The aim of this research program is to develop and validate a smartphone app-based digital measurement concept that: * Objectively quantifies the severity of Parkinson's Disease (PD) related vocal and speech symptoms; * Accurately and sensitively identifies vocal and speech abnormalities associated with the prodromal stage of PD.
Although multiple approaches to this problem have been proposed in addition to commercially available speech analytics platforms, there is currently no established measure which incorporates the disparate aspects of affected speech to fully characterize Parkinson's symptom progression, particularly in the prodromal phase. The measurement concept being evaluated in the present study utilizes a custom smartphone-based speech assessment tool to extract multiple hypothesis-driven acoustic features from patient speech in a real-life environment. The resultant features will be used to train a pair of supervised machine learning models to predict clinical PD symptom severity scores, and to distinguish prodromal PD patients from both healthy matched controls and PD patients in more advanced phases of disease progression.
Study Type
OBSERVATIONAL
Enrollment
91
A custom smartphone-based speech assessment tool to extract multiple hypothesis-driven acoustic features from patient speech in a real-life environment.
Northwestern University
Chicago, Illinois, United States
Compliance of digital speech assessment data recorded via smartphone assessments
o % Interpretable minutes of data per patient
Time frame: 8 weeks
Quality of digital speech assessment data recorded via smartphone assessments
o % Interpretable vs. expected number of minutes of data per patient by complete days on study
Time frame: 8 weeks
Usability of digital speech assessments
o SUS Usability scores by score, grade and adjective rating
Time frame: 8 weeks
Content validity of digital speech assessments
o Percent of patients that score Excellent or Good for usability ratings
Time frame: 8 weeks
Characterization and reliability of digital speech assessment features
o Candidate feature characterization: response distributions, and outlier analysis. Stratification of sustained phonation measures by MDS-UPDRS relevant speech items
Time frame: 8 weeks
Reliability of digital speech assessment features
internal consistency and test-retest reliability
Time frame: 8 weeks
Predictive performance of machine learning (ML) regression model
o Construct validity: convergent validity of each model output versus relevant MDS-UPDRS speech items and Parts I-IV total score, respectively; and versus Hoehn \& Yahr Stage
Time frame: 8 weeks
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.
Predictive performance of ML classification model
o Known group validity by cohort (including by H\&Y Stage/ MDS-UPDRS Parts I-IV total score)
Time frame: 8 weeks