← Back to Explore

Systematic Parkinson Audio Recognition Construct (SPARC): A Novel Approach Implementing A Machine Learning Method to Diagnose Parkinson’s Disease Using Voice Features

JSHS · 2022

Overview

(UAMS); Dr. Yasir Rahmatallah, Professor, Department of Biomedical Informatics, University of Arkansas Medical Sciences (UAMS); Mr. Aaron Kemp, Lead Research Manager, Neurocognitive Dynamics Lab, Psychiatric Research Institute, University of Arkansas for Medical Sciences (UAMS) Parkinson’s disease (PD) is a neurodegenerative disorder primarily prominent in individuals 65 years and older (the elderly population). Despite advances in the medical field, the diagnosis of PD requires examination in a clinical setting. However, due to the ongoing coronavirus pandemic in the United States (January 2020-present), requesting individuals to visit their local clinic can place them at potential risk for coronavirus. A literature search with Google Scholar and PubMed databases from January 2020 to January 2022 determined that currently, no machine learning model (n=0/188) has an accuracy of 90% or higher in predicting PD from vocal features. We propose our model SPARC, the Systematic Parkinson Audio Recognition Construct, a virtual diagnostic tool for the screening of patients with Parkinson’s disease. Project SPARC consisted of the following steps: a) data collection, b) filter audio files, c) feature analysis, d) audio files to images, e) train a convolutional neural network (CNN), and f) determine which test is more accurate (Ah or Rainbow passage) to assist with PD diagnosis. By training a random forest algorithm to extract vocal biometric data for feature selection and a transfer learning CNN on waveform and mel spectrogram images, SPARC is a state-of-the-art model that identified vital vocal features in PD and determined that mel spectrograms and the Ah test are accurate identifiers for PD. Project SPARC is successful in providing an accurate and effective method for PD diagnosis (94% accurate) in a clinical or virtual setting through a vocal feature-based machine learning model.

Competition history

  • JSHS 2022 Category not listed

Resources

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Source: Junior Science and Humanities Symposium

Save projects to your library

Sign in with Google to keep track of projects you find interesting, organized into folders. An account also raises your daily allowance for “Has this been done?”, and lets you create a key for the MCP server with a much higher limit than anonymous use. Browsing stays public.

Continue with Google