Restoration of digital multiplane tomosynthesis by a constrained iteration method

Area of Science:

• Medical Imaging
• Image Processing
• Computational Science

Background:

• Tomosynthetic reconstructions often contain superimposed blurred images from out-of-plane structures.
• This superimposition obscures details within the desired tomosynthetic plane, potentially hindering accurate diagnosis.

Purpose of the Study:

• To develop and validate a constrained iterative restoration method for reducing superimpositions in tomosynthetic imaging.
• To generalize the method for simultaneous deconvolution across multiple planes.

Main Methods:

• A constrained iterative restoration algorithm was developed, generalized for multi-plane deconvolution.
• Theoretical conditions for iterative convergence in noise-free scenarios were derived.
• Noise buildup and limiting noise variance in reconstructions were investigated theoretically and experimentally.

Main Results:

• The proposed method effectively reduces undesired superimpositions in tomosynthetic images.
• Experimental results confirm that satisfying noise-free convergence conditions enhances stability.
• An approximation for limiting noise variance was derived and experimentally verified.

Conclusions:

• The developed iterative restoration method offers a viable solution for improving tomosynthetic image quality.
• Satisfying derived convergence conditions is crucial for stable and effective image restoration.
• Understanding noise buildup aids in establishing formal stopping criteria for the iterative process.