/TrackPain

Hack4rare Track 3 Rasopathy solution. The last part of a solution that dedicates to help predict occurrences of pain in non-communicative Rasopathy patients.

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

TrackPain

installation dependencies

Get pomegranate installed in Python3:

pip3 install pomegranate

Or get the latest version of pomegranate

git clone https://github.com/jmschrei/pomegranate cd pomegranate python setup.py install

Please refer to https://pgmpy.org/models/bayesiannetwork.html for more tutorials on creating a Bayesian network

Introduction

This is the third part of the Hack$rare solution Trackpain, an inference model that can help predict pain levels for non-communicative children. Please refer to this page for the complete solution.

To show a proof of concept, we use an imaginary person whose biological data and pain scales are recorded. We will show how to construct a network for this particular individual patient.

Data preparation

  • The sensor will collect biological data including heart rate variability, respiration rate, actigraphy, and skin conductance
  • The phone app will collect self-reported metrics obtained from visual pain analog, including facial and emotional scores; the third party caregivers can also make observations about eating, sleeping, social, and verbal patterns.
  • Refer to this notebook for data cleaning probabilty calculation.

Modeling and inference

  • Given probabilities value, create a static binary discrete Bayesian network
  • Refer to this notebook for model creation and prediction

Future consideration

  • Make a continuous instead of discrete network
  • following this pipeline, construct a dynamic network that fits all patients data

Debug

  • Note that we are trying to solve the error in network creation. See Issue Section.

Newest version on google colab