Faculty Bibliography

PURPOSE/OBJECTIVE:Through-plane motion introduces uncertainty in 3D motion monitoring when using single slice on board imaging (OBI) modalities such as cine MRI. We propose a principal component analysis (PCA) based framework to determine the optimal imaging plane to minimize the through-plane motion for single slice imaging based motion monitoring METHODS: Four-dimensional computed tomography (4DCT) images of 8 thoracic cancer patients were retrospectively analyzed. The target volumes were manually delineated at different respiratory phases of 4DCT. We performed automated image registration to establish the 4D respiratory target motion trajectories for all patients. PCA was conducted using the motion information to define the three principal components of the respiratory motion trajectories. Two imaging planes were determined perpendicular to the second and third principal component respectively to avoid imaging with the primary principal component of the through-plane motion. Single slice images were reconstructed from 4DCT in the PCA-derived orthogonal imaging planes, and were compared against the traditional AP/Lateral image pairs on through-plane motion, residual error in motion monitoring, absolute motion amplitude error, and the similarity between target segmentations at different phases. We evaluated the significance of the proposed motion monitoring improvement using paired t-test analysis. RESULTS:The PCA-determined imaging planes had overall less through-plane motion compared against the AP/Lateral image pairs. For all patients, the average through-plane motion was 3.6 mm (range: 1.6-5.6 mm) for the AP view, and 1.7 mm (range: 0.6-2.7 mm) for the lateral view. With PCA optimization, the average through-plane motion was 2.5 mm (range: 1.3-3.9 mm) and 0.6 mm (range: 0.2-1.5 mm) for the two imaging planes respectively. The absolute residual error of the reconstructed max-exhale-to-inhale motion averaged 0.7 mm (range: 0.4-1.3 mm, 95% CI: 0.4-1.1 mm) using optimized imaging planes, averaged 0.5 mm (range: 0.3-1.0 mm, 95% CI: 0.2-0.8 mm) using an imaging plane perpendicular to the minimal motion component only, and averaged 1.3 mm (range: 0.4-2.8 mm, 95% CI: 0.4-2.3 mm) in AP/Lat orthogonal image pairs. The root mean square error of reconstructed displacement was 0.8 mm for optimized imaging planes, 0.6 mm for imaging plane perpendicular to the minimal motion component only, and 1.6 mm for AP/Lat orthogonal image pairs. When using the optimized imaging planes for motion monitoring, there was no significant absolute amplitude error of the reconstructed motion (p=0.0988), while AP/Lat images had significant error (p=0.0097) with a paired t-test. The average surface distance (ASD) between overlaid 2D tumor segmentation at end-of-inhale and end-of-exhale for all eight patients was 0.6±0.2 mm in optimized imaging planes and 1.4±0.8 mm in AP/Lat images. The Dice similarity coefficient (DSC) between overlaid 2D tumor segmentation at end-of-inhale and end-of-exhale for all eight patients was 0.96±0.03 in optimized imaging planes and 0.89±0.05 in AP/Lat images. Both ASD (p=0.034) and DSC (p=0.022) were significantly improved in the optimized imaging planes. CONCLUSIONS:Motion monitoring using imaging planes determined by the proposed PCA-based framework had significantly improved performance. Single slice image based motion tracking can be used for clinical implementations such as MR Image Guided Radiation Therapy (MR-IGRT).

Purpose: Rapid dose fall off from planning target volume (PTV) is required to spare organs at risk (OAR) that is hallmark of advanced treatment techniques (IMRT, VMAT, and SBRT). Along with conformity index (CI) and homogeneity index (HI) recently, gradient index (GI) was introduced to provide a measurable quality index for dose fall off from PTV. However, it is not clear if GI can be used in IMRT and if there are consistent differences between the two techniques. Methods: Dose volume histogram data for 700 patients equally divided between SBRT and IMRT were retrospectively analyzed. GI was calculated for each patient which is ratio of the volume of at half the prescription (PIV50%) to prescription isodose volume (PIV100%) which was calculated treatment plan and from PTV and CI. Physical distance ratio between OAR and PTV coverage was determined based on the cube root of the GI as an additional tool along with gradient measure distance (rPIV50%-rPIV100%) for the evaluation of SBRT and IMRT plans. Results: The GI varied widely between IMRT and SBRT patients due to inherent differences in techniques. The GIs are nearly constant 4.3 +/-1.1 and 12.8 +/-4.4 for SBRT and IMRT, respectively. The size of PTV is inversely related to GI with data more consistent in SBRT compared to IMRT. The gradient measure increases with PTV size in a well-defined way (0.5-1.5 cm) compared to IMRT where data is widely spread. This can be used as a surrogate for distance between PTV and OAR in IMRT and SBRT. Conclusion: Gradient index and measure are effective parameters in evaluating dose gradient that show consistent differences between treatment techniques. These tools could provide an opportunity for plan evaluation especially the distance from the PTV to OAR to optimize the dose to reduce complications

Dosimetric accuracy of a volumetric modulated arc therapy (VMAT) plan is directly related to the beam model, particularly with multileaf collimator characterization. Inappropriate dosimetric leaf gap (DLG) value can lead to a suboptimal beam model, with significant failure in patient-specific quality assurance (PSQA) of VMAT plans. This study addressed the systematic issue of beam modeling and developed a practical method to determine the optimal DLG value for a beam model. Several complex VMAT plans were selected for the quality assurance analysis using the variable DLG values. The results of three-dimensional (3D) Gamma analysis as a function of the DLG at 3%/3 mm, 2%/2 mm, and 1%/1 mm criteria were fitted by a polynomial curve. The DLG value corresponding to the maximum Gamma passing rate for each polynomial fitting function was derived, and the average was calculated to be the optimal DLG value for each model. The 3D Gamma analysis was repeated with the optimal DLG value to verify the dosimetric accuracy of each VMAT case by PSQA. Gamma passing rates are seen to vary considerably with the DLG values and different analysis criteria (3%/3 mm, 2%/2 mm, and 1%/1 mm) for each case. The optimal DLG derived for each model was 1.16 mm and 1.10 mm, much larger than the measured value (about 0.3 mm). The beam models with the optimal DLG was able to produce an average Gamma passing rate of 97.1% (range, 94.6%- 99.1%) at 3%/3 mm and 93.5% (range, 89.0%- 96.5%) at 2%/2 mm for one beam model, and 97.1% (range, 94.8%- 99.1%) at 3%/3 mm, and 93.3% (range, 88.8%- 96.7%) at 2%/2 mm for another. The overall accuracy of dose calculation for VMAT plans should be optimized with a compromise of varied modulation complexities in a beam model. We have developed a practical method to derive the optimal DLG value for each beam model based on the Gamma passing criterion. This technique should be applicable in general for all beam energies and patient cases.

PURPOSE/OBJECTIVE:The purpose of this study was to assess the dosimetric equivalence of magnetic resonance (MR)-generated synthetic CT (synCT) and simulation CT for treatment planning in radiotherapy of rectal cancer. METHODS:This study was conducted on eleven patients who underwent whole-body PET/MR and PET/CT examination in a prospective IRB-approved study. For each patient synCT was generated from Dixon MR using a model-based method. Standard treatment planning directives were used to create a four-field box (4F), an oblique four-field (O4F) and a volumetric modulated arc therapy (VMAT) plan on synCT for treatment of rectal cancer. The plans were recalculated on CT with the same monitor units (MUs) as that of synCT. Dose-volume metrics of planning target volume (PTV) and organs at risk (OARs) as well as gamma analysis of dose distributions were evaluated to quantify the difference between synCT and CT plans. All plans were calculated using the analytical anisotropic algorithm (AAA). The VMAT plans on synCT and CT were also calculated using the Acuros XB algorithm for comparison with the AAA calculation. RESULTS:Medians of absolute differences in PTV metrics between synCT and CT plans were 0.2%, 0.2% and 0.3% for 4F, O4F and VMAT respectively. No significant differences were observed in OAR dose metrics including bladder V40Gy, mean dose in bladder, bowel V45Gy and femoral head V30Gy in any techniques. Gamma analysis with 2%/2mm dose difference/distance to agreement criteria showed median passing rates of 99.8% (range: 98.5 to 100%), 99.9% (97.2 to 100%), and 99.9% (99.4 to 100%) for 4F, O4F and VMAT, respectively. Using Acuros XB dose calculation, 2%/2mm gamma analysis generated a passing rate of 99.2% (97.7 to 99.9%) for VMAT plans. CONCLUSION/CONCLUSIONS:SynCT enabled dose calculation equivalent to conventional CT for treatment planning of 3D conformal treatment as well as VMAT of rectal cancer. The dosimetric agreement between synCT and CT calculated doses demonstrated the potential of MR-only treatment planning for rectal cancer using MR generated synCT.

BACKGROUND: Interest in MR-only treatment planning for radiation therapy is growing rapidly with the emergence of integrated MRI/linear accelerator technology. The purpose of this study was to evaluate the feasibility of using synthetic CT images generated from conventional Dixon-based MRI scans for radiation treatment planning of lung cancer. METHODS: Eleven patients who underwent whole-body PET/MR imaging following a PET/CT exam were randomly selected from an ongoing prospective IRB-approved study. Attenuation maps derived from the Dixon MR Images and atlas-based method was used to create CT data (synCT). Treatment planning for radiation treatment of lung cancer was optimized on the synCT and subsequently copied to the registered CT (planCT) for dose calculation. Planning target volumes (PTVs) with three sizes and four different locations in the lung were planned for irradiation. The dose-volume metrics comparison and 3D gamma analysis were performed to assess agreement between the synCT and CT calculated dose distributions. RESULTS: Mean differences between PTV doses on synCT and CT across all the plans were -0.1% +/- 0.4%, 0.1% +/- 0.5%, and 0.4% +/- 0.5% for D95, D98 and D100, respectively. Difference in dose between the two datasets for organs at risk (OARs) had average differences of -0.14 +/- 0.07 Gy, 0.0% +/- 0.1%, and -0.1% +/- 0.2% for maximum spinal cord, lung V20, and heart V40 respectively. In patient groups based on tumor size and location, no significant differences were observed in the PTV and OARs dose-volume metrics (p > 0.05), except for the maximum spinal-cord dose when the target volumes were located at the lung apex (p = 0.001). Gamma analysis revealed a pass rate of 99.3% +/- 1.1% for 2%/2 mm (dose difference/distance to agreement) acceptance criteria in every plan. CONCLUSIONS: The synCT generated from Dixon-based MRI allows for dose calculation of comparable accuracy to the standard CT for lung cancer treatment planning. The dosimetric agreement between synCT and CT calculated doses warrants further development of a MR-only workflow for radiotherapy of lung cancer.

Purpose: The purpose of this study is to assess the dosimetric equivalence of MR-generated synthetic CT and conventional CT for treatment planning in radiotherapy of rectal cancer. Methods: This study was conducted on eleven patients who underwent whole-body PET/MR and PET/CT examination in a prospective IRB-approved study. MR data were obtained on a 3T Siemens PET/MR hybrid scanner using a 2-point Dixon sequence. Synthetic CT (synCT) was generated from Dixon MR using a model-based method and then registered to CT using a deformable registration. Rectal tumors were artificially contoured in the synCT by a physician to define a planning tumor volume (PTV) for treatment planning comparison. Standard treatment planning directives were used to create a four-field box (4F), an oblique four-field box (O4F) and a volumetric modulated arc therapy (VMAT) plans on synCT for each PTV. The plans were recalculated on registered planCT with the same MUs as on synCT. Dose-volume metrics and gamma analysis were evaluated between synCT and CT plans. Results: Differences in PTV Dmin, D95% and Dmax between synCT and CT plans across all patients were 0.02 +/- 0.82% for 4F, -0.16 +/- 0.78% for O4F and -0.31 +/- 0.97% for VMAT plans. In any of the treatment techniques, no significant differences were observed in organs at risk (OAR) dose metrics including small bowel (V45 Gy), bladder (V40 Gy) and femoral head (V30 Gy). Gamma Analysis with 2%/2 mm dose difference/distance to agreement showed percentage pass rates of 98.7 +/- 3.1, 98.0 +/- 3.6, and 98.8 +/- 2.7 for 4F, O4F and VMAT, respectively between SynCT and CT plan. Conclusion: Planning on synCT agreed well with the dose recalculated on planning CT for conventional treatment techniques in rectal cancer radiotherapy. These results suggest the potential of MR-only radiotherapy using MR generated synCT