Ateneo Innovation Center

Body

AI Ventilator

The COVID-19 pandemic has created an urgent need for mechanical ventilators to treat respiratory issues. This demand has strained healthcare workers and increased the risk of infection due to the surge in COVID-19 cases.

Responding to this problem, the Ateneo Innovation Center (AIC) at Ateneo de Manila University, designed the A-Vent, a portable and cost-effective ventilator.

A-vent: Minimum Viable Ventilator prototype

A-vent-prototype

A-vent is an affordable, open-source non-invasive ventilator that relies on pneumatic components. It uses two air compressors and an oxygen tank to supply air to the patient. The air is filtered through activated carbon and particulate filters in a modified water canister. Pressure is regulated by a stainless angle valve, and pressure levels are monitored with gauges. The system calculates the fraction of expired oxygen (FiO2) to control oxygen delivery.

Safety is ensured through two one-way angle valves that allow volume adjustment. A bubble humidifier adds moisture to the air, and electronic pressure sensors measure the volume of air delivered to the patient. A commercial water meter serves as an airflow meter, and manual monitoring is performed for flow rate. No electronic components are used in this part of the system.

However, as of 2021, many open-source ventilators lacked crucial features, such as data telemetry and processing. These features are essential for telemedicine, which plays a vital role in supporting healthcare services during the pandemic.

ai_ventilator_1
ai_ventilator_2

In order to bridge this gap, Ateneo Innovation Center's researcher. Paul Ryan Santiago introduced expertise in embedded systems to enhance the A-Vent's capabilities. Specifically, it focuses on providing real-time monitoring of patient and ventilator waveforms, along with telemedicine solutions. To achieve this, two AI/ML algorithms are employed:

  • Early Warning Score Recognition

    A multilayer perceptron (MLP) algorithm is used to identify patterns associated with early warning signs and predict clinical risk in patients. This is crucial for timely intervention and care.

  • Ventilator Waveform Interpretation

    A Keras Neural Network is utilized to interpret the ventilator waveform and detect the interaction between the patient and the ventilator. This information aids in understanding how well the ventilator is assisting the patient.

What makes this approach particularly innovative is the use of embedded machine learning, often referred to as TinyML. This involves deploying the AI/ML models mentioned above on embedded devices. By doing so, medical devices equipped with TinyML can analyze biological data in real-time at a low cost. This not only addresses the shortage of healthcare personnel but also offers a safer and more efficient way to track patients' clinical information, especially during a pandemic.

In essence, this research endeavors to enhance the A-Vent ventilator by integrating cutting-edge technology, enabling it to provide better patient care and support healthcare professionals more effectively during these challenging times.

Here are the videos featuring the capabilities of AI Ventilator.

Journals

Ateneo Innovation Center

CTC 416, PLDT Convergent Technologies Center

Ateneo de Manila University Loyola Heights campus

Katipunan Avenue, Loyola Heights

1108 Quezon City

Philippines

Telephone +63 2 8426-6002 ext 5636
Direct Line +63 2 8927-2147

aic.sose@ateneo.edu