Acoustic Biomarker Extraction for Early Detection of Neurological Disorders Using Generative Speech Representations

Authors

  • Rune Mohammed Department of Computer Science, George Mason University, Fairfax, VA, USA. Author
  • Ada Heinz School of Computing, Clemson University, Clemson, SC, USA. Author
  • Jinthe Lagerweij Department of Computer Science, Colorado State University, Fort Collins, CO, USA. Author
  • Eliza Schenkel Department of Computer Science, Binghamton University, Binghamton, NY, USA. Author

Keywords:

acoustic biomarkers; self-supervised speech representation; generative latent modeling; counterfactual normalization; neurological disorder detection; calibration

Abstract

Early neurological disorders can alter speech before routine clinical diagnosis, yet acoustic biomarkers remain difficult to transfer across speakers, languages, recording conditions, and disease stages. Conventional features such as jitter, shimmer, formant trajectories, pause duration, and articulation rate are interpretable but brittle, while self-supervised speech representations are information-rich but not directly equivalent to measurable biomarkers. This manuscript develops a methodological framework for acoustic biomarker extraction that couples self-supervised speech encoders with a generative latent model, counterfactual resynthesis, calibrated risk estimation, and longitudinal scoring. The central contribution is a representation-level construction in which speech is decomposed into content, speaker style, and a latent neuromotor state; biomarkers are defined as segment-specific deviations from a conditional normative manifold rather than as raw classifier scores. Counterfactual normalization preserves linguistic content and speaker identity while replacing the estimated neuromotor state with a reference state, yielding residual acoustic energy and latent displacement measures that can be interpreted and audited. The framework specifies training losses, invariance constraints, uncertainty handling, decision thresholds, and evaluation endpoints suitable for early detection studies. It also identifies failure modes that can invalidate apparent biomarker performance, including disentanglement leakage, microphone shift, label noise, and over-normalization of disease-related variation. The approach provides a technically coherent route from generative speech representations to reproducible, clinically meaningful acoustic biomarkers without treating synthetic plausibility as diagnostic evidence.

References

1. Jankovic, J. (2008). Parkinson's disease: clinical features and diagnosis. Journal of Neurology, Neurosurgery & Psychiatry, 79(4), 368–376. https://doi.org/10.1136/jnnp.2007.131045

2. Sapir, S., Ramig, L. O., & Fox, C. M. (2008). Speech and swallowing disorders in Parkinson disease. Current Opinion in Otolaryngology & Head and Neck Surgery, 16(3), 205–210. https://doi.org/10.1097/MOO.0b013e3282febd3a

3. Fraser, K. C., Meltzer, J. A., & Rudzicz, F. (2016). Linguistic features identify Alzheimer's disease in narrative speech. Journal of Alzheimer's Disease, 49(2), 407–422. https://doi.org/10.3233/JAD-150520

4. Ahmed, S., Haigh, A.-M., de Jager, C. A., & Garrard, P. (2013). Connected speech as a marker of disease progression in autopsy-proven Alzheimer's disease. Brain, 136(12), 3727–3737. https://doi.org/10.1093/brain/awt269

5. Rusz, J., Cmejla, R., Ruzicka, E., & Ruszova, H. (2011). Quantitative acoustic measurements for characterization of speech and voice disorders in early untreated Parkinson's disease. The Journal of the Acoustical Society of America, 129(1), 350–367. https://doi.org/10.1121/1.3514381

6. Tsanas, A., Little, M. A., McSharry, P. E., Spielman, J., & Ramig, L. O. (2012). Novel speech signal processing algorithms for high-accuracy classification of Parkinson's disease. IEEE Transactions on Biomedical Engineering, 59(5), 1264–1271. https://doi.org/10.1109/TBME.2012.2183367

7. Baevski, A., Zhou, Y., Mohamed, A., & Auli, M. (2020). wav2vec 2.0: A framework for self-supervised learning of speech representations. In Advances in Neural Information Processing Systems (Vol. 33, pp. 12449–12460).

8. Hsu, W.-N., Bolte, B., Tsai, Y.-H. H., Lakhotia, K., Salakhutdinov, R., & Mohamed, A. (2021). HuBERT: Self-supervised speech representation learning by masked prediction of hidden units. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 29, 3451–3460. https://doi.org/10.1109/TASLP.2021.3122291

9. Kingma, D. P., & Welling, M. (2014). Auto-encoding variational Bayes. In Proceedings of the 2nd International Conference on Learning Representations.

10. Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., & Bengio, Y. (2014). Generative adversarial nets. In Advances in Neural Information Processing Systems (Vol. 27, pp. 2672–2680).

11. Ho, J., Jain, A., & Abbeel, P. (2020). Denoising diffusion probabilistic models. In Advances in Neural Information Processing Systems (Vol. 33, pp. 6840–6851).

12. Zhang, Y., Huang, R., Li, R., He, J., Xia, Y., Chen, F., ... & Zhao, Z. (2024, March). Stylesinger: Style transfer for out-of-domain singing voice synthesis. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 38, No. 17, pp. 19597-19605).

13. van den Oord, A., Dieleman, S., Zen, H., Simonyan, K., Vinyals, O., Graves, A., Kalchbrenner, N., Senior, A., & Kavukcuoglu, K. (2016). WaveNet: A generative model for raw audio. arXiv preprint arXiv:1609.03499.

14. Jeancolas, L., Petrovska-Delacrétaz, D., Mangone, G., Benkelfat, B.-E., Corvol, J.-C., Vidailhet, M., Roze, E., & Lehéricy, S. (2021). X-vectors: New quantitative biomarkers for early Parkinson's disease detection from speech. Frontiers in Digital Health, 3, 578369. https://doi.org/10.3389/fdgth.2021.578369

15. Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017). On calibration of modern neural networks. In Proceedings of the 34th International Conference on Machine Learning (Vol. 70, pp. 1321–1330).

16. Vickers, A. J., & Elkin, E. B. (2006). Decision curve analysis: a novel method for evaluating prediction models. Medical Decision Making, 26(6), 565–574. https://doi.org/10.1177/0272989X06295361

17. Ben-David, S., Blitzer, J., Crammer, K., Kulesza, A., Pereira, F., & Vaughan, J. W. (2010). A theory of learning from different domains. Machine Learning, 79(1–2), 151–175. https://doi.org/10.1007/s10994-009-5152-4

18. Marquand, A. F., Rezek, I., Buitelaar, J., & Beckmann, C. F. (2016). Understanding heterogeneity in clinical cohorts using normative models: beyond case-control studies. Biological Psychiatry, 80(7), 552–561. https://doi.org/10.1016/j.biopsych.2015.12.023

19. Woo, C.-W., Chang, L. J., Lindquist, M. A., & Wager, T. D. (2017). Building better biomarkers: brain models in translational neuroimaging. Nature Neuroscience, 20(3), 365–377. https://doi.org/10.1038/nn.4478

20. Topol, E. J. (2019). High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine, 25(1), 44–56. https://doi.org/10.1038/s41591-018-0300-7

Downloads

Published

2026-11-11