Chest compression frequency measurement from smartphone video

Abstract

Dispatcher assisted telephone CPR plays an important role in increasing sur- vival rate after cardiac arrest. Providing the medical emergency dispatcher with real time chest compression frequency measurements can improve this even further. The dispatcher would be able to give more detailed instructions and reduce unnecessary hands-o time by getting reports that the compres- sions has stopped. The main goal in this thesis was to create an algorithm that can measure the frequency of chest compressions based on video from a smartphone, in real time. Several prototype algorithms has been implemented and tested in MATLAB in order to nd the best method. We used video annotations to compare the results to the reference frequency in several test videos. Our proposed algorithm uses a live camera feed from a smartphone and a combination of di erential motion analysis and discrete Fourier transform, to measure the chest compression frequency in real time. The algorithm utilizes a region of interest strategy to avoid interference from other people in the frame. The nal algorithm has been implemented as an Android application, where all the calculations are done on the phone. The nal algorithm generally shows very good results. It handles both con- tinuous compressions and CPR 30:2, as well as variations in compression rate. There are one major issue however, if the person doing CPR has long and loose hair the measured frequency is often half of the true frequency. Future work will be to x this issue, as well as making a Android library based on the algorithm, so that other applications can easily use it.