Machine learning based reconstruction for laser diode photoacoustic imaging

Photoacoustic (PA) imaging has shown great potential in cancer detection, tissue characterization, and therapeutic applications. Laser diode-based PA imaging is a new development in this field that can be used for image guidance, nerve or cancer imaging, and the localization of a needle target during biopsy procedures. The PA imaging requires very accurate time synchronization between the laser system and data acquisition system which is not available in widely used clinical ultrasound machines. In this project, a machine learning based photoacoustic reconstruction without the need for such synchronization will be proposed. The location and intensity information of a point source will be recovered based on the recorded channel data. The proposed method has two benefits: first, it democratizes the use of PA imaging by allowing conventional ultrasound systems to be used for PA. Second, it also allows us to use the ultrasound machines beyond its frequency limitation. This allows for monitoring fast changing and dynamic targets, such as nerve activities.

Faculty Supervisor:

Septimiu (Tim) Salcudean

Student:

Partner:

Eindhoven University of Technology

Discipline:

Engineering

Sector:

Education

University:

The University of British Columbia

Program: