Preliminary Evaluation of a Novel Neural Network-Based Hybrid Simulator for Surgical Training and Performance Assessment of Neonatal Thoracentesis.

Abstract

Tension pneumothorax is a rare and life-threatening situation in neonates requiring immediate intervention through thoracentesis. Significant complications can arise while performing thoracentesis in the case of inadequate skill level or exposure to the condition. Although simulation-based training (SBT) has proven to be effective in learning surgical skills, training sessions are long, subjective, and expensive, because of which they cannot be held regularly. This article attempts to improve traditional SBT for neonatal thoracentesis through an autonomous simulator that can provide real-time objective feedback during surgical training and assessment. The simulator incorporates a custom manikin and virtual reality software interfaced through electromagnetic sensors that track the motion of surgical instruments. The software application reads and stores instrument motion information to replicate physical actions in the virtual environment, play back previously stored surgical performances and analyze data through a pretrained neural network. The simulator encapsulates the experience of SBT by allowing trainees to watch and replicate an ideal method of conducting the procedure, providing simplified, real-time autonomous guidance during practice and an objective taskwise assessment of the performance during testing. The preliminary trial held at the University of Illinois Hospital in the presence of 1 neonatologist and 4 fellows revealed that all the participants used the autonomous guidance more than once, and all found simulation experience to be accurate and overall effective in learning thoracentesis. Although the sample size is small, the simulator shows potential in being a viable alternative approach for training and assessment for thoracentesis.