Skip to content

bhanmrinal/fPI-Parkison-Analyser-using-Acoustic-Sound-Features

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

fPI: Predicting Parkinson's with Acoustic Features

What is Parkinson's Disease?

Parkinson's disease is a neurodegenerative disorder that affects millions of people worldwide. It's characterized by the gradual loss of dopaminergic neurons in the brain, leading to tremors, rigidity, and difficulty with movement.

Dataset

We've utilized a dataset containing biomedical voice measurements from 198 individuals. Here's a breakdown:

  • Participants: 198 individuals (140 with Parkinson's, 58 healthy)\
  • Format: Initially ASCII CSV, converted to .xlsx for easier analysis

Attributes:

  • Name: Subject name and recording number

  • Vocal Fundamental Frequency (MDVP:Fo, Fhi, Flo) ️

  • Jitter and Shimmer Measures (variation in frequency and amplitude)

  • Noise-to-Honer Ratio (NHR, HNR) ⚖️

  • Nonlinear Dynamical Complexity (RPDE, D2)

  • Signal Fractal Scaling Exponent (DFA)

  • Nonlinear Measures of Frequency Variation (spread1, spread2, PPE) 〰️

  • status: Health status (1 = Parkinson's, 0 = Healthy)

Novelty

Based on our correlation map analysis, we identified a potentially valuable feature combination: Denoted as the fPI Analyser

log10(DFA * D2 * spread2)

  • D2: Nonlinear dynamical complexity measure
  • DFA: Signal fractal scaling exponent
  • spread2: Nonlinear measure of fundamental frequency variation This combined feature may offer stronger predictive power for Parkinson's detection using machine learning models.

Model Development and Evaluation

We investigated the effectiveness of fPI by training and evaluating four machine learning models: Decision Trees, Support Vector Machines, Random Forest, and XGBoost. We used four performance metrics to assess their effectiveness: Precision, Accuracy, F1-score, and Recall. This comprehensive evaluation helped us identify the best model for Parkinson's detection using fPI.

Deployment

This project is deployed as a full-stack web application, enabling global access to the Parkinson's disease detection system. The application leverages the following technologies:

Frontend: Flask for building dynamic and interactive user interfaces.

Backend: Django with Django REST framework (DRF) for rapid development of APIs and web applications. DRF provides a robust foundation for building RESTful APIs.

Results

image

Metric scores for different ML Models

image

Metric plots for True Positive and False Positive Estimation

About

Detecting Parkinson's Disease using Acoustic Sound Features

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors