2020-862 Early Prediction of Mechanical Ventilation Need in COVID-19 Patients Using Clinician-Constrained Decision Trees

SUMMARY

UCLA researchers in the Department of Computer Science have developed a tree model with machine learning to determine the need for mechanical ventilators, which is important during a shortage of mechanical ventilators in a hospital.

BACKGROUND

Mechanical ventilation is used to help a patient breath when they find it difficult to breathe on their own by pushing airflow into the patient’s lungs. While mechanical ventilators are an essential part on the care of respiratory diseases (such as COVID-19), they can also induce injury by volutrauma or atelectrauma if not used appropriately. Furthermore, shortage of mechanical ventilators makes it crucial to determine who is in need of one when. An assessment tool to predict the need of a respiratory need patients, such as a COVID-19 positive patient, to use a mechanical ventilator.

INNOVATION

UCLA researchers in the Department of Computer Science developed a new class of tree model that utilizes machine learning to predict future mechanical ventilator use. The tree model uses knowledge of hospital workflow and physiologically-relevant know risk thresholds, and incorporates external prior patient information to assess the need of a mechanical ventilator. Especially for patients in the gray zone,  this prediction model will improve patient outcomes especially among positive COVID-19 patients.

POTENTIAL APPLICATIONS

  • Clinical assessment of need for mechanical ventilator
  • Assessment of artificial assistance in patients
  • COVID-19

ADVANTAGES

  • Prediction of need of mechanical ventilator
  • Assistance on clinical workflow

STATUS OF DEVELOPMENT

Initial Conception and prototype.

Patent Information:
For More Information:
Joel Kehle
Business Development Officer
joel.kehle@tdg.ucla.edu
Inventors:
Elior Rahmani
Brian Hill
Zeyuan (Johnson) Chen
Eran Halperin
Vladimir Manuel