A nonrigid registration of MR breast images using complex-valued wavelet transform

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Image Processing

Background:

• Accurate registration of dynamic breast MRI is crucial for detecting subtle changes and improving diagnostic accuracy.
• Nonrigid motion in dynamic imaging can lead to misregistration, obscuring important diagnostic information.

Purpose of the Study:

• To present a novel, fast, slice-by-slice, nonrigid registration algorithm for dynamic breast MRI.
• To evaluate the performance and clinical utility of this new registration method.

Main Methods:

• The algorithm employs multiresolution motion estimation using complex discrete wavelet transform (CDWT).
• A hierarchical phase-matching algorithm based on oriented complex subimages was utilized for motion estimation.
• Performance was assessed using correlation coefficient and normalized mutual information, alongside clinical scoring.

Main Results:

• The CDWT-based registration significantly improved correlation coefficient and normalized mutual information compared to pre- and post-contrast images.
• The proposed nonrigid registration outperformed rigid and affine registration methods.
• Clinical validation showed enhanced readability of subtraction images, with better definition of breast and lesion borders.

Conclusions:

• The developed CDWT registration algorithm offers a fast and effective solution for nonrigid motion correction in dynamic breast MRI.
• This method improves lesion detectability and border definition, aiding in more accurate breast cancer diagnosis.