{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,19]],"date-time":"2026-10-19T20:42:19Z","timestamp":1792442539945,"version":"4.5.0"},"reference-count":50,"publisher":"JMIR Publications Inc.","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["J Med Internet Res"],"abstract":"Abstract<\/jats:title>\n \n Background<\/jats:title>\n Voice disorders are common, yet access to timely specialist assessment is often limited. Automated analysis of short voice recordings may support early screening, but achieving robust performance on heterogeneous clinical datasets remains challenging.<\/jats:p>\n <\/jats:sec>\n \n Objective<\/jats:title>\n This study aims to investigate the impact of synthetic speech augmentation on voice disorder recognition and classification using the Saarbr\u00fccken Voice Database (SVD) and to develop an adaptive ensemble with learnable weighting and optional sex embedding.<\/jats:p>\n <\/jats:sec>\n \n Methods<\/jats:title>\n Sentence recordings from the SVD were used for binary voice disorder recognition (VDR, n=1988) and 4-class voice disorder classification (VDC; n=1199). Two text-to-speech (TTS) models, XTTS and Kartoffelbox, generated 1 synthetic sample for each training instance. Synthetic speech quality was assessed in a blinded listening evaluation with 40 samples per TTS model. Baseline models included support vector machine (SVM), ResNet-18, long short-term memory, and transformer-based architectures, including Hidden-Unit BERT (HuBERT), Wav2Vec 2.0, and WavLM. Ensemble learning and learnable fusion (LF), with and without a sex embedding (LFS), were evaluated using bootstrapped performance estimates and standard classification metrics. Area under the receiver operating characteristic curve (AUC) was reported for VDR.<\/jats:p>\n <\/jats:sec>\n \n Results<\/jats:title>\n \n In the blinded evaluation, Kartoffelbox-generated synthetic samples were less often identified as synthetic than XTTS-generated synthetic samples (10\/19, 52.6% vs 17\/19, 89.5%). In pooled original and synthetic ratings, Kartoffelbox showed higher pathological-status identification than XTTS for pathological samples (22\/29, 75.9% vs 19\/29, 65.5%) and nonpathological samples (8\/10, 80.0% vs 6\/10, 60.0%), as well as higher diagnostic-class assignment (20\/39, 51.3% vs 14\/39, 35.9%). Among the baseline models, HuBERT showed the most balanced VDR performance (accuracy=0.85,\n F<\/jats:italic>\n 1<\/jats:sub>\n -score=0.84). Kartoffelbox-based augmentation further improved performance, with HuBERT achieving 0.88 sensitivity, 0.87 accuracy, and 0.85\n F<\/jats:italic>\n 1<\/jats:sub>\n -score. LF showed minor improvements in threshold-dependent metrics (accuracy=0.87,\n F<\/jats:italic>\n 1<\/jats:sub>\n -score=0.86), while LFS further balanced sensitivity and specificity (accuracy=0.88, AUC=0.92). The equal-weight ensemble achieved the highest AUC among ensemble approaches (AUC=0.93). For VDC, HuBERT was the strongest baseline (accuracy=0.72, median\n F<\/jats:italic>\n 1<\/jats:sub>\n -score=0.63), and LF achieved the highest overall fusion performance (accuracy=0.73,\n F<\/jats:italic>\n 1<\/jats:sub>\n -score=0.66), while laryngitis remained one of the more challenging classes.\n <\/jats:p>\n <\/jats:sec>\n \n Conclusions<\/jats:title>\n High-quality synthetic speech augmentation can improve automated voice disorder analysis across heterogeneous clinical data, with effectiveness varying by the TTS model. Multimodel fusion of complementary architectures provided robust performance for screening and multiclass classification, with the strongest approach varying by metric and task. These findings support further evaluation of AI-assisted preassessment in clinical and telemedicine settings.<\/jats:p>\n <\/jats:sec>","DOI":"10.2196\/94285","type":"journal-article","created":{"date-parts":[[2026,9,9]],"date-time":"2026-09-09T19:05:06Z","timestamp":1788980706000},"page":"e94285-e94285","source":"Crossref","is-referenced-by-count":0,"title":["Detection of Pathological Voice Samples Using a Text-to-Speech\u2013Inspired Approach With Multiple Model Architectures: Method Development and Comparative Evaluation Study"],"prefix":"10.2196","volume":"28","author":[{"ORCID":"https:\/\/orcid.org\/0009-0004-6777-565X","authenticated-orcid":false,"given":"Marie Sophie","family":"Walter","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0008-8484-9322","authenticated-orcid":false,"given":"Muhittin","family":"Demir","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7489-1972","authenticated-orcid":false,"given":"Mattias","family":"Heinrich","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5811-7100","authenticated-orcid":false,"given":"Felix","family":"Nensa","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-4420-9664","authenticated-orcid":false,"given":"Stephan","family":"Lang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0207-7621","authenticated-orcid":false,"given":"Nadia","family":"Sadok","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1760-2342","authenticated-orcid":false,"given":"Ren\u00e9","family":"Hosch","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1010","published-online":{"date-parts":[[2026,10,19]]},"reference":[{"key":"R1","unstructured":"Stimmst\u00f6rung (dysphonie) [Article in German]. Deutscher Bundesverband f\u00fcr Logop\u00e4die eV (dbl). URL: https:\/\/www.dbl-ev.de\/fachwissen-logopaedie\/stimmstoerung-dysphonie\/ [Accessed 20-11-2025]"},{"key":"R2","doi-asserted-by":"crossref","unstructured":"Kroegel C Costabel U Bals R Taube C . Referenz Pneumologie [Book in German]. 1st ed. Georg Thieme Verlag KG; 2024. [doi: 10.1055\/b-006-163265]","DOI":"10.1055\/b-006-163265"},{"key":"R3","unstructured":"Leitlinien-details [Article in German]. AWMF-Leitlinienregister. 2025. URL: https:\/\/register.awmf.org\/de\/leitlinien\/detail\/049-008 [Accessed 24-11-2025]"},{"issue":"1","key":"R4","doi-asserted-by":"publisher","first-page":"75","DOI":"10.1016\/S0030-6665(03)00169-5","article-title":"Evidence-based treatment of paralytic dysphonia: making sense of outcomes and efficacy data","volume":"37","author":"Behrman","journal-title":"Otolaryngol Clin North Am"},{"issue":"3","key":"R5","doi-asserted-by":"publisher","first-page":"263","DOI":"10.1016\/j.anl.2025.04.008","article-title":"The efficacy of early voice therapy in patients with vocal fold paralysis after thyroidectomy","volume":"52","author":"Kim","journal-title":"Auris Nasus Larynx"},{"key":"R6","doi-asserted-by":"publisher","DOI":"10.1016\/j.jvoice.2026.02.014","article-title":"Treatment efficacy for functional voice disorders with dysphonia: a network meta-analysis","author":"Barsties V Latoszek","journal-title":"J Voice"},{"key":"R7","doi-asserted-by":"crossref","unstructured":"Koudounas A Ciravegna G Fantini M et al. Voice disorder analysis: a transformer-based approach. 2024:3040-3044. [doi: 10.21437\/Interspeech.2024-1122]","DOI":"10.21437\/Interspeech.2024-1122"},{"issue":"1","key":"R8","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-023-34461-9","article-title":"Voice disorder classification using convolutional neural network based on deep transfer learning","volume":"13","author":"Peng","journal-title":"Sci Rep"},{"key":"R9","doi-asserted-by":"publisher","DOI":"10.1016\/j.jvoice.2024.09.002","article-title":"Voice disorder classification using Wav2vec 2.0 feature extraction","author":"Cai","journal-title":"J Voice"},{"issue":"2","key":"R10","doi-asserted-by":"publisher","first-page":"288","DOI":"10.1016\/j.jvoice.2020.05.029","article-title":"Deep neural network for automatic classification of pathological voice signals","volume":"36","author":"Chen","journal-title":"J Voice"},{"key":"R11","doi-asserted-by":"publisher","DOI":"10.1016\/j.compeleceng.2018.04.008","article-title":"A new database of healthy and pathological voices","volume":"68","author":"Cesari","journal-title":"Comput Electr Eng"},{"key":"R12","unstructured":"P\u00fctzer M Koreman J . A German database of patterns of pathological vocal fold vibration. Phonus Research Report. Zenodo; 1997. URL: https:\/\/zenodo.org\/records\/16258835 [Accessed 30-09-2026]"},{"key":"R13","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2019.2913444","article-title":"Dysphonia detection index (DDI): a new multi-parametric marker to evaluate voice quality","volume":"7","author":"Verde","journal-title":"IEEE Access"},{"key":"R14","doi-asserted-by":"publisher","DOI":"10.1016\/j.cmpbup.2022.100074","article-title":"Voice pathology detection using convolutional neural networks with electroglottographic (EGG) and speech signals","volume":"2","author":"Islam","journal-title":"Comput Methods Programs Biomed Update"},{"key":"R15","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2026.3658198","article-title":"A novel and accurate recurrence plot-based characterization of speech dynamics in hyperfunctional dysphonia, laryngitis, and vocal fold polyp","volume":"14","author":"Krishnan","journal-title":"IEEE Access"},{"key":"R16","doi-asserted-by":"crossref","unstructured":"Casanova E Davis K G\u00f6lge E et al. XTTS: a massively multilingual zero-shot text-to-speech model. Sep 2024:4978-4982. [doi: 10.21437\/Interspeech.2024-2016]","DOI":"10.21437\/Interspeech.2024-2016"},{"issue":"8","key":"R17","doi-asserted-by":"publisher","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","author":"Hochreiter","journal-title":"Neural Comput"},{"key":"R18","unstructured":"The cepstrum, mel-cepstrum and mel-frequency cepstral coefficients (MFCCs). Introduction to Speech Processing. 2025. URL: https:\/\/speechprocessingbook.aalto.fi\/representations\/melcepstrum\/ [Accessed 24-11-2025]"},{"key":"R19","unstructured":"SebastianBodza\/kartoffelbox-v0.1. Hugging Face. 2025. URL: https:\/\/huggingface.co\/SebastianBodza\/Kartoffelbox-v0.1 [Accessed 20-11-2025]"},{"issue":"2","key":"R20","doi-asserted-by":"publisher","first-page":"367","DOI":"10.1109\/JSTSP.2019.2957988","article-title":"Analysis and detection of pathological voice using glottal source features","volume":"14","author":"Kadiri","journal-title":"IEEE J Sel Top Signal Process"},{"key":"R21","unstructured":"Gupta V . Voice disorder detection using long short term memory (LSTM) model. arXiv. Preprint posted online on Dec 4, 2018. [doi: 10.48550\/arXiv.1812.01779]"},{"key":"R22","doi-asserted-by":"publisher","DOI":"10.1109\/ICMERALDA60125.2023.10458192","article-title":"Children with speech disorders voice classification: LSTM and BiLSTM approach based on MFCC features","author":"Riyanto","journal-title":"ICMERALDA 2023"},{"key":"R23","doi-asserted-by":"crossref","unstructured":"Hastie T Tibshirani R Friedman J . Support vector machines and flexible discriminants. In: The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd ed. Springer; 2009:417-458. [doi: 10.1007\/978-0-387-84858-7_12]","DOI":"10.1007\/978-0-387-84858-7_12"},{"key":"R24","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90","article-title":"Deep residual learning for image recognition","author":"He","journal-title":"IEEE Conf Comput Vis Pattern Recognit"},{"key":"R25","unstructured":"Ioffe S Szegedy C . Batch normalization: accelerating deep network training by reducing internal covariate shift. In: Bach F Blei D , editors. ICML\u201915: Proceedings of the 32nd International Conference on Machine Learning - Volume 37. JMLR; 2015:448-456. URL: https:\/\/dl.acm.org\/doi\/10.5555\/3045118.3045167 [Accessed 16-09-2026]"},{"key":"R26","unstructured":"ReLU. PyTorch. 2025. URL: https:\/\/docs.pytorch.org\/docs\/stable\/generated\/torch.nn.ReLU.html [Accessed 24-11-2025]"},{"key":"R27","unstructured":"Otte S . Recurrent neural networks for sequential pattern recognition applications [Dissertation]. Verlag Dr. Hut; 2017. URL: https:\/\/portal.dnb.de\/opac.htm?method=simpleSearch&cqlMode=true&query=idn%3D1149579382 [Accessed 16-09-2026]"},{"key":"R28","doi-asserted-by":"publisher","DOI":"10.1109\/TASLP.2021.3122291","article-title":"HuBERT: self-supervised speech representation learning by masked prediction of hidden units","volume":"29","author":"Hsu","journal-title":"IEEE\/ACM Trans Audio Speech Lang Process"},{"key":"R29","unstructured":"Baevski A Zhou H Mohamed A Auli M . Wav2vec 2.0: a framework for self-supervised learning of speech representations. In: Larochelle H Ranzato M Hadsell R Balcan MF Lin HT , editors. NIPS \u201920: Proceedings of the 34th International Conference on Neural Information Processing Systems. Curran Associates Inc; 2020:12449-12460. URL: https:\/\/dl.acm.org\/doi\/abs\/10.5555\/3495724.3496768 [Accessed 16-09-2026]"},{"issue":"6","key":"R30","doi-asserted-by":"publisher","first-page":"1505","DOI":"10.1109\/JSTSP.2022.3188113","article-title":"WavLM: large-scale self-supervised pre-training for full stack speech processing","volume":"16","author":"Chen","journal-title":"IEEE J Sel Top Signal Process"},{"key":"R31","unstructured":"Kingma DP Ba J . Adam: a method for stochastic optimization. arXiv. Preprint posted online on Dec 22, 2014. [doi: 10.48550\/arXiv.1412.6980]"},{"key":"R32","unstructured":"CrossEntropyLoss. PyTorch. 2025. URL: https:\/\/docs.pytorch.org\/docs\/stable\/generated\/torch.nn.CrossEntropyLoss.html [Accessed 24-11-2025]"},{"key":"R33","unstructured":"Linear. PyTorch. 2025. URL: https:\/\/docs.pytorch.org\/docs\/stable\/generated\/torch.nn.Linear.html [Accessed 24-11-2025]"},{"issue":"3-4","key":"R34","doi-asserted-by":"publisher","first-page":"591","DOI":"10.1093\/biomet\/52.3-4.591","article-title":"An analysis of variance test for normality (complete samples)","volume":"52","author":"Shapiro","journal-title":"Biometrika"},{"key":"R35","doi-asserted-by":"publisher","DOI":"10.1016\/j.procs.2019.12.233","article-title":"Transfer learning with AudioSet to voice pathologies identification in continuous speech","volume":"164","author":"Guedes","journal-title":"Procedia Comput Sci"},{"key":"R36","doi-asserted-by":"crossref","unstructured":"Rahman MU Direkoglu C . Multi-class classification of voice disorders using deep transfer learning. In: M\u00e1rquez FP Jamil A Ramirez IS Eken S Hameed AA , editors. Computing, Internet of Things and Data Analytics Selected papers from the International Conference on Computing, IoT and Data Analytics (ICCIDA). Springer; 2024:262-270. [doi: 10.1007\/978-3-031-53717-2_25]","DOI":"10.1007\/978-3-031-53717-2_25"},{"key":"R37","doi-asserted-by":"publisher","DOI":"10.1109\/OJSP.2023.3242862","article-title":"Hierarchical multi-class classification of voice disorders using self-supervised models and glottal features","volume":"4","author":"Tirronen","journal-title":"IEEE Open J Signal Process"},{"issue":"4","key":"R38","doi-asserted-by":"publisher","first-page":"195","DOI":"10.4103\/tjem.tjem_182_23","article-title":"Receiver operating characteristic curve analysis in diagnostic accuracy studies: a guide to interpreting the area under the curve value","volume":"23","author":"\u00c7orbac\u0131o\u011flu","journal-title":"Turk J Emerg Med"},{"key":"R39","doi-asserted-by":"crossref","unstructured":"Efron B Tibshirani RJ . An Introduction to the Bootstrap. 1st ed. Chapman and Hall\/CRC; 1993. [doi: 10.1007\/978-1-4899-4541-9]","DOI":"10.1007\/978-1-4899-4541-9_1"},{"key":"R40","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2021-698","article-title":"AST: audio spectrogram transformer","author":"Gong","journal-title":"Interspeech 2021"},{"key":"R41","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2021-1775","article-title":"SUPERB: speech processing universal performance benchmark","author":"Yang","journal-title":"Interspeech 2021"},{"key":"R42","unstructured":"Titze IR . Principles of Voice Production. 2nd ed. National Center for Voice and Speech; 2000. ISBN: 9780874141221"},{"key":"R43","doi-asserted-by":"crossref","unstructured":"Kreiman J Sidtis D . Foundations of Voice Studies: An Interdisciplinary Approach to Voice Production and Perception. 1st ed. Wiley-Blackwell; 2011. ISBN: 9780631222972","DOI":"10.1002\/9781444395068"},{"issue":"1","key":"R44","doi-asserted-by":"publisher","first-page":"24","DOI":"10.1038\/s41591-018-0316-z","article-title":"A guide to deep learning in healthcare","volume":"25","author":"Esteva","journal-title":"Nat Med"},{"issue":"1","key":"R45","doi-asserted-by":"publisher","first-page":"307","DOI":"10.3389\/fpubh.2017.00307","article-title":"Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice","volume":"5","author":"Trevethan","journal-title":"Front Public Health"},{"issue":"5","key":"R46","doi-asserted-by":"publisher","first-page":"804","DOI":"10.1016\/j.jvoice.2018.04.016","article-title":"A comparison of voice activity and participation profiles among etiological groups","volume":"33","author":"Lee","journal-title":"J Voice"},{"issue":"4","key":"R47","doi-asserted-by":"publisher","first-page":"905","DOI":"10.1016\/s0030-6665(05)70251-6","article-title":"Outcome measurements and quality of life in voice disorders","volume":"33","author":"Murry","journal-title":"Otolaryngol Clin North Am"},{"issue":"2","key":"R48","doi-asserted-by":"publisher","first-page":"241","DOI":"10.1016\/S0893-6080(05)80023-1","article-title":"Stacked generalization","volume":"5","author":"Wolpert","journal-title":"Neural Netw"},{"issue":"9","key":"R49","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3649447","article-title":"Deep multimodal data fusion","volume":"56","author":"Zhao","journal-title":"ACM Comput Surv"},{"key":"R50","unstructured":"Saarbruecken voice database. Zenodo. URL: https:\/\/zenodo.org\/records\/16874898 [Accessed 16-09-2026]"}],"container-title":["Journal of Medical Internet Research"],"original-title":[],"language":"en","deposited":{"date-parts":[[2026,10,19]],"date-time":"2026-10-19T20:22:47Z","timestamp":1792441367000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.jmir.org\/2026\/1\/e94285"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,19]]},"references-count":50,"URL":"https:\/\/doi.org\/10.2196\/94285","relation":{},"ISSN":["1438-8871"],"issn-type":[{"value":"1438-8871","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,19]]},"article-number":"v28i1e94285"}}