Predicting Delay-Bound Violations for Cellular Transmissions: Pre-Hospital ECGs Uploaded from Moving Vehicles

Time-critical data uploads from mobile hosts are challenging when there is high variability in connectivity to the base stations. In these situations, it can be difficult to know in advance if a transmission will meet a delay-bound guarantee. One example of this type of transmission is photographs of 12 lead ECGs. The images are captured by EMTs using the printouts from machines in ambulances, and they are sent to physicians for pre-hospital diagnosis of heart attacks. Current systems do not adapt their operation to the quality of service of the local network, and as a result, they do not provide guarantees or feedback to users about the success or failure of the transmission. A data-centric, application-specific approach and validation method are presented to predict the likelihood of failure in real-time for one of these transmissions. The prediction informs both ends of a voice conversation - between EMT and physician - allowing them to adapt to the knowledge of whether or not the image will be available. Our approach is implemented as an algorithm in an iPhone application that manages the capture and transmission of these diagnostic photographs. Field experiments validate the efficacy of the predictor; it is able to distinguish successful transmissions from failed ones 96% of the time. The framework and end-to-end transmission system are designed to allow for generalizability and ease of extensibility to other networks, cities, and future network improvements.