Investigation of In-Vehicle Cardiac Monitoring and Severe Event Prediction

Determining what safety systems should be implemented to mitigate accidents related to cardiac events, including driver warnings upon possible event detection and then a slowing of the vehicle and eventually full stop if no response is given.

The Problem

Heart disease is the leading cause of death among both men and women in the United States, according to the Center for Disease Control and Prevention. About 630,000 Americans die from heart disease conditions every year -- representing one in four deaths in the country. As of 2010, the senior population in the US (age 65 and older) was projected to double by 2030, at which point this group will represent nearly quarter of the population. Approximately 70 percent of Americans over the age of 65 have cardiovascular diseases and the potential to get behind the wheel. Currently, no safety systems exist to monitor the cardiac activities that may lead to loss of vehicle control and hazardous situations.

The Question 

How can we determine a driver's cardiac condition by examining signals from non-clinical devices -- such as wearables on in-vehicle systems -- using methods that are robust enough to function with in-vehicle noise, vibration and interference. 

What We Did 

We are integrating a number clinical databases and analyzing this data, focusing on cardiac feature extraction. 

1. Atrial Fibrillation Detection (Afib)

  • Afib classification: sensitivity – 84 percent
  • Non-Afib classification: specificity – 93 percent
  • Area Under Curve (AUC) – 0.95
  • F1 Score – 0.74

2. Atrial Fibrillation Prediction (Afib)

  • Afib classification: sensitivity – 86 percent
  • Non-Afib classification: specificity – 80 percent
  • Area Under Curve (AUC) – 0.91
  • F1 Score – 0.75

This is a project in collaboration with University of Michigan Center for Integrative Research in Critical Care