AI-EgyVent: Artificial Neural Network Assisted Control based Low-Cost Ventilator Implementation in Response to COVID-19 Pandemic

It has become crucial to develop low-cost ventilators as the demand for these life-saving devices has increased during the COVID-19 pandemic. A pneumatic system of four ventilation modes is implemented with an overall cost under ${\$}$400. The proposed design utilizes sensor fusion to achieve precise flow values even if low-accuracy sensors are used. Fluke biomedical monitoring device is used to verify the functionality correctness of the ventilator. An Artificial Neural Network (ANN) is used to enhance the PID control process of the ventilator, providing better accuracy and faster response times. The ANN is trained using a dataset generated from the ventilator system using two test lungs with different compliance values. The proposed ANN achieves a high determination coefficient of 87.97% and a low mean absolute error equals to 3.26 while maintaining a low number of parameters equals to 70, to be feasible for the hardware implementation.