Multi-fidelity high-performance solver for patient-specific treatment planning in therapeutic ultrasound

Project ID: 2531ad1560

(You will need this ID for your application)

UCL Lead department: Mechanical Engineering

Project Summary:

Therapeutic ultrasound is a non-invasive treatment option for conditions like cancer, neurological disorders, and musculoskeletal diseases. It works by focusing ultrasound energy on diseased tissue, sparing surrounding healthy tissue. To ensure success, a patient-specific treatment plan is essential, tailoring ultrasound parameters like frequency, intensity, and focus. This strategy requires quickly and accurately simulating the acoustic field in the body, which presents computational challenges, especially at operating frequencies and realistic domains.

To simulate focused ultrasound, the supervisory team (from UCL and PUC) has been developing mathematical formulations using boundary element methods (BEM) and the open-source Python library OptimUS. OptimUS has been benchmarked internationally, including applications in transcranial neuromodulation, and is used globally in research and education.

The current BEM models in OptimUS solve the 3D linear full-wave propagation problem, represented by the Helmholtz equation, across various tissue types (e.g., soft tissues like the liver and hard tissues like bone or dense tumours). Accurately modelling and rapidly solving ultrasound propagation in domains with mixed soft and hard tissues, particularly at high ultrasound frequencies, presents significant challenges. This project aims to overcome computational limitations by developing a multi-fidelity solver that integrates high-fidelity BEM with low-fidelity surrogate models (e.g., machine learning). Key objectives include creating efficient formulations for mixed tissue domains, designing an optimal computational strategy informed by anatomical data from MRI/CT scans and treatment goals, implementing these advances in OptimUS, and testing them on patient-specific scenarios. The enhanced simulations will improve OptimUS for clinical applications, enabling personalized treatment planning and ultimately improving patient outcomes.

The student will focus on developing surrogate models and multi-fidelity formulations, contributing to OptimUS’s software in Python while applying best practices in software engineering . We seek a student with a background in mathematics, physical sciences, or engineering who is passionate about advancing healthcare through computational research.