PROJECT SUMMARY
Over 100,000 newborns receive mechanical ventilation through an endotracheal tube
(ETT) each year in the United States. Intubating newborns is challenging due to their
size and delicate nature, and unfortunately, nearly 40% of the initial intubation attempts
are incorrect, and the tube is inadvertently placed in the esophagus instead of the
trachea, or too deep in the main stem bronchus, leading to ventilation of only one lung,
or with the tip of the tube too high in the trachea. It is critical to detect malpositioning of
the tube promptly. The goal of this research is to develop Simultaneous Multi-Source
Electrical Impedance Tomography (SMS-EIT) technology for the bedside to correctly
and instantly identify ETT position or malposition. In this application we will combine (1)
deep learning EIT-based confirmation of ETT placement with (2) EIT images of lungs
being ventilated. Together, this would provide clinicians and bedside staff with a real-
time, closed-loop system for determining if (1) the ETT was inserted in the correct
lumen (trachea, not esophagus) and (2) if the lungs are being ventilated appropriately to
detect left or right mainstem bronchial malplacement. The same system with no change
in electrode placement could be used to monitor for inadvertent extubation and for the
onset of emergency conditions such as pneumothorax.
EIT is a noninvasive, non-ionizing functional imaging technique in which images are
formed from voltages measured on electrodes on the body arising from imperceptible
applied currents. Since EIT is a safe and portable technology with no damaging side
effects, it can be used both for continuous monitoring and as needed. Our
interdisciplinary team from GE Research, Colorado State University, and Stanford
University will develop and validate the specialized SMS-EIT system through three
specific aims. The first aim is to develop and implement an electrode configuration,
reconstruction algorithms, and hardware modifications of the GE SMS-EIT system for
the special needs of neonates and this project. In the second aim, training data and a
deep learning classification algorithm to classify intubation as correct, esophageal, too
high, or mainstem bronchial misplacement will be developed. The efficacy and clinical
feasibility of the SMS-EIT system and algorithms for the real-time detection and
classification of ETT malplacement will be evaluated in a study of 30 infants in the Level
IV NICU at Stanford University Medical Center.