2D Fluoroscopic Image Registration for Target Localization and Patient Monitoring in Image-Guided Radiation Therapy

Many hospitals use image-guided cone-beam computed tomography (CT) to provide qualitative treatment for cancer patients. Although cone-beam CT provides good volumetric images, it takes a while to reconstruct them, which limits its applications because in many cases, real-time spatial information about patient’s internal structures is needed. To overcome this slow image reconstruction problem and missing of the movement data in between the reconstructions, we will use the same cone-beam device to obtain a rapid series of 2D projection images, which have a high temporal resolution but low spatial resolution. In this project, the intern will investigate the possibility and effectiveness of mapping this image sequence onto the high resolution reference 3D image (known as image registration), so that the 3D image provides necessary detail while 2D image sequence gives information about moves and deformations. This will enable continuous detection of inaccuracies in the patient’s position or anatomy, and image guidance for the treatment procedure with precision never possible before. To achieve this, novel algorithmic and software tools for 2D image sequence analysis will be developed. In addition, the team will benchmark existing algorithms for image registration and develop their new variants specifically for our problems, analyze and test these novel methodologies on real large-scale patient data.

Faculty Supervisor:

Dr. Tamas Terlaky

Student:

Olesya Peshko

Partner:

Princess Margaret Hospital

Discipline:

Engineering

Sector:

Life sciences

University:

McMaster University