journal article·2022-11-02Open access

The Role of Data Analytics in the Assessment of Pathological Speech—A Critical Appraisal

Pedro Gómez-Vilda, Andrés Gómez-Rodellar, Daniel Palacios-Alonso, Victoria Rodellar-Biarge, Agustín Álvarez-Marquina

Applied Sciences· Volume 12· Issue 21

Abstract

Pathological voice characterization has received increasing attention over the last 20 years. Hundreds of studies have been published showing inventive approaches with very promising findings. Nevertheless, methodological issues might hamper performance assessment trustworthiness. This study reviews some critical aspects regarding data collection and processing, machine learning-oriented methods, and grounding analytical approaches, with a view to embedding developed clinical decision support tools into the diagnosis decision-making process. A set of 26 relevant studies published since 2010 was selected through critical selection criteria and evaluated. The model-driven (MD) or data-driven (DD) character of the selected approaches is deeply examined considering novelty, originality, statistical robustness, trustworthiness, and clinical relevance. It has been found that before 2020 most of the works examined were more aligned with MD approaches, whereas over the last two years a balanced proportion of DD and MD-based studies was found. A total of 15 studies presented MD characters, whereas seven were mainly DD-oriented, and four shared both profiles. Fifteen studies showed exploratory or prospective advanced statistical analysis. Eighteen included some statistical validation to avail claims. Twenty-two reported original work, whereas the remaining four were systematic reviews of others’ work. Clinical relevance and acceptability by voice specialists were found in 14 out of the 26 works commented on. Methodological issues such as detection and classification performance, training and generalization capability, explainability, preservation of semantic load, clinical acceptance, robustness, and development expenses have been identified as major issues in applying machine learning to clinical support systems. Other important aspects to be taken into consideration are trustworthiness, gender-balance issues, and statistical relevance.
Source: crossref · Quality: baseline

Research context

OpenAlex topics: Voice and Speech Disorders · Speech Recognition and Synthesis · Reliability and Agreement in Measurement

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

NeuSpeLab, CTB, Universidad Politécnica de Madrid, Pozuelo de Alarcón, 28220 Madrid, Spain

Usher Institute, Faculty of Medicine, University of Edinburgh, Edinburgh EH8 9YL, UK

Escuela Técnica Superior de Ingeniería Informática, Universidad Rey Juan Carlos, Móstoles, 28933 Madrid, Spain

Funding

King Juan Carlos University

References (94)

Show source-provided references
1. Hybrid expert systems: A survey of current approaches and applications10.1016/j.eswa.2011.08.130 ↗
2. Keener, J., and Sneyd, J. (2009). Mathematical Physiology: II: Systems Physiology, Springer.10.1007/978-0-387-79388-7 ↗
3. Titze, I. (1994). Principles of Voice Production, Prentice-Hall.
4. Hastie, T., Tibshirani, R., Friedman, J.H., and Friedman, J.H. (2009). The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Springer.10.1007/978-0-387-84858-7 ↗
5. Chaos: Useful at last?10.1126/science.1087822 ↗
6. Glottal source information for pathological voice detection10.1109/ACCESS.2020.2986171 ↗
7. The source filter concept in voice production
8. A Survey of Methods for Explaining Black Box Models10.1145/3236009 ↗
9. Automated analysis of connected speech reveals early biomarkers of Parkinson’s disease in patients with rapid eye movement sleep behaviour disorder10.1038/s41598-017-00047-5 ↗
10. ASHA (2022, July 13). Voice Disorders. Available online: https://www.asha.org/Practice-Portal/Clinical-Topics/Voice-Disorders/.
11. Schmid, L., Gerharz, A., Groll, A., and Pauly, M. (2022). Machine Learning for Multi-Output Regression: When should a holistic multivariate approach be preferred over separate univariate ones?. arXiv.
12. (2022, October 21). Saarbrücken Voice Database. Available online: http://www.stimmdatenbank.coli.uni-saarland.de/help_en.php4.
13. An algorithm for Parkinson’s disease speech classification based on isolated words analysis10.1007/s13755-021-00162-8 ↗
14. Pathological Likelihood Index as a Measurement of the Degree of Voice Normality and Perceived Hoarseness10.1016/j.jvoice.2009.04.003 ↗
15. MEEI Database, Massachusetts Eye and Ear Infirmary Voice and Speech Lab, Boston, MA. &. KayPENTAX, Kay Elemetrics Disordered Voice Database, Model 4337. Kay Elemetrics, Lincoln Park, NJ, USA. 1996–2005.
16. Causability and explainability of artificial intelligence in medicine10.1002/widm.1312 ↗
17. Robust and complex approach of pathological speech signal analysis10.1016/j.neucom.2015.02.085 ↗
18. Detection of different voice diseases based on the nonlinear characterization of speech signals10.1016/j.eswa.2017.04.012 ↗
19. A basic protocol for functional assessment of voice pathology, especially for investigating the efficacy of (phonosurgical) treatments and evaluating new assessment techniques10.1007/s004050000299 ↗
20. Voice Pathology Detection and Classification Using Auto-Correlation and Entropy Features in Different Frequency Regions
21. Magner, L.N., and Kim, O.J. (2017). A History of Medicine, CRC Press.
22. A survey on explainable artificial intelligence (XAI): Toward medical XAI10.1109/TNNLS.2020.3027314 ↗
23. Madiega, T.A. (2022, July 13). EU Guidelines on Ethics in Artificial Intelligence: Context and Implementation. EPRS: European Parliamentary Research Service. Available online: https://policycommons.net/artifacts/1337743/eu-guidelines-on-ethics-in-artificial-intelligence/1945725/.
24. Lightweight Attention Convolutional Neural Network for Retinal Vessel Image Segmentation10.1109/TII.2020.2993842 ↗
25. Quo vadis artificial intelligence?10.1007/s44163-022-00022-8 ↗
26. Steps to avoid overuse and misuse of machine learning in clinical research10.1038/s41591-022-01961-6 ↗
27. Categorizing Normal and Pathological Voices: Automated and Perceptual Categorization10.1016/j.jvoice.2010.04.009 ↗
28. Data dependent random forest applied to screening for laryngeal disorders through analysis of sustained phonation: Acoustic versus contact microphone10.1016/j.medengphy.2014.12.005 ↗
29. Voice Disorders: Etiology and Diagnosis10.1016/j.jvoice.2015.09.017 ↗
30. High-performance medicine: The convergence of human and artificial intelligence10.1038/s41591-018-0300-7 ↗
31. Towards robust voice pathology detection10.1007/s00521-018-3464-7 ↗
32. Expert systems: Definitions, advantages and issues in medical field applications10.1016/j.eswa.2021.114900 ↗
33. From certainty factors to belief networks10.1016/0933-3657(92)90036-O ↗
34. The perceptron: A probabilistic model for information storage and organization in the brain10.1037/h0042519 ↗
35. Logical versus analogical or symbolic versus connectionist or neat versus scruffy
36. Current topics in voice production mechanisms10.3109/00016489309135838 ↗
37. Generalized mel frequency cepstral coefficients for large-vocabulary speaker-independent continuous-speech recognition10.1109/89.784104 ↗
38. Relevance, redundancy and complementarity trade-off (RRCT): A generic, efficient, robust feature selection tool10.1016/j.patter.2022.100471 ↗
39. Statistical Modeling: The two cultures10.1214/ss/1009213726 ↗
40. Forsyth, D. (2019). Applied Machine Learning, Springer.10.1007/978-3-030-18114-7 ↗
41. Kelly, C.J., Karthikesalingam, A., Suleyman, M., Corrado, G., and King, D. (2019). Key challenges for delivering clinical impact with artificial intelligence. BMC Med., 17.10.1186/s12916-019-1426-2 ↗
42. Ashri, R. (2022, July 14). Building AI Software: Data-Driven vs. Model-Driven AI and Why We Need an AI-Specific Software (Issues Brief). Available online: https://hackernoon.com/building-ai-softwaredata-driven-vs-model-driven-ai-and-why-we-need-an-specific-software-640f74aaf78f.
43. Maruyama, Y. (2020). Symbolic and statistical theories of cognition: Towards integrated artificial intelligence. International Conference on Software Engineering and Formal Methods, Springer.10.1007/978-3-030-67220-1_11 ↗
44. Taroni, F., Bozza, S., Biedermann, A., Garbolino, P., and Aitken, C. (2010). Data Analysis in Forensic Science: A Bayesian Decision Perspective, John Wiley & Sons.10.1002/9780470665084 ↗
45. Sex hormones and the female voice10.1016/S0892-1997(99)80048-4 ↗
46. Anatomy of the larynx and pharynx: Effects of age, gender and height revealed by multidetector computed tomography10.1111/joor.12298 ↗
47. Davatz, G.C., Yamasaki, R., Hachiya, A., Tsuji, D.H., and Montagnoli, A.N. (2021). Source and Filter Acoustic Measures of Young, Middle-Aged and Elderly Adults for Application in Vowel Synthesis. J. Voice, in press.
48. Some acoustic characteristics in the voices of 6- to 10-year-old children and adults: A comparative sex and developmental perspective10.1080/14015430050175851 ↗
49. Glottal Source biometrical signature for voice pathology detection10.1016/j.specom.2008.09.005 ↗
50. Sex and gender differences and biases in artificial intelligence for biomedicine and healthcare10.1038/s41746-020-0288-5 ↗
51. Mellino, S., Morey, C., and Rohner, C. (2022). Biases in digital health measures. Sex and Gender Bias in Technology and Artificial Intelligence, Elsevier.10.1016/B978-0-12-821392-6.00001-7 ↗
52. Bouckaert, R.R., and Frank, E. (2004, January 26–28). Evaluating the replicability of significance tests for comparing learning algorithms. Proceedings of the Pacific-Asia Conference on Knowledge Discovery and Data Mining, Sydney, NSW, Australia.10.1007/978-3-540-24775-3_3 ↗
53. Statistical comparisons of classifiers over multiple data sets
54. A Simple Generalisation of the Area Under the ROC Curve for Multiple Class Classification Problems10.1023/A:1010920819831 ↗
55. Phoneme recognition using time-delay neural networks10.1109/29.21701 ↗
56. Voice Pathology Detection and Classification by Adopting Online Sequential Extreme Learning Machine10.1109/ACCESS.2021.3082565 ↗
57. Quick extreme learning machine for large-scale classification10.1007/s00521-021-06727-8 ↗
58. Voice Pathologies Classification and Detection Using EMD-DWT Analysis Based on Higher Order Statistic Features10.1016/j.irbm.2019.11.004 ↗
59. Pathological assessment of patients’ speech signals using nonlinear dynamical analysis10.1016/j.compbiomed.2009.10.011 ↗
60. Tennenholtz, G., Zahavy, T., and Mannor, S. (2018). Train on validation: Squeezing the data lemon. arXiv.
61. Automatic Detection of Pathological Voices Using Complexity Measures, Noise Parameters, and Mel-Cepstral Coefficients
62. Predicting electrical power output of combined cycle power plants using a novel artificial neural network optimized by electrostatic discharge algorithm10.1016/j.measurement.2022.111405 ↗
63. Subset simulation with adaptable intermediate failure probability for rogust reliability analysis: And unsupervised learning-based approach10.1007/s00158-022-03260-7 ↗
64. The impact of artificial intelligence in medicine on the future role of the physician10.7717/peerj.7702 ↗
65. Lee, J.-Y. (2021). Experimental evaluation of deep learning methods for an intelligent pathological voice detection system using the Saarbruecken voice database. Appl. Sci., 11.10.3390/app11157149 ↗
66. On the design of automatic voice condition analysis systems. Part I: Review of concepts and an insight to the state of the art10.1016/j.bspc.2018.12.024 ↗
67. A survey on signal processing based pathological voice detection techniques10.1109/ACCESS.2020.2985280 ↗
68. A Survey on Machine Learning Approaches for Automatic Detection of Voice Disorders10.1016/j.jvoice.2018.07.014 ↗
69. Characterization Methods for the Detection of Multiple Voice Disorders: Neurological, Functional, and Laryngeal Diseases10.1109/JBHI.2015.2467375 ↗
70. An efficient voice pathology classification scheme based on applying multi-layer linear discriminant analysis to wavelet packet-based features10.1016/j.bspc.2013.11.002 ↗
71. Voice Pathology Detection Using Modulation Spectrum-Optimized Metrics
72. Voice disorder identification by using machine learning techniques10.1109/ACCESS.2018.2816338 ↗
73. Pützer, M., and Wokurek, W. (2021). Electroglottographic and Acoustic Parametrization of Phonatory Quality Provide Voice Profiles of Pathological Speakers. J. Voice, in press.
74. Voice data mining for laryngeal pathology assessment10.1016/j.compbiomed.2015.07.026 ↗
75. Kullback–Leibler divergence and sample skewness for pathological voice quality assessment10.1016/j.bspc.2019.101697 ↗
76. Multi-modal voice pathology detection architecture based on deep and handcrafted feature fusion
77. Deep connected attention (DCA) ResNet for robust voice pathology detection and classification10.1016/j.bspc.2021.102973 ↗
78. Analysis and detection of pathological voice using glottal source features10.1109/JSTSP.2019.2957988 ↗
79. Deep neural network for automatic classification of pathological voice signals10.1016/j.jvoice.2020.05.029 ↗
80. Mohammed, M.A., Abdulkareem, K.H., Mostafa, S.A., Ghani, M.K.A., Maashi, M.S., Garcia-Zapirain, B., Oleagordia, I., AlHakami, H., and Al-Dhief, F.T. (2020). Voice Pathology Detection and Classification Using Convolutional Neural Network Model. Appl. Sci., 10.10.3390/app10113723 ↗
81. Investigation and Evaluation of Glottal Flow Waveform for Voice Pathology Detection10.1109/ACCESS.2020.3046767 ↗
82. Gammatone spectral latitude features extraction for pathological voice detection and classification10.1016/j.apacoust.2021.108417 ↗
83. Olson, R.S., and Moore, J.H. (2016, January 24). TPOT: A Tree-Based Pipeline Optimization Tool for Automating Machine Learning. Proceedings of the Workshop on Automatic Machine Learning, New York, NY, USA. Available online: http://proceedings.mlr.press/v64/olson_tpot_2016.pdf.
84. LeDell, E., and Poirier, S. (2020, January 17–18). H2O automl: Scalable automatic machine learning. Proceedings of the AutoML Workshop at ICML, online.
85. Generative Adversarial Networks: An Overview10.1109/MSP.2017.2765202 ↗
86. Open set task augmentation facilitates generalization of deep neural networks trained on small data sets10.1007/s00521-021-06753-6 ↗
87. Recommended Protocols for Instrumental Assessment of Voice: American Speech-Language-Hearing Association Expert Panel to Develop a Protocol for Instrumental Assessment of Vocal Function10.1044/2018_AJSLP-17-0009 ↗
88. James, G., Witten, D., Hastie, T., and Tibshirani, R. (2013). An Introduction to Statistical Learning, Springer.10.1007/978-1-4614-7138-7 ↗
89. Feature Selection10.1145/3136625 ↗
90. Relief-based feature selection: Introduction and review10.1016/j.jbi.2018.07.014 ↗
91. Bernard, M., Poli, M., Karadayi, J., and Dupoux, E. (2021). Shennong: A Python toolbox for audio speech features extraction. arXiv.
92. Generative adversarial networks: Introduction and outlook10.1109/JAS.2017.7510583 ↗
93. A generative adversarial network for synthetization of regions of interest based on digital mammograms10.1038/s41598-022-09929-9 ↗
94. Artificial intelligence within the interplay between natural and artificial computation: Advances in data science, trends and applications10.1016/j.neucom.2020.05.078 ↗