Session Item

Radiotherapy of lung cancer often involves considerable irradiation doses to normal tissue because safety margins are needed to account for respiratory motion and other geometrical uncertainties. Treatment in deep inspiration breath-hold (DIBH) may reduce the motion and the needed margins. However, poor correlation between the external signal used to gate the DIBH treatment and the internal target position can limit the targeting accuracy and thus the benefit of the DIBH treatment. In this study, we develop a method for markerless lung tumor localization in cine MV images of IMRT treatment fields and demonstrate use of the method to determine the tumor position error during DIBH treatments. 

25 lung cancer patients were treated with IMRT in DIBH in 20-33 fractions with daily cone-beam CT (CBCT) guided setup based on the primary tumor. Cine MV images were recorded for all treatment fields at most fractions. Post-treatment, the shift of the tumor position in the MV images relative to the intended position from the CBCT setup procedure was found as described in the following and summarized in Fig 1. First the individual cine MV frames with IMRT segments were pieced together to form a composite MV image that covered the entire MV beam aperture (L1 in Fig 1). Here, the possible tumor motion during the DIBH field delivery was neglected. Next, an in-house developed ray tracing algorithm was used to generate digitally reconstructed radiographs (DRRs) based on the entire CBCT (DDR_total), a slab of the CBCT near the isocenter plane containing the tumor volume (DDR_tumor), and the entire CBCT except for the tumor containing slab (DDR_non-tumor) (L2). All DRRs were rescaled to have similar contrasts as the composite cine MV image (L3). DDR_non-tumor was then subtracted from the composite cine MV image to generate a tumor-enhanced cine MV image (L4) in which the tumor position was found by template-based segmentation (L6) using the projected tumor shape in DDR_tumor as template (L5). The tumor segmentation has been performed for one patient at the time of abstract deadline and is currently being extended to all 25 patients.

Figure 2A shows examples of the CBCT-based tumor template and 2B shows the tumor-enhanced cine MV image with the segmented tumor position. In this case, the internal tumor shift in the patient between setup CBCT and treatment delivery was 2.9 mm in the cranial direction. Figure 2C shows the tumor shift at all imaged fractions for this patient. The mean shift was 3.4 mm (SD = 2.1 mm).

A method for markerless lung tumor localization in cine MV images was developed and applied to determine the intrafraction tumor shift between setup imaging and treatment delivery in lung cancer DIBH IMRT treatments. The method is well-suited for DIBH treatments, where the optimally small residual lung tumor motion during IMRT field delivery allows generation of a composite cine MV image from the IMRT segments.