Non-Contact Monitoring of Respiratory Rate for Infants in the NICU using Depth and RGB Cameras

Speaker: Paul Addison and Soodeh Ahani

November 8th, 2023, 12:00pm

FSC2330, Hyrbid

Join us on November 8th as Dr. Paul Addison and Dr. Soodeh Ahani explore "Non-Contact Monitoring of Respiratory Rate for Infants in the NICU using Depth and RGB Cameras".

Abstract

In our presentation, we explore non-contact respiratory rate estimation for NICU infants, delving into both depth-sensing and RGB video-based methods. We've conducted extensive evaluations with 13 infants, each monitored for 4 consecutive hours through the Mitacs Elevate program.

First, we spotlight Paul Addison's pioneering work in continuous respiratory rate monitoring using depth-sensing cameras. His research underscores the crucial role of accurate respiratory rate monitoring, a vital sign with critical clinical implications. It acts as an early indicator of acute illnesses, offering timely intervention opportunities and significantly impacting patient outcomes.

Expanding on this foundation, I introduce our innovative RGB video-based respiratory rate (RR) estimation method, tailored for sensitive NICU infants. Leveraging Eulerian video magnification (EVM), we extract subtle respiratory signals from video data. We've addressed unique NICU challenges, such as low light, shallow breathing, and swaddled infants.

A standout contribution is our adaptive peak prominence threshold estimation, ensuring accurate RR readings despite these challenges. Our method also includes automatic ROI selection based on motion signals, enabling precise breathing pulse estimation and potential disorder detection.

Additionally, Bland-Altman analysis highlights our algorithm's superior agreement with visually counted RR, surpassing impedance sensors and reference values. This demonstrates our algorithm's potential for accurate NICU RR estimation, even under complex conditions.

In conclusion, our presentation sheds light on non-contact NICU infant respiratory rate monitoring. It showcases depth-sensing and RGB video-based methods, offering a holistic view of this transformative technology's implications for neonatal care.

Event directions

DFP Classroom 2300A (2nd Floor), Forestry Science Centre, 2424 Main Mall.

Once you are inside the Forestry Science building, walk to the rear (South-East) of the building and pass through the large open study area then up the stairs to the 2nd floor student lounge (“Treetop”) area. Turn left, pass through a double door and the room 2300A is the first on the right.

2023 Series

DFP Seminar

About the speaker

Paul Addison and Soodeh Ahani

Paul Addison, PhD, Chief Scientist Medtronic Patient Monitoring, Technical Lead Algorithm and AI Group, Professor of Fluids Engineering | Soodeh Ahani, PhD, Research Associate, UBC

Dr. Paul Addison is the Chief Scientist for Medtronic Patient Monitoring where he is the subject matter expert across various biomedical technologies. He is also technical lead for the Algorithm and AI group in Edinburgh, UK which he set up, and leads the AI landscaping and IP generation activities in the business. Paul joined Medtronic in 2008 through the acquisition of CardioDigital, a start-up he co-founded. Prior to that he was a tenured professor of fluids engineering.

Paul has a Masters in Engineering and a PhD in Fluid Mechanics, both from the University of Glasgow, UK. He has worked on a wide range of biosignals and non-contact monitoring vision technologies and is inventor on 175 US Patents as well as over 200 published articles in the area. He has also written two textbooks in signal processing. Paul is a Medtronic Bakken fellow, one of 70 engineers and scientists tasked with maintaining the highest scientific standards across the company.

Dr. Soodeh Ahani is a seasoned researcher with extensive experience in biomedical signal processing, machine learning, and deep learning. Since 2016, she has closely collaborated with Prof. Guy Dumont at the University of British Columbia (UBC), specializing in the analysis of EEG data for elderly individuals under anesthesia and video data analysis from neonates in the Neonatal Intensive Care Unit (NICU). Soodeh is a recipient of the prestigious Mitacs Elevate grant, which supported a two-year partnership project with Medtronic, focusing on non-contact vital sign detection for NICU infants.

Her research journey encompasses a wide array of expertise, particularly in biomedical signal processing, machine learning, deep learning, and pattern recognition. A significant part of her work involves handling extensive datasets and conducting in-depth research in signal and image processing.

In the upcoming presentation, Soodeh will collaborate with Paul Addison from Medtronic to showcase their groundbreaking research in the NICU. Together, they will demonstrate the transformative potential of non-contact vital sign detection technologies, highlighting the profound impact of their work on neonatal care.