Treffer: A Real-Time Heart Attack Detection and Warning System for Drivers Using Neural Network.

The high rate of vehicle accidents is a major cause for concern, especially those caused by drivers suffering heart attacks while driving. Such accidents lead to a tragic loss of life and significant material damage, posing a serious threat not only to the drivers themselves but also to everyone with whom they share the road. Several solutions have been proposed to identify heart attack risk factors among drivers, but they have proven to be inadequate and lack accuracy and speed. In addition, these methods often require drivers to perform complex tasks during a heart attack due to the complex procedures involved in the examination process. This study determines the design of a system capable of identifying the heart attack risk areas. It also implements an alert mechanism that can save the life of a driver who has a heart attack while driving and at the same time reduce the risk of accidents involving pedestrians or other drivers. The system starts by recognizing the driver through a camera inside the car, then makes a quick check by collecting data from sensors and wearable devices such as heart rate, body temperature, and blood pressure, then processes it to detect possible heart attack risks in a non-surgical and cost-effective way. The Max30100 sensor is used to collect heart rate and blood oxygen levels, while the MLX90614 sensor captures body temperature. The ESP32 board acts as a bridge connecting the sensors to the Jetson Nano board. Blood pressure data is collected through wearable devices. Then all this data is processed using a Neural Network (NN) algorithm, which is implemented on an intelligent microcontroller built into the Jetson Nano board. The warning system is triggered if the algorithm determines that the user is at risk of a heart attack. The system was evaluated using a dataset created specifically for this study by collecting data for 1467 real cases, some of which suffered heart attacks, others did not experience problems with the change in their vital symptoms. A comparison was made between the SVM, random forest, logistic regression, and NN algorithms using Python. Used 20% for validation, the metrics using F1 score, recall, precision, and accuracy, which achieved 99.2% on the NN algorithm. [ABSTRACT FROM AUTHOR]