Technical Note: Noise models for virtual clinical trials of digital breast tomosynthesis

Area of Science:

• Medical Imaging
• Computational Phantoms
• Image Noise Modeling

Background:

• Virtual clinical trials (VCT) require accurate noise simulation for digital breast tomosynthesis (DBT).
• Existing noise models may not fully capture the complexities of quantum noise and gain variations in DBT systems.

Purpose of the Study:

• To evaluate an affine-variance noise model incorporating correlated quantum noise and spatially dependent quantum gain for DBT VCT.
• To assess the model's accuracy in simulating noise for both direct and indirect conversion detectors.

Main Methods:

• A VCT framework generated noise-free projections of a simulated phantom.
• The noise model introduced correlated noise and spatially dependent gain to simulate noisy DBT data.
• Simulated and real DBT projections were compared using signal-to-noise ratio (SNR) and normalized noise power spectrum (NNPS).

Main Results:

• The affine-variance noise model achieved simulation errors below 4.4% (SNR) and 7.0% (NNPS) compared to clinical data.
• Uncorrelated noise models introduced significant errors (up to 65.8%) for indirect detection systems.
• Assuming spatially independent quantum gain resulted in 11.2% error.

Conclusions:

• The proposed noise model accurately reproduces clinical DBT noise characteristics.
• Correlated noise simulation is essential, particularly for indirect conversion detectors.
• Uncorrelated noise assumptions are acceptable for direct detectors with minimal crosstalk.