Cough analysis

Group members

Pooya Andy Ells Sarah Nix Glen

TODO

Below is a list of things that we will need to do for the project.

• IRB approval or individual consent?
• Website
• Database
• HTML5 sound collection
• Discuss appropriate metadata collection
• Spectrum analysis
• Complex variable analysis?

Project narrative

We have evolved a cough mechanism in order to breakup our mucus to help our mucocilliary clearance remove the mucus for processing. Cystic fibrosis patients have more viscous mucus than healthy individuals and have an impeded mucocilliary clearance mechanism, so it is reasonable to assume that the way they cough is also impeded somehow. This project will investigate, through crowd-sourcing data, if there exists a "normal" cough by analyzing the coughs of both healthy and unhealthy patients. If there exists a normal frequency spectrum of healthy coughs, then the possibility exists that applying the "normal" cough response to cystic fibrosis patients may help them move mucus from their lungs and thus evading infection.

Introduction

Mucus is a visocoelastic material that is a line of defense to the immune system and it is the first line of defense for the lungs. If a foreign object is between 1–5 µm in aerodynamic diameter it has a high probability to reach the deep lungs. Anything smaller is typically breathed out due to its momentum within the airflow of an inhalation, i.e. it stays within a flow stream of the inhalation. If it is larger, then it has enough momentum to impact the esophagus before it reaches the lungs and is swallowed where the stomach then processes it. However, if the foreign object reaches the lungs, then it will land in the mucus that coats the airways. Defense mechanisms will then respond to the foreign object to remove it from the system before it reaches the bloodstream. If the foreign object stays within the mucus, then the lungs will move the mucus through mucocilliary clearance mechanisms. When the process of mucocilliary clearance needs help moving mucus through the lungs for processing, a cough will be issued. Since mucus is visoelastic, a sharp percussive motion can break up the mucus (i.e. a cough), which then allows the mucocilliary clearance to do its job.

Specific aims

Specific aim 1: Data analysis for this project requires a large amount of data. Collecting data from the participants in the CSSS 2014 meeting will allow us to determine the feasibility of our project, however, we will need more data in order to obtain any statistically significant results. Because of this, our first aim will be to create a working website that users can submit their cough and metadata to. The website will collect metadata about an individual as well as take a recording of the individual cough. This data will be stored in a database and any interesting results will be displayed on the site.

Specific aim 2: From the crowd sourced data, we will initially study the spectrum of each cough. While it seems reasonable that there will be differences between healthy and unhealthy lungs, there may be more complex interactions between the metadata collected and the signal analysis we do on the cough. Because of this, we will use principle component analysis and other data analysis techniques in order to determine if a significant difference between healthy and unhealthy lungs can be distinguished. If a statistical difference can be determined, we will use machine learning algorithms to create a classifier that is capable of distinguishing a healthy lung from an unhealthy lung.

Specific aim 3: If we are successful in find differences in lung types, an analysis of creating an "artificial-normal" cough will be conducted. The hope is that it leads to the advancement for treatment of chronic pulmonary diseases.