Planning CT-guided robust and fast cone-beam CT scatter correction using a local filtration technique

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Image Processing

Background:

• Cone-beam CT (CBCT) is crucial for image-guided radiotherapy.
• Accurate CBCT imaging is needed for dose calculation and adaptive planning.
• Existing methods for CBCT scatter correction using planning CT (pCT) struggle with anatomical differences, increasing computational cost and limiting performance.

Purpose of the Study:

• To develop a robust and fast CBCT scatter correction method using local filtration and rigid registration (LF-RR).
• To improve CBCT image quality by addressing scatter artifacts.
• To overcome limitations of existing pCT-based scatter correction methods, particularly in cases of large anatomical discrepancies.

Main Methods:

• Rigidly registered pCT with CBCT, followed by forward projection to generate scatter-free projections.
• Calculated raw scatter signals by subtracting scatter-free projections from measured CBCT projections.
• Selected reliable scatter signals based on frequency and intensity criteria, then applied local filtration for global scatter estimation.
• Reconstructed corrected CBCT using the FDK algorithm after subtracting scatter estimation from raw CBCT projections.

Main Results:

• The LF-RR method demonstrated effective scatter removal on an anthropomorphic pelvis phantom, even with intentional anatomical mismatches.
• Quantitative analysis on clinical CBCT images showed significant reduction in CT number errors and spatial non-uniformity compared to uncorrected and other corrected CBCT images.
• The contrast-to-noise ratio between muscle and adipose tissue was improved, and scatter correction was completed within 10 seconds, with corrected volumetric images obtained within 2 minutes.

Conclusions:

• A fast and robust pCT-based CBCT scatter correction method was developed, utilizing local filtration for accurate scatter estimation.
• The LF-RR method is resistant to pCT-to-CBCT registration uncertainties, offering improved imaging accuracy.
• Both phantom and patient studies confirmed the superiority of the proposed method in terms of imaging accuracy and computational efficiency, indicating its potential for clinical application.