Selective-diffusion regularization for enhancement of microcalcifications in digital breast tomosynthesis...

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Digital breast tomosynthesis (DBT) improves mass detection but faces challenges in identifying subtle microcalcifications due to large breast volumes.
• Simultaneous algebraic reconstruction technique (SART) used in DBT can amplify image noise with increased iterations, impacting microcalcification detection.
• Noise regularization is crucial for enhancing the contrast-to-noise ratio (CNR) of microcalcifications in DBT images for both human and machine interpretation.

Purpose of the Study:

• To develop and evaluate a selective-diffusion (SD) method for noise regularization with SART in DBT.
• To improve the contrast-to-noise ratio (CNR) of microcalcifications in DBT slices.
• To enhance the detection of subtle microcalcifications for improved diagnostic accuracy.

Main Methods:

• The selective-diffusion (SD) method was integrated into the SART reconstruction process to regularize noise during projection view updates.
• Local gradient information was used to differentiate microcalcifications from background noise, applying differential regularization to signal and noise classes.
• The SD method was compared against quadratic Laplacian (QL), total variation (TV), and nonconvex total p-variation (TpV) methods using DBT scans of a phantom and a human subject, evaluating CNR and full width at half maximum (FWHM).

Main Results:

• The SD method achieved comparable CNR to the nonconvex TpV method for large microcalcifications, outperforming QL and TV methods.
• For subtle microcalcifications, the SD method demonstrated superior CNR compared to all other evaluated methods.
• The SD method preserved microcalcification sharpness, showing comparable FWHM to non-regularized images for large microcalcifications and to TV methods for subtle ones.

Conclusions:

• The selective-diffusion (SD) method effectively regularizes SART reconstruction, enhancing microcalcification CNR while preserving image sharpness.
• SD regularization proved superior to QL, TV, and TpV methods, particularly for improving the detection of subtle microcalcifications in DBT.
• This technique holds promise for improving the accuracy of microcalcification detection in digital breast tomosynthesis.