The edge visualization metric: Quantifying the improvement of lung SBRT target definition with 4D CBCT

Purpose: Four-dimensional cone-beam CT (4D CBCT) incorporates oversampling of 3D data to reconstruct multi-phase CBCT data sets representing distinct phases of the breathing cycle based on a diaphragmatic correlate of respiratory motion. Motion artifacts and blurring can be reduced relative to three-dimensional cone-beam (3D CBCT), allowing clinicians to better visualize motion of targets. To quantitatively understand the degree to which target visualization is improved by 4D CBCT, an edge visualization metric (EVM) has been developed to describe the change in voxel intensities at the edge of targets in 4D CBCT maximum intensity projection images relative to 3D CBCT images. Methods: The EVM describes the median distance where voxel intensities drop from 80% to 20% of target voxel values. The EVM was evaluated in a phantom study with a CIRS dynamic thorax phantom and with eleven on-treatment lung SBRT patients. Results: In the phantom study, the EVM was improved for 4D CBCT relative to 3D CBCT for one-cm targets (2.43 ± 0.22 mm vs. 2.67 ± 0.31 mm, p = 0.04) and for 2-cm targets (2.60 ± 0.35 mm vs. 3.46 ± 1.03 mm, p = 0.02). In patients, the EVM was 3.59 ± 1.01 mm vs. 4.25 ± 1.24 mm (p < 0.05). Conclusions: When evaluating an imaging acquisition's degree of motion blurring and ability to delineate target edges, EVM may provide a less biased way to evaluate edge detection in the presence of motion when compared to traditional methods.