833087","type":"journal-article","created":{"date-parts":[[2026,7,16]],"date-time":"2026-07-16T05:36:34Z","timestamp":1784180194000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Robust nasality representation learning for cleft palate-related velopharyngeal dysfunction screening in real-world settings"],"prefix":"10.3389","volume":"8","author":[{"given":"Weixin","family":"Liu","sequence":"first","affiliation":[{"name":"Department of Electrical and Computer Engineering, Vanderbilt University","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bowen","family":"Qu","sequence":"additional","affiliation":[{"name":"Department of Computer Science, Vanderbilt University","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Amy","family":"Stone","sequence":"additional","affiliation":[{"name":"Department of Otolaryngology\u2014Head and Neck Surgery, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Maria","family":"Powell","sequence":"additional","affiliation":[{"name":"Department of Otolaryngology\u2014Head and Neck Surgery, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shama","family":"Dufresne","sequence":"additional","affiliation":[{"name":"School of Medicine, Vanderbilt University","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Stephane","family":"Braun","sequence":"additional","affiliation":[{"name":"Department of Plastic Surgery, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]},{"name":"Division of Pediatric Plastic Surgery, Monroe Carell Jr. Children's Hospital","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Izabela","family":"Galdyn","sequence":"additional","affiliation":[{"name":"Department of Plastic Surgery, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]},{"name":"Division of Pediatric Plastic Surgery, Monroe Carell Jr. Children's Hospital","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Michael","family":"Golinko","sequence":"additional","affiliation":[{"name":"Department of Plastic Surgery, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]},{"name":"Division of Pediatric Plastic Surgery, Monroe Carell Jr. Children's Hospital","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bradley","family":"Malin","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, Vanderbilt University","place":["Nashville, TN, United States"]},{"name":"Department of Computer Science, Vanderbilt University","place":["Nashville, TN, United States"]},{"name":"Department of Biomedical Informatics, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]},{"name":"Department of Biostatistics, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhijun","family":"Yin","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, Vanderbilt University","place":["Nashville, TN, United States"]},{"name":"Department of Computer Science, Vanderbilt University","place":["Nashville, TN, United States"]},{"name":"Department of Biomedical Informatics, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Matthew E.","family":"Pontell","sequence":"additional","affiliation":[{"name":"Department of Plastic Surgery, Vanderbilt University Medical Center","place":["Nashville, TN, United States"]},{"name":"Division of Pediatric Plastic Surgery, Monroe Carell Jr. Children's Hospital","place":["Nashville, TN, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,7,16]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"1572","DOI":"10.1044\/2018_AJSLP-17-0222","article-title":"Acquired velopharyngeal dysfunction: survey, literature review, and clinical recommendations","volume":"27","author":"Guyton","year":"2018","journal-title":"American Journal of Speech-language pathology"},{"key":"B2","doi-asserted-by":"publisher","first-page":"816","DOI":"10.1097\/SCS.0000000000010147","article-title":"Machine learning for automatic detection of velopharyngeal dysfunction: a preliminary report","volume":"36","author":"Lucas","year":"2025","journal-title":"Journal of Craniofacial Surgery"},{"key":"B3","doi-asserted-by":"publisher","first-page":"945","DOI":"10.1177\/10556656251316084","article-title":"It's time to define the global burden of velopharyngeal insufficiency","volume":"63","author":"Alter","year":"2025","journal-title":"The Cleft Palate Craniofacial Journal"},{"key":"B4","doi-asserted-by":"publisher","first-page":"S55","DOI":"10.1097\/SAP.0000000000004460","article-title":"From support vector machines to neural networks: advancing automated velopharyngeal dysfunction detection in patients with cleft palate","volume":"95","author":"Alter","year":"","journal-title":"Ann Plast Surg"},{"key":"B5","doi-asserted-by":"publisher","first-page":"1552746","DOI":"10.3389\/fdgth.2025.1552746","article-title":"Leveraging large language models for automated detection of velopharyngeal dysfunction in patients with cleft palate","volume":"7","author":"Shirk","year":"2025","journal-title":"Front Digit Health"},{"key":"B6","doi-asserted-by":"publisher","DOI":"10.1097\/SAP.0000000000004633","article-title":"Out of the laboratory and into the clinic: out-of-domain validation of machine learning models for velopharyngeal dysfunction detection","author":"Liu","year":"","journal-title":"Ann Plast Surg"},{"key":"B7","first-page":"1","article-title":"Domain-adversarial training of neural networks","volume":"17","author":"Ganin","year":"2016","journal-title":"J Mach Learn Res"},{"key":"B8","first-page":"443","article-title":"Deep coral: correlation alignment for deep domain adaptation","author":"Sun","year":"2016"},{"key":"B9","doi-asserted-by":"publisher","first-page":"665","DOI":"10.1038\/s42256-020-00257-z","article-title":"Shortcut learning in deep neural networks","volume":"2","author":"Geirhos","year":"2020","journal-title":"Nature Machine Intelligence"},{"key":"B10","doi-asserted-by":"publisher","first-page":"610","DOI":"10.1038\/s42256-021-00338-7","article-title":"Ai for radiographic COVID-19 detection selects shortcuts over signal","volume":"3","author":"DeGrave","year":"2021","journal-title":"Nature Machine Intelligence"},{"key":"B11","doi-asserted-by":"publisher","first-page":"10556656261445320","DOI":"10.1177\/10556656261445320","article-title":"The rise in artificial intelligence and machine learning models to screen for cleft-related velopharyngeal dysfunction: a systematic review","author":"Isber","year":"2026","journal-title":"Cleft Palate Craniofac J"},{"key":"B12","doi-asserted-by":"publisher","first-page":"2986","DOI":"10.1109\/TBME.2021.3058424","article-title":"A deep learning algorithm for objective assessment of hypernasality in children with cleft palate","volume":"68","author":"Mathad","year":"2021","journal-title":"IEEE Trans Biomed Eng"},{"key":"B13","author":"Song","year":"2022"},{"key":"B14","doi-asserted-by":"publisher","first-page":"e0315452","DOI":"10.1371\/journal.pone.0315452","article-title":"Computing nasalance with\u00b4 mfccs and convolutional neural networks","volume":"19","author":"Lozano","year":"2024","journal-title":"PLoS One"},{"key":"B15","doi-asserted-by":"publisher","first-page":"2714","DOI":"10.3390\/diagnostics13162714","article-title":"Automatic detection system for velopharyngeal insufficiency based on acoustic signals from nasal and oral channels","volume":"13","author":"Zhang","year":"","journal-title":"Diagnostics"},{"key":"B16","first-page":"113","article-title":"Multi-task learning and gan-based enhancement for automated assessment of cleft palate speech","author":"Rohith","year":"2026"},{"key":"B17","doi-asserted-by":"publisher","first-page":"1800132","DOI":"10.3389\/fdgth.2026.1800132","article-title":"Voice disorders classification using machine learning: a scoping review","volume":"8","author":"Gupta","year":"2026","journal-title":"Front Digit Health"},{"key":"B18","doi-asserted-by":"publisher","first-page":"1752356","DOI":"10.3389\/fdgth.2026.1752356","article-title":"Classifying voice disorders for machine learning: pilot study of the usvac-c2025 framework","volume":"8","author":"Madill","year":"2026","journal-title":"Front Digit Health"},{"key":"B19","doi-asserted-by":"crossref","first-page":"73","DOI":"10.1007\/978-3-032-03134-1_6","article-title":"Ai models for voice disorders: considerations from development to deployment","volume-title":"Voice-related Biomarkers","author":"Das","year":"2026"},{"key":"B20","doi-asserted-by":"publisher","first-page":"1754426","DOI":"10.3389\/fdgth.2026.1754426","article-title":"Translating ai research into reality: summary of the 2025 voice ai symposium and hackathon","volume":"8","author":"Salvi Cruz","year":"2026","journal-title":"Front Digit Health"},{"key":"B21","doi-asserted-by":"publisher","first-page":"00194","DOI":"10.1016\/j.jvoice.2026.04.024","article-title":"Ai-driven detection and classification of voice disorders using acoustic recordings","author":"Berteloot","year":"2026","journal-title":"J Voice"},{"key":"B22","doi-asserted-by":"publisher","first-page":"253","DOI":"10.1007\/s10489-026-07264-4","article-title":"Voicepathnet: classifier fusion with cnn-driven multiclass voice disorder diagnosis with enhanced preprocessing","volume":"56","author":"Roy","year":"2026","journal-title":"Applied Intelligence"},{"key":"B23","doi-asserted-by":"publisher","first-page":"12449","DOI":"10.48550\/arXiv.2006.11477","article-title":"Wav2vec 2.0: a framework for self-supervised learning of speech representations","volume":"33","author":"Baevski","year":"2020","journal-title":"Adv Neural Inf Process Syst"},{"key":"B24","doi-asserted-by":"publisher","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":"Hsu","year":"2021","journal-title":"IEEE\/ACM Transactions on Audio, Speech, and Language Processing"},{"key":"B25","first-page":"1298","article-title":"Data2vec: a general framework for self-supervised learning in speech, vision and language","author":"Baevski","year":"2022"},{"key":"B26","first-page":"28492","article-title":"Robust speech recognition via large-scale weak supervision","author":"Radford","year":"2023"},{"key":"B27","author":"Liu","year":"2026"},{"key":"B28","doi-asserted-by":"publisher","first-page":"18661","DOI":"10.48550\/arXiv.2004.11362","article-title":"Supervised contrastive learning","volume":"33","author":"Khosla","year":"2020","journal-title":"Adv Neural Inf Process Syst"},{"key":"B29","first-page":"5206","article-title":"Librispeech: an asr corpus based on public domain audio books","author":"Panayotov","year":"2015"},{"key":"B30","author":"Lugosch","year":"2019"},{"key":"B31","first-page":"498","article-title":"Montreal Forced aligner: trainable text-speech alignment using kaldi","author":"McAuliffe","year":"2017"},{"key":"B32","author":"Gilkey","year":"2026"},{"key":"B33","author":"Lugosch","year":"2019"},{"key":"B34","first-page":"1597","article-title":"A simple framework for contrastive learning of visual representations","author":"Chen","year":"2020"},{"key":"B35","doi-asserted-by":"crossref","first-page":"785","DOI":"10.1145\/2939672.2939785","article-title":"XGBoost: A scalable tree boosting system","volume-title":"Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining","author":"Chen","year":"2016"},{"key":"B36","doi-asserted-by":"publisher","first-page":"357","DOI":"10.1109\/TASSP.1980.1163420","article-title":"Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences","volume":"28","author":"Davis","year":"1980","journal-title":"IEEE transactions on Acoustics, Speech, and Signal Processing"},{"key":"B37","doi-asserted-by":"publisher","first-page":"1348","DOI":"10.1109\/TASLP.2023.3261753","article-title":"Robust vocal quality feature embeddings for dysphonic voice detection","volume":"31","author":"Zhang","year":"","journal-title":"IEEE\/ACM Transactions on Audio, Speech, and Language Processing"},{"key":"B38","first-page":"4776","article-title":"From voice to diagnosis: a hybrid approach to multi-label disease classification with uncertainty awareness","author":"Liu","year":""}],"container-title":["Frontiers in Digital Health"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdgth.2026.1833087\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,16]],"date-time":"2026-07-16T05:36:36Z","timestamp":1784180196000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdgth.2026.1833087\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,16]]},"references-count":38,"alternative-id":["10.3389\/fdgth.2026.1833087"],"URL":"https:\/\/doi.org\/10.3389\/fdgth.2026.1833087","relation":{},"ISSN":["2673-253X"],"issn-type":[{"value":"2673-253X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,16]]},"article-number":"1833087"}}