Understanding Expiratory Occlusion
The rising prevalence of respiratory diseases is placing a growing strain on healthcare systems, driven by increasing cases of asthma, chronic obstructive pulmonary disease (COPD), and respiratory infections. Environmental factors such as air pollution and occupational hazards, along with lifestyle choices like smoking and vaping, further contribute to this burden. Smoking is a well-known cause of lung damage, leading to chronic inflammation, airway obstruction, and diseases like COPD and lung cancer. Vaping, often marketed as a safer alternative, has also been linked to respiratory complications, including airway irritation, lung injury, and conditions such as EVALI (e-cigarette or vaping-associated lung injury). As these issues continue to rise, there is a pressing need to better understand respiratory mechanics and develop more effective disease management strategies.
Understanding the distinct breathing patterns across different lung conditions is crucial for effective diagnosis and treatment. By visualizing and analyzing these distinct patterns, we can better understand how different conditions affect respiratory function, leading to more accurate assessments and modeling a personalized treatment.
The Data
We used data from the paper Respiratory dataset from PEEP study with expiratory occlusion. The data was collected from 80 participants with various lung conditions. The participants were split evenly into groups based on sex and lung condition as either asthmatic, smoker, vaper, or healthy. The information about the participants was self-reported as well as measured. An example of the setup is shown below.
The respiratory data was collected using a custom-calibrated venturi-based flow and pressure sensor device with one-way valves to separate inspiration and expiration. A CPAP machine provided PEEP, and a filter with a full-face mask was used at the patient interface. Aeration data was simultaneously recorded using an electrical impedance tomography (EIT) device, with an electrode belt placed around the chest. The participant remained seated, and EIT and circumference data were continuously collected after calibration.
You can download the dataset here.
Participant Demographics
Explore the demographics breakdown of the 80 participants.
Respiratory Flow Visualization
This visualization shows breathing patterns across different participant groups. It also provides insights into how lung conditions affect airflow, volume, and breathe rate.
Current Flow
Current Volume
Breathing Rate
Insights
Chest vs. Abdominal Motion Comparison
Interactive Model Simulation
With the growing strain of respiratory illnesses on healthcare, there's a strong need for automated models to personalize patient care. We've identified key features and done feature engineering to predict individuals as asthmatic, smokers, vapers, or healthy. We've developed an XGB Classifier model and our best combination achieved an accuracy of 81.25%, can you beat our accuracy?
Model Results
Model Accuracy: