Machine Learning Smart System for Parkinson Disease Classification Using the Voice as a Biomarker
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Objectives: This study presents PD Predict, a machine learning system for Parkinson disease classification using voice as abiomarker. Methods: We first created an original set of recordings from the mPower study, and then extracted several audiofeatures, such as mel-frequency cepstral coefficient (MFCC) components and other classical speech features, using a windowingprocedure. The generated dataset was then divided into training and holdout sets. The training set was used to traintwo machine learning pipelines, and their performance was estimated using a nested subject-wise cross-validation approach. The holdout set was used to assess the generalizability of the pipelines for unseen data. The final pipelines were implementedin PD Predict and accessed through a prediction endpoint developed using the Django REST Framework. PD Predict is atwo-component system: a desktop application that records audio recordings, extracts audio features, and makes predictions;and a server-side web application that implements the machine learning pipelines and processes incoming requests with theextracted audio features to make predictions. Our system is deployed and accessible via the following link: https://pdpredict. herokuapp.com/. Results: Both machine learning pipelines showed moderate performance, between 65% and 75% using thenested subject-wise cross-validation approach. Furthermore, they generalized well to unseen data and they did not overfit thetraining set. Conclusions: The architecture of PD Predict is clear, and the performance of the implemented machine learningpipelines is promising and confirms the usability of smartphone microphones for capturing digital biomarkers of disease.
Abstract
Objectives: This study presents PD Predict, a machine learning system for Parkinson disease classification using voice as abiomarker. Methods: We first created an original set of recordings from the mPower study, and then extracted several audiofeatures, such as mel-frequency cepstral coefficient (MFCC) components and other classical speech features, using a windowingprocedure. The generated dataset was then divided into training and holdout sets. The training set was used to traintwo machine learning pipelines, and their performance was estimated using a nested subject-wise cross-validation approach. The holdout set was used to assess the generalizability of the pipelines for unseen data. The final pipelines were implementedin PD Predict and accessed through a prediction endpoint developed using the Django REST Framework. PD Predict is atwo-component system: a desktop application that records audio recordings, extracts audio features, and makes predictions;and a server-side web application that implements the machine learning pipelines and processes incoming requests with theextracted audio features to make predictions. Our system is deployed and accessible via the following link: https://pdpredict. herokuapp.com/. Results: Both machine learning pipelines showed moderate performance, between 65% and 75% using thenested subject-wise cross-validation approach. Furthermore, they generalized well to unseen data and they did not overfit thetraining set. Conclusions: The architecture of PD Predict is clear, and the performance of the implemented machine learningpipelines is promising and confirms the usability of smartphone microphones for capturing digital biomarkers of disease.
참고문헌 (29)
- Weintraub D, "Parkinson’s disease : Part 1 : pathophysiology, symptoms, burden, diagnosis, and assessment", Am J Manag Care, 14(2 Suppl), pp.S40~S48 (2008)[DOI]
- Goetz CG, "Evidencebased medical review update : pharmacological and surgical treatments of Parkinson’s disease : 2001 to 2004", Mov Disord, 20(5), pp.523~539 (2005)[DOI]
- Ho AK, "Speech impairment in a large sample of patients with Parkinson’s disease", Behav Neurol, 11(3), pp.131~137 (1998)[DOI]
- Perry B, "Use of mobile devices to measure outcomes in clinical research, 2010-2016 : a systematic literature review", Digit Biomark, 2(1), pp.11~30 (2018)[DOI]
- Sakar CO, "A comparative analysis of speech signal processing algorithms for Parkinson’s disease classification and the use of the tunable Q-factor wavelet transform", Appl Soft Comput, 74, pp.255~263 (2019)[DOI]
- Gunduz H, "Deep learning-based Parkinson’s disease classification using vocal feature sets", IEEE Access, 7, pp.115540~115551 (2019)[DOI]
- Haq AU, "Feature selection based on L1-norm support vector machine and effective recognition system for Parkinson’s disease using voice recordings", IEEE Access, 7, pp.37718~37734 (2019)[DOI]
- Almeida JS, "Detecting Parkinson’s disease with sustained phonation and speech signals using machine learning techniques", Pattern Recognit Lett, 125, pp.55~62 (2019)[DOI]
- Berus L, "Classifying Parkinson’s disease based on acoustic measures using artificial neural networks", Sensors(Basel), 19(1), pp.16~ (2018)[DOI]
- Benba A, "Analysis of multiple types of voice recordings in cepstral domain using MFCC for discriminating between patients with Parkinson’s disease and healthy people", Int J Speech Technol, 19(3), pp.449~456 (2016)[DOI]
- Bot BM, "The mPower study, Parkinson disease mobile data collected using ResearchKit", Sci Data, 3, pp.160011~ (2016)[DOI]
- "mPower: mobile Parkinson disease study" (2019)[URL]
- "Synapse REST API: basics table query" (2020)[URL]
- Wong SL, "Parkinson's disease : prevalence, diagnosis and impact", Health Rep, 25(11), pp.10~14 (2014)
- Bonet-Sola D, "A comparative survey of feature extraction and machine learning methods in diverse acoustic environments", Sensors(Basel), 21(4), pp.1274~ (2021)[DOI]
- Giannakopoulos T, "Introduction to audio analysis: a MATLAB approach" (2014), Academic Press
- "Surfboard:audio feature extraction for modern machine learning" (2020)[URL]
- Ilias Tougui, "Impact of the Choice of Cross-Validation Techniques on the Results of Machine Learning-Based Diagnostic Applications", Healthcare Informatics Research, 27(3), pp.189~199 (2021), 대한의료정보학회
- Kotsiantis SB, "Data preprocessing for supervised leaning", Int J Comput Sci, 1(2), pp.111~117 (2006)
- Chandrashekar G, "A survey on feature selection methods", Comput Electr Eng, 40(1), pp.16~28 (2014)[DOI]
- Zou H, "Regularization and variable selection via the elastic net", J R Stat Soc Series B Stat Methodol, 67(2), pp.301~320 (2005)[DOI]
- Bergstra J, "Random search for hyper-parameter optimization", J Mach Learn Res, 13(2), pp.281~305 (2012)
- Vabalas A, "Machine learning algorithm validation with a limited sample size", PLoS One, 14(11), pp.e0224365~ (2019)[DOI]
- Virbel M, "Kivy : a framework for rapid creation of innovative user interfaces", Workshop-Proceedings der Tagung Mensch & Computer 2011, pp.69~73 (2011)
- "PyAudio" (2006)[URL]
- "PyInstaller Manual" (2020)[URL]
- "Inno Setup" (2022)[URL]
- "Django: the web framework for perfectionists with deadlines" (2022)[URL]
- Pedregosa F, "Scikit-learn : machine learning in Python", J Mach Learn Res, 12, pp.2825~2830 (2011)