Few-view cone-beam CT reconstruction with deformed prior image

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Prior images enhance cone-beam CT (CBCT) reconstruction quality from sparse-view or low-dose projections.
• Geometric deformation between prior and target images can degrade reconstruction performance.
• Addressing this mismatch is crucial for accurate CBCT imaging in clinical settings.

Purpose of the Study:

• To develop and evaluate a deformed prior image-based reconstruction (DPIR) strategy.
• To mitigate the geometric deformation between prior and target CBCT images.
• To improve the quality of CBCT images reconstructed from sparse-view or low-dose projections.

Main Methods:

• A projection-based registration approach estimates deformation vector fields.
• Iterative matching of forward-projected deformed prior images with on-treatment projections.
• The deformed prior image is integrated into the prior image constrained compressed sensing (PICCS) algorithm.

Main Results:

• The deformed prior image exhibits closer geometric alignment with the on-treatment CBCT compared to the original prior.
• DPIR strategy shows improved performance over standard PICCS for few-view projections.
• In phantom studies, DPIR reduced root mean squared error (10% vs. 14%) and increased universal quality index (0.92 vs. 0.88).

Conclusions:

• The DPIR approach offers a practical solution for prior-target image geometric mismatch.
• DPIR enhances the performance of PICCS for CBCT reconstruction with limited projection data.
• This method improves the reliability of CBCT imaging in scenarios with sparse or low-dose data acquisition.