Noise Suppression for Dual-Energy CT Through Entropy Minimization

Area of Science:

• Medical Imaging
• Computed Tomography
• Image Processing

Background:

• Noise amplification in dual-energy CT (DECT) limits the usefulness of basis material images.
• Clinically relevant objects often comprise a limited number of materials, suggesting opportunities for targeted noise reduction.

Purpose of the Study:

• To introduce and evaluate an Image-domain Decomposition method through Entropy Minimization (IDEM) for effective noise suppression in DECT.
• To assess the capability of IDEM in preserving spatial resolution and quantitative accuracy compared to existing methods.

Main Methods:

• IDEM transforms decomposed DECT image pixels into 2D data clusters.
• An optimal projection axis is identified by minimizing cluster data entropy.
• Noise suppression is achieved by estimating cluster centers perpendicular to the projection axis, without reducing spatial variation.

Main Results:

• IDEM reduced noise standard deviation in DECT decomposed images by approximately one order of magnitude.
• The method preserved spatial resolution and image noise power spectra (NPS).
• Compared to filtering and iterative methods, IDEM produced high-resolution images with fewer artifacts and less than 2% bias error in electron density measurements.

Conclusions:

• IDEM effectively suppresses noise in DECT for quantitative applications.
• The algorithm uniquely preserves image spatial resolution and NPS while reducing noise.
• IDEM offers a promising approach for improving the diagnostic utility of DECT imaging.