Clinical Study of Orthogonal-View Phase-Matched Digital Tomosynthesis for Lung Tumor Localization

Abstract

Background and purpose: Compared to cone-beam computed tomography, digital tomosynthesis imaging has the benefits of shorter scanning time, less imaging dose, and better mechanical clearance for tumor localization in radiation therapy. However, for lung tumors, the localization accuracy of the conventional digital tomosynthesis technique is affected by the lack of depth information and the existence of lung tumor motion. This study investigates the clinical feasibility of using an orthogonal-view phase-matched digital tomosynthesis technique to improve the accuracy of lung tumor localization.

Materials and methods: The proposed orthogonal-view phase-matched digital tomosynthesis technique benefits from 2 major features: (1) it acquires orthogonal-view projections to improve the depth information in reconstructed digital tomosynthesis images and (2) it applies respiratory phase-matching to incorporate patient motion information into the synthesized reference digital tomosynthesis sets, which helps to improve the localization accuracy of moving lung tumors. A retrospective study enrolling 14 patients was performed to evaluate the accuracy of the orthogonal-view phase-matched digital tomosynthesis technique. Phantom studies were also performed using an anthropomorphic phantom to investigate the feasibility of using intratreatment aggregated kV and beams' eye view cine MV projections for orthogonal-view phase-matched digital tomosynthesis imaging. The localization accuracy of the orthogonal-view phase-matched digital tomosynthesis technique was compared to that of the single-view digital tomosynthesis techniques and the digital tomosynthesis techniques without phase-matching.

Results: The orthogonal-view phase-matched digital tomosynthesis technique outperforms the other digital tomosynthesis techniques in tumor localization accuracy for both the patient study and the phantom study. For the patient study, the orthogonal-view phase-matched digital tomosynthesis technique localizes the tumor to an average (± standard deviation) error of 1.8 (0.7) mm for a 30° total scan angle. For the phantom study using aggregated kV-MV projections, the orthogonal-view phase-matched digital tomosynthesis localizes the tumor to an average error within 1 mm for varying magnitudes of scan angles.

Conclusion: The pilot clinical study shows that the orthogonal-view phase-matched digital tomosynthesis technique enables fast and accurate localization of moving lung tumors.

Keywords: MV imaging; digital tomosynthesis; orthogonal-view; phase-matched DTS; tumor localization.

Figure 1.

The acquisition scheme for CBCT and DTS. The CBCT is reconstructed from projections covering a full scan angle. The DTS image is reconstructed from limited-angle projections. CBCT indicates cone-beam computed tomography; DTS, digital tomosynthesis.

Figure 2.

A, The single-view DTS that is reconstructed by limited-angle projections from a single direction. B, The proposed orthogonal-view DTS that acquires complementary projections from orthogonal directions for reconstruction to improve the depth information. Note that the scan angle in (B) is halved at each scan direction to achieve the same total scan angle as (A). DTS indicates digital tomosynthesis.

Figure 3.

Localization scheme of DTS: a RDTS is synthesized from the RCT and registered to the onboard DTS for tumor localization. Each DRR is projected to match the corresponding OBP based on the same imaging geometry. DRR indicates digitally reconstructed radiograph; RDTS, reference digital tomosynthesis; OBP, onboard projection; RCT, reference computed tomography.

Figure 4.

Localization scheme of phase-matched DTS: a RDTS is synthesized from the 4D-RCT and registered to the onboard DTS for tumor localization. Each DRR is projected to match the corresponding OBP based on the same respiratory phase and imaging geometry. 4D-RCT indicates 4-dimensional reference computed tomography; DRR, digitally reconstructed radiograph; RDTS, reference digital tomosynthesis; OBP, onboard projection.

Figure 5.

Different scan angle schemes evaluated in this study: the first row is for single-view DTS and the second row is for orthogonal-view DTS. Note that for the 0° scan, only one projection was acquired. DTS indicates digital tomosynthesis.

Figure 6.

The axial CT slice image of the CIRS phantom and the 150° dynamic conformal arc plan designed for treating the target. CIRS indicates computerized imaging reference systems; CT, computed tomography.

Figure 7.

Generation of the synthetic kV projections based on a linear relationship derived between kV and MV projection intensities using pretreatment projection pairs. Note that the synthetic kV projection is slightly truncated at the boundary to remove the leakage signal from the multileaf collimator leaves.

Figure 8.

A, Scheme of incorporating kV imaging and BEV cine MV imaging into plan delivery through the Varian TrueBeam research mode. B, Flowchart of using the aggregated kV and BEV cine MV projections to reconstruct OV-PMDTS for tumor localization. Pretreatment is the stage at which a linear correlation between kV and MV projection intensities is derived using pretreatment kV and MV projections. The linear correlation is then applied to intratreatment BEV cine MV projections to synthesize kV projections. Note that the 2 intratreatment kV projection sets marked by stars are the same projection set duplicated to facilitate scheme drawing. BEV indicates beams’ eye view; OV-PMDTS, orthogonal-view phase-matched digital tomosynthesis.

Figure 9.

The scheme for continuous DTS localization in the kV–MV study. A verification point is assessed when the gantry rotates every 10°. For each verification point, the orthogonal-view DTS is reconstructed using 10° kV projections and 10° BEV cine MV projections for tumor localization. BEV indicates beams’ eye view; DTS, digital tomosynthesis.

Figure 10.

Patient OBDTS images reconstructed from single-view projections (A and B) and orthogonal-view projections (C and D). The total scan angle (15°) is preserved through splitting it into equal halves (7.5°) for the orthogonal-view acquisition. OBDTS indicates onboard digital tomosynthesis.

Figure 11.

Visual comparison between the OBDTS images, the RDTS images synthesized by the conventional DTS technique, and the RDTS images synthesized by the phase-matched DTS technique. The OBDTS images are reconstructed from single-view 30° projections. Each row shows image sets of a different patient. The left column shows the OBDTS images. The middle column shows the RDTS images synthesized using the conventional DTS technique. The right column shows the RDTS images synthesized using the phase-matched DTS technique. OBDTS indicates onboard digital tomosynthesis; RDTS, reference digital tomosynthesis.

Figure 12.

Comparison of tumor localization accuracy (relative to CBCT) between different DTS techniques for the retrospective patient study. Each boxplot contained 30 localization results corresponding to the 30 cone-beam projection sets studied for the 14 patients. The localization errors of the DTS techniques were calculated relative to the localization results of CBCT, which were treated as the “standard” for reference. The scan angle shown beside each DTS technique label is either the total scan angle (for single-view DTS) or the halved total scan angle at each orthogonal direction (for orthogonal-view DTS). In each boxplot, the upper edge, the central line, and the lower edge of the box represent the 75 percentile (Q3), median, and 25 percentile (Q1) of the data, respectively. The lower whisker extends to the datum no smaller than  Q1−1.5×(Q3−Q1), and the upper whisker extends to the datum no larger than  Q3+1.5×(Q3−Q1). The “+” in the plots are outliers outside the whiskers. The boxplots make no assumptions of the data distribution and are nonparametric. CBCT indicates cone-beam computed tomography; DTS, digital tomosynthesis.

Figure 13.

Comparison between onboard DTS and different RDTS sets generated by different techniques for the kV–MV study. RDTS indicates reference digital tomosynthesis.

Figure 14.

A, Tumor localization error boxplots of different DTS techniques with fixed scan directions but varying total scan angles ranging from 0° to 60° (0°, 5°, 10°, 20°, 40°, and 60°). Each scan angle has 5 localization results, corresponding to 5 onboard tumor deviation scenarios created: no shift, 5 mm shift along the longitudinal direction, 5 mm shift along the vertical direction, 5 mm shift along the lateral direction, and 5 mm shifts along each of the 3 directions. Thus in each boxplots there are 30 data points. B, Tumor localization errors of different DTS techniques with fixed total scan angles (20°) but varying scan directions (each column shows results from a different scan direction). The tumor position was deviated onboard with 5 mm shifts along each of the 3 directions. DTS indicates digital tomosynthesis.