Encrypted login | home

Program Information

Liver 4D-CBCT Imaging by a Motion Modeling and Biomechanical Modeling-Guided Reconstruction Technique (MM-Bio-Recon)

Y Zhang

Y Zhang1*, J Meyer1 , L Ren2 , J Nasehi Tehrani1 , J Wang1 , (1) UT Southwestern Medical Ctr at Dallas, Dallas, TX, (2) Duke University Medical Center, Durham, NC


TH-EF-605-5 (Thursday, August 3, 2017) 1:00 PM - 3:00 PM Room: 605

Purpose: To develop a motion modeling and biomechanical modeling-guided technique(MM-Bio-Recon) to enable accurate liver 4D-CBCT reconstruction from phase-sorted sparse-view projections.

Methods: MM-Bio-Recon reconstructs new CBCTs by deforming a prior high-quality CT/CBCT using a deformation-vector-field(DVF), which is developed on the foundation of ‘2D-3D deformation’. However, 2D-3D deformation solves the DVF purely by intensity-matching 2D projections simulated from the deformed volume to acquired on-board projections, whose accuracy is limited in low-contrast liver regions. MM-Bio-Recon incorporates two strategies to enhance the accuracy of 2D-3D deformation: 1. Liver boundary motion modeling from a prior liver 4D-CT/4D-CBCT; and 2. Intra-liver DVF optimization by biomechanical modeling. Specifically, MM-Bio-Recon first contours and density-overrides the liver volume from each 4D-CT/4D-CBCT phase to enhance its boundary contrast(inferior boundary especially), followed by inter-phase deformable registration and principal-component-analysis to extract a motion model. From the patient-specific motion model and phase-specific on-board projections, DVFs are initialized and subsequently input into 2D-3D deformation for refinement. The refined DVFs at the liver boundary are used to drive finite-element-analysis based biomechanical modeling to fine-tune intra-liver DVFs. The biomechanically-corrected DVFs are fed back into 2D-3D deformation for further refinement, which forms an iterative loop until convergence.The efficacy of MM-Bio-Recon was evaluated on both the extended-cardiac-torso(XCAT) digital phantom and liver cancer patients. The low-contrast tumors contoured on the prior images were deformed by MM-Bio-Recon to new tumor contours, and compared against ‘ground-truth’ ones for evaluation.

Results: By using 20 projections for reconstruction, the average(±s.d.) DICE coefficients between the deformed liver tumor contours and the ‘ground-truth’ ones for the XCAT study are 0.57±0.31, 0.83±0.21, and 0.89±0.11 for 2D-3D deformation, Bio-Recon(without motion modeling) and MM-Bio-Recon techniques, respectively. The corresponding results for the patient study are 0.58±0.14, 0.81±0.13 and 0.88±0.07, respectively.

Conclusion: MM-Bio-Recon substantially improves the accuracy of solved DVFs for liver tumor propagation/tracking, dose accumulation and adaptive radiotherapy.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by grants from the American Cancer Society (RSG-13-326-01-CCE), from the US National Institutes of Health (R01 EB020366), and from the Cancer Prevention and Research Institute of Texas (RP130109).

Contact Email: