Image denoising based on multiscale singularity detection for cone beam CT breast imaging.

Area of Science:

• Medical Imaging
• Radiology
• Signal Processing

Background:

• Cone-beam computed tomography (CBCT) breast imaging offers improved detection of small tumors.
• Current CBCT protocols use X-ray doses comparable to mammography.
• Reducing radiation exposure in medical imaging is a critical goal.

Purpose of the Study:

• To develop an efficient denoising algorithm for CBCT breast imaging.
• To enable a significant reduction in X-ray exposure levels while preserving image quality.
• To enhance the clinical applicability of CBCT for breast cancer screening.

Main Methods:

• A novel wavelet-based denoising algorithm was developed.
• Wavelet coefficients were classified into irregular and edge-related/regular categories.
• Noise reduction strategies were applied based on coefficient classification and decomposition level.

Main Results:

• The algorithm effectively reduced noise in irregular coefficients without introducing artifacts.
• Edge-related and regular coefficients were selectively denoised at the first decomposition level.
• Application to filtered projection images allowed up to a 60% reduction in X-ray dose.
• Clinically acceptable image quality was maintained at reduced exposure levels.

Conclusions:

• The proposed denoising algorithm significantly lowers the radiation dose required for CBCT breast imaging.
• This advancement facilitates safer and more accessible breast cancer detection using CBCT.
• The method holds potential for widespread clinical adoption to minimize patient radiation exposure.