The Yale Department of Psychiatry (in conjunction with Dr. Benjamin Marlin, an assistant computer science professor at University of Massachusetts Amherst) is currently researching methods of detecting cocaine use using wireless on-body sensors. There are five parts to the heartbeat, P, Q, R, S, and T with each letter representing a peak or valley in the ECG (1).
Large parts of the medical community believe that discreet cocaine use causes the prolongation of the QRS peak and QT interval.
The heartbeat of a subject is measured using an electrocardiogram, ECG, to determine discreet cocaine use. Cocaine causes the body’s heartbeat to become irregular, among other symptoms. The characteristic shape of the heartbeat on heart rate monitors becomes distorted and elongated at certain, regular points (1).
Cocaine also causes increased breathing, heart rate, blood pressure, and the constriction of blood vessels. However, a variety of other activities such as exercise and smoking can also cause these symptoms. The arrhythmia that cocaine causes is more unique to cocaine (1).
Working with Dr. Marlin, the Yale team collected heartbeat data from 10 subjects during cocaine use in fixed and self-administrated sessions. Data was also collected during a baseline session, as well as sessions involving physical exercise, smoking, and methylphenidate use (Ritalin). Further data collection is planned.
The ECG chest band used to collect data from the subjects is low power and the ECG waveform that it creates is not very precise. Therefore it is difficult to identify accurately each part of the heartbeat and to determine irregularities.
To solve this problem, Dr. Marlin identifies potential individual parts of the heartbeat and uses a classifier program to compare the potential individual parts of the heartbeat against a known part of the subject’s heartbeat. Using this system, Dr. Marlin was able to achieve above 95% accuracy (1).
Dr. Marlin hopes to refine his process further as new data from the Yale team becomes available to him. He plans on expanding his model to not only incorporate other sensors such as accelerometers in mobile phones, but also to determine other behaviors such as smoking and eating detection.
Dr. Marlin wishes to have this research support further study into cocaine addiction behavior.
Source:
1. Marlin, B. (22 January 2015). Detecting Cocaine Use with Wireless On-Body Sensors. Computer Science Colloqium. Lecture conducted at
Dartmouth College, Hanover, NH.
Leave a Reply