Machine learning for ultrasound-guided interventions

Lead Research Organisation: University College London
Department Name: Medical Physics and Biomedical Eng

Abstract

1) Brief description of the context of the research including potential impact

Ultrasound is the most widely used imaging technique in medicine. Three-dimensional (3D) ultrasound, in which 3D images of anatomical and pathological lesions are acquired, has been feasible for many years. Despite the usefulness of this technique for a wide range of applications, most clinical ultrasound imaging still involves acquiring two-dimensional (2D) images, which represent cross-sections through the body. The so-called "freehand 3D ultrasound" approach enables 3D images to be acquired using a standard (2D) handheld clinical ultrasound probe but requires an additional 3D tracking device to measure the position of the probe as it is moved during a scan. Such devices are highly specialised and add significant cost.

2) Novelty of the research methodology
Recent advances in artificial intelligence (AI) open up the exciting possibility of image-based probe motion estimation without the need for an external tracking device. This novel approach has its foundation in the ability of humans can perceive patterns in images and estimate motion while scanning and correlated imaging features such as speckle patterns between adjacent 2D images. The resulting efficient AI models have the potential to dramatically increase the accessibility to 3D ultrasound imaging through simpler and affordable computer-assisted systems for imaging and training.

3) Aims and objectives

In the proposed research, the core focus and objective output of the work will encompass the creation of a robust solution to the problem of estimating probe motion and image location from real-time 2D ultrasound scans for 3D ultrasound imaging.

4) Alignment to EPSRC's strategies and research areas

This research aligns to the EPSRC's alignments with artificial intelligence technologies and medical imaging through its use of artificially intelligent methods for 3D image reconstruction of ultrasound imaging for use in medical applications.

5) Any companies or collaborators involved

None

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/S021930/1 30/09/2019 30/03/2028
2361326 Studentship EP/S021930/1 22/09/2019 29/09/2023 Zachary Michael Baum