Building and Coding a Non-Invasive Ventilator Using the Arduino Platform to Improve the Clinical Efficacy of Non-Invasive Ventilation Systems by Remotely Controlling the Respiratory Status of COVID-19 Patients

Overview

The purpose of this experiment was to build, code, and test a non-invasive ventilator that could be monitored and controlled from a remote location via a mobile device. During this pandemic, people were being admitted into the hospital suffering from acute respiratory distress syndrome. To ease respiratory strain in these patients, they were connected to non-invasive ventilators (NIVs). Healthcare providers had to enter these patients’ rooms frequently in order to control the pressures on the NIV, risking exposure to the virus. Having a NIV that could be remotely controllable via a mobile device from outside the patient’s room is critical to mitigating the spread of COVID-19. The total production cost of this NIV was $230 and was coded using the Arduino programming language. Regarding the initial testing portion of the prototype, a testing lung used in my home was connected to the NIV in either CPAP or BiPAP and remote mode. I controlled the pressures on my device using the Arduino Io Tapp. When changes were made on the device, the NIV adjusted itself accordingly without physical contact. The clinician can also see the patient’s oxygen saturation on their mobile device to better monitor their progression. The final stage tested the functionality of the NIV at NYU Langone Medical Center apart from its remote capabilities. All tests performed in CPAP and BiPAP modes had percent errors below the 15% maximum accepted in clinical practice, thereby supporting the efficacy of this prototype as a potential alternative for current NIV technologies. Creating a “Third Eye” for the Visually Impaired with Object Classification Kevin Taylor Paul D. Schreiber High School, Port Washington, NY With 285 million visually impaired individuals in the world, it is important for there to be technological innovations aimed at making their everyday lives easier and safer. After spending time volunteering at the Helen Keller National Center and analyzing injury trends, it was determined that, due to the prevalence of the white cane, people who are blind are most vulnerable in their head and neck region. To combat this issue, an object detection system was developed compatible with a hat or headband. In order to create such a device, an Arduino Uno R3 was utilized along with the Arduino IDE coding system. An ultrasonic sensor, buzzer, vibration motor, button, and battery were all connected to the Uno using wires, some of which had to be soldered to utilize the limited space available. Code was then developed that creates bursts of feedback (either sound or vibrations), with set break intervals depending on the distance from the object. For example, an object 150cm away will beep or vibrate once every 5 seconds while an object 50cm away will repeat every 1.4 seconds. Then the button was programmed to allow a user to toggle between vibrational feedback and sound, depending on whether the user is blind, deafblind, or in a social situation that would be more appropriate to use one or the other. Afterwards, it was found that these capabilities could be taken further, allowing a user to be informed on the types of objects in their surroundings, rather than solely the distance. Using Google Colab, a machine learning environment was created catered towards object classification. 24 everyday objects were taken out of the CIFAR-100 dataset. Then utilizing VGG-16, a type of convolutional neural network, a model was trained to recognize these 24 objects. After obtaining a training accuracy of 98% and a testing accuracy near 70%, the model was exported to XCode, allowing it to be run on a mobile device’s camera. Finally, the model was successfully tested in real-world environments, placing all 24 objects in the camera’s view and printing its predictions every second. With text-to-speech, these object recognitions are told to a visually impaired individual, allowing them to make decisions based on the specifics of their surroundings. They would now know to avoid walking into a table or car, and conversely, would know that they are able to sit on a chair or couch. With object classification, the possibilities for future technology in the blind community are endless. 24 specific objects can be expanded to thousands, giving the visually impaired a complete understanding of the world around them.

Competition history

• JSHS 2022 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.