Automated Localization for Image GuidaNce in fetal ultrasound

Research summary

The purpose of this study is to create an ultrasound dataset of sonographer-mother interactions during fetal biometry scans. The dataset will contain ultrasound videos, probe position and orientation (pose), probe accelerations and probe interaction forces as well as infrared/RGB sensor streams and a depth map of the scan area. Data acquisition will be conducted using an instrumented Philips Lumify probe with a Samsung tablet computer for display, as a sonographer performs the scan. The pose of the probe will be tracked with a stereo depth sensor. This data will be used to train a novel machine learning model that learns from expert demonstrations to predict the ultrasound probe’s pose for acquiring standard imaging planes. As such, this project aims to produce probe motion guidance with 6D pose information (3D position and 3D rotation) for obstetric sonographers for standard plane searching in fetal ultrasound. This data will be used to train a novel machine learning model that learns from expert demonstrations to predict the US probe’s pose for acquiring standard imaging planes. The goal of this study is to collect 6D pose (3D position and 3D rotation), force, and a surface map of the scan area in order to train a model for US pose estimation. As such, this project aims to produce probe motion guidance with 3D pose information for obstetric sonographers.

Principal Investigator

Prof Aris Papageorghiou

Contact us

Email: osprea@wrh.ox.ac.uk

IRAS number

349696