Skip to content

About

Machine learning prototype that detects vestibular problems in concussion patients from eye-tracking data.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

3 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 

Repository files navigation

Vestibular Problems Detector

This repository contains a Jupyter notebook for exploring and prototyping a machine-learning approach to detect vestibular problems in concussion patients from eye-tracking data.

The notebook focuses on two related goals:

  • Training a model from historical scans paired with clinically collected VVE labels.
  • Running prediction on a new eye-tracking scan to estimate whether vestibular issues are present.

Overview

The working assumption behind this project is that eye-tracking signals collected during an EyeBOX exam can be used to predict vestibular problems without relying on the full VVE exam at inference time.

The notebook documents:

  • Data visualization and preprocessing.
  • Correlation analysis to identify repetitive or highly related features.
  • Sample extraction from scan files.
  • A prediction flow that uses a saved MiniRocket-based feature transformer and classifier.

Notebook Contents

  • Vestibular Problem Detector.ipynb - main analysis and prototype notebook.
  • Colab launch link - included at the top of the notebook for browser-based execution.

Data Used

The notebook works with scan data that includes these features:

  • LeftX
  • LeftY
  • RightX
  • RightY
  • LeftPupilRadius
  • RightPupilRadius

It also references the VVE abnormal-measure count as the label source for training samples.

Workflow

  1. Load and clean scan data.
  2. Slice scans into fixed-length samples.
  3. Separate positive and negative training samples using VVE thresholds.
  4. Scale and transform the input data.
  5. Apply the saved transformer and classifier to produce a prediction.

Notes

  • The notebook is intended for experimentation and model prototyping.
  • Some file paths in the notebook are hard-coded for the original working environment and may need to be updated before reuse.
  • The notebook has not been executed in this workspace yet.

Getting Started

Open Vestibular Problem Detector.ipynb in Jupyter Notebook, VS Code, or Colab and run the cells from top to bottom after updating any local data paths.

If you are using the Colab link embedded in the notebook, make sure the referenced data and model files are available in your Colab drive or mounted storage.

About

Machine learning prototype that detects vestibular problems in concussion patients from eye-tracking data.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages