1. Liver CBCT Estimation Using Limited-view Projections and A Biomechanical Model
You Zhang, Jeffrey Meyer, Joubin Nasehi Tehrani, Jing Wang
UT Southwestern Medical Center
Dallas, Texas, USA
Introduction
Delivering effective radiation therapy for liver cancer requires daily tumor localization prior to treatment, but respiration-induced liver motion complicates this task in the clinic. We have developed a biomechanical modeling-guided cone-beam computed tomography estimation technique (Bio-CBCT-est) to generate new liver 4D-CBCT images by deforming a prior high-quality CT/CBCT using deformation vector fields (DVFs), which enables fast and accurate on-board liver tumor localization to guide the radiotherapy treatment.
Materials and Methods
To solve the DVFs, the Bio-CBCT-est technique employs an iterative scheme that alternates between intensity-driven 2D-3D deformation and biomechanical modeling-guided DVF optimization. The 2D-3D deformation step solves DVFs by matching digitally reconstructed radiographs of the 3D deformed prior volume to 2D phase-sorted on-board projections according to imaging intensities (Fig. 1).
Since the intensity-driven 2D-3D deformation technique incorporates high-quality prior information into the new image estimation process, it potentially enables accurate, high-quality image estimation using only sparse-view projections to save imaging time and dose. However, for liver imaging, the liver tumors’ low contrast limits the accuracy of 2D-3D deformation, as the liver tumor fails to provide sufficient intensity variation from the liver parenchyma to drive the intensity difference-based deformation.
Figure 1. The general scheme of the intensity-driven 2D-3D deformation technique.
To boost the DVF accuracy in these low-contrast regions, we use the intensity-driven DVF solved at higher-contrast liver boundaries to fine-tune the intra-liver DVF by finite element analysis-based biomechanical modeling (Fig. 2). The biomechanically-corrected DVFs are then fed back into 2D-3D deformation to form an iterative loop until final convergence.
Figure 2. The general scheme of the biomechanical modeling-based intra-liver DVF fine-tuning process.
Evaluation Results
The clinical FDK algorithm generated CBCT images with excessive streak artifacts that prevent accurate tumor localization, because of the 20 sparse-view projections used for reconstruction. In contrast, the 2D-3D deformation and Bio-CBCT-est algorithms preserved the high-quality information from the prior CT volume through the DVF-driven image estimation approach (Fig. 3). However, 2D-3D deformation did not accurately deform the low-contrast tumor regions. By comparison, the Bio-CBCT-est technique estimated tumor regions that matched better with the “gold-standard” CT reference image.
When applied to estimate different respiratory phases of 4D-CBCT (Table 1), the Bio-CBCT-est technique generated automatic liver tumor contours (by DVF propagation) with significantly improved DICE values compared to the intensity-driven 2D-3D deformation technique alone ( < 10 −6 by Wilcoxon signed-rank test).
Figure 3. From left to right: three-view slice cuts of the prior CT volume (before DVF-based deformation), the CBCT volume reconstructed by FDK, the CBCT volume estimated by 2D-3D deformation, the CBCT volume estimated by Bio-CBCT-est, and the “gold-standard” CT reference volume. CBCT reconstructions/estimations used 20 projections.
Conclusion
The Bio-CBCT-est technique substantially improves the quality and accuracy of the estimated liver 4D-CBCT images. The liver tumor contours are correctly deformed to match the manual contours on “gold-standard” 4D-CT reference images, indicating that the technique can improve the registration and localization accuracy of liver tumors to promote safe and effective liver cancer radiation therapy.
Table 1. DICE similarity values between the automatic tumor contours on the estimated CBCT images propagated by solved DVFs, and the manually contoured tumors on the “gold-standard” CT reference images (with 4D phases 10-90%) for patients 1-4, using different estimation methods and 20 projections.