Fast triage of Covid-19 patients in hospitals by means of remote respiration rate determination

Due to the Covid-19 pandemic, hospitals are experiencing a significant workload of patients with respiratory diseases. These highly contagious patients require special treatment and isolation to protect other patients and medical staff. Particularly critical is the central admission in the hospital, where all actors come together unprotected. Therefore, a fast triage with a separation into potential Covid-19 patients and the remaining patients is crucial. The two typical symptoms of Covid-19 are fever and coughing or shortness of breath due to a severe infection of the lower respiratory tract. When recording these two symptoms, care must be taken to ensure that no care givers are infected and that there is no physical contact between the measuring devices. While fever can be detected relatively easily and quickly with a non-contact thermal imaging camera, the determination of shortness of breath is more difficult. As part of a rapid response to these requirements, we propose a deliberately simple measurement setup in the form of an RGB camera and a connected computing unit as well as a method for image-based detection of the respiratory rate. Our method first determines a ROI on the thorax, since this is where the greatest movement occurs during breathing. Subsequently, Minimum Eigen-features are detected in the ROI and tracked continuously by means of the optical flow according to Lucas-Kanade. In addition, a bandpass filter, a PCA and a frequency determination using FFT are applied. An easy-to-use GUI ensures that the application can be operated by anyone. The accuracy of the measurement method was evaluated with the NeXus-10 MKII. Currently, the proposed system is being used at the Universitätsklinikum Essen. At present, the measuring interval plus processing takes about one minute. We aim to halve this time in the future in order to achieve a higher frequency of measurements.