Encrypted login | home

Program Information

Potential Gains of Incorporating Iterative Reconstruction Into 4DCT

C Glide-Hurst

C Glide-Hurst1*, R Price2, J Adams2 IJ Chetty1 (1) Henry Ford Health System, Detroit, MI, (2) Wayne State University, Detroit, MI

TH-C-103-9 Thursday 10:30AM - 12:30PM Room: 103

Purpose: IGART requires high quality, volumetric imaging for routine feedback and cumulative dose analysis. We propose to incorporate iterative reconstruction (IR) into 4DCT to reduce imaging dose while maintaining image quality, to facilitate routine 4DCT for IGART.

Methods: Static thorax phantom CTs (100-500 mAs) and 4DCTs (200-800 mAs) of an in-house deformable lung phantom undergoing respiration were acquired. IR was applied to phantom data at Levels 1-6 (L1-L6), corresponding to noise reduction factors of 0.89-0.55, with respect to filtered backprojection (FBP). Mean and standard deviation (μ and σ (image noise), respectively) were calculated via region of interest (ROI) analysis. Contours were assessed via Dice Similarity Coefficients (DSCs). Profile analysis evaluated boundary preservation between IR levels. Contrast to noise ratio (CNR) and dose savings were quantified. Raw 4DCT projections for 10 patient datasets (5 lung, 5 liver, 120-140 kVp, ~800 mAs) were retrospectively reconstructed with FBP and IR levels L1-L6. ROI analyses (tumor and nearby lung/liver) were conducted.

Results: For the thorax phantom, comparable noise was obtained between L6 IR at 100 mAs and FBP at 600 mAs at 6-fold dose savings. In the deformable phantom, maximum CT number discrepancy occurred between FBP and L6 (4 HU, 300 mAs). Profile analysis revealed negligible boundary differences between FBP and IR. CNR increased with increased IR level across mAs settings (0.7-2.2% for L3 and 1.4-3.5% for L6 with respect to FBP). Near perfect DSC was calculated at all IR levels. Substantial noise reduction was observed for all liver cases (L3 and L6 reduced ~24 ± 2.8% and 45.2 ± 2.9%, respectively), although not for lung (~2-4%).

Conclusion: IR reduced noise while improving CNR and preserving boundary information. ~6-fold dose reduction was possible while maintaining similar noise characteristics. A prospective trial is warranted to optimize mAs and IR, and further integrate IR-4DCT into IGART.

Funding Support, Disclosures, and Conflict of Interest: The submitting institution holds research agreement with Philips Healthcare.

Contact Email: