Feasibility Study of Dose Reduction in Digital Breast Tomosynthesis Using Non-Local Denoising Algorithms

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Digital Breast Tomosynthesis (DBT) involves higher radiation doses than standard mammography due to "combo-mode" imaging.
• Reducing radiation dose in DBT is crucial for patient safety, but it increases quantum noise and degrades image quality.

Purpose of the Study:

• To evaluate the effectiveness of denoising algorithms in reducing radiation dose for DBT examinations.
• To assess if denoising techniques can maintain acceptable image quality despite dose reduction.

Main Methods:

• Two "state of the art" denoising techniques, Non-local Means (NLM) and Block-matching 3D (BM3D), were investigated.
• DBT projections were acquired from a breast phantom at various radiation dose levels.
• Optimal filtering parameters were determined to recover image quality.

Main Results:

• BM3D demonstrated superior noise reduction and less image blurring compared to NLM.
• BM3D achieved better noise adjustment (mean difference in peak signal to noise ratio < 0.1dB).
• BM3D resulted in less image blurring (mean difference in image sharpness ~ 6%).

Conclusions:

• Denoising algorithms, particularly BM3D, show significant potential for reducing radiation dose in DBT.
• BM3D effectively mitigates noise and preserves image quality in low-dose DBT projections.
• These findings support the clinical implementation of dose reduction strategies in DBT imaging.