Robotic Ultrasound-Guided Instrument Localization in Fetoscopy

Abstract

Fetoscopic Endoluminal Tracheal Occlusion (FETO) is a minimally invasive fetal surgery (MIFS) aimed at mitigating the effects of Congenital Diaphragmatic Hernia (CDH). During FETO, a latex balloon is introduced in the fetal trachea using a fetoscope. Typically, this surgery is performed under ultrasound guidance which is provided by a sonographer who manually operates the ultrasound probe. This manual operation imposes a considerable physical and cognitive demand, placing a burden on the sonographer. This paper proposes a robotic ultrasound-based instrument tracking system that automates the probe position control while ensuring continuous visibility of the fetoscope in ultrasound images. The development of the proposed system is achieved with the completion of two tasks. Firstly, a series of fetoscope localization algorithms are developed and compared. Secondly, a task-based control for a robotic ultrasound system is developed. The localization algorithms’ performance is evaluated on annotated ultrasound datasets. The OEU-Net algorithm is selected based on this evaluation and is implemented in the instrument tracking system. The performance assessment of the tracking system shows that it is capable of tracking the fetoscope with a mean error below 4 mm. Thus, the developed system represents a significant advancement toward automatic robotic assistance for ultrasound guidance during FETO.