Image/patient registration from (partial) projection data by the Fourier phase matching method

Area of Science:

• Medical Imaging
• Image Registration
• Computational Anatomy

Background:

• Accurate image registration is crucial for medical imaging and patient positioning in radiotherapy.
• Current methods often require image reconstruction, increasing computational time and potential for artifacts.

Purpose of the Study:

• To introduce a novel projection-based Fourier phase matching (PFPM) method for 2D and 3D image/patient registration.
• To enable direct registration from tomographic projection data, bypassing image reconstruction.
• To evaluate the performance and advantages of PFPM compared to existing methods.

Main Methods:

• Developed a PFPM technique utilizing Fourier invariant descriptors and Fourier phase matching in projection space.
• Applied the method to both synthetic and experimental 2D and 3D data, including partial projection data.
• Compared PFPM performance against image-based Fourier phase matching (IFPM) in terms of robustness, efficiency, and registration time.

Main Results:

• PFPM demonstrated comparable performance to IFPM for 2D image registration (translation, rotation, scaling).
• PFPM offers advantages such as subpixel resolution and increased insensitivity to image noise.
• The method successfully generalized to 3D registration and partial projection data, enabling real-time applications.

Conclusions:

• PFPM provides an efficient and robust solution for 2D/3D image and patient registration directly from projection data.
• Eliminating the need for image reconstruction reduces computational load and artifacts in radiotherapy setup verification.
• The method's ability to handle partial data and achieve real-time registration opens possibilities for dynamic verification in tomotherapy.