Adaptive Real-Time Data Collection Device to Centralize Mechanical Ventilator Monitoring using Pytesseract-OCR

Abstract Respiratory therapists from different hospitals in the Philippines are currently performing room-to-room visitation to patients with respiratory conditions to monitor the data from mechanical ventilator machines. However, this exposes them to various respiratory diseases and puts them at risk of infection. Utilizing Internet-of-Things is imperative to design a remote monitoring system in implementing a solution to lessen the exposure of respiratory therapists and healthcare workers and prevent the possible further spreading of various respiratory diseases. While ventilators can be designed and replaced with built-in telemonitoring capability, acquiring new equipment with such functionality can be costly. This study proposes to design a device that can be attached to old model ventilators to make them able to be remotely viewable in a centralized monitoring station. This study takes advantage of using an open-source optical character recognition tool (Pytesseract-OCR) so that instead of transferring images that require a lot of data, the device will only transfer text data of vital information. The images captured are pre-processed and warped to increase the confidence level of character recognition of Pytesseract-OCR. The device was able to recognize text from ventilator monitors with confidence level of 58.6% to 94.3% with an average of 81.568%. The team were also able to display the captured data to a centralized monitoring station.