Assessing Dosimetric Benefits of Cone Beam Computed Tomography-Guided Online Adaptive Radiation Treatment Frequencies for Lung Cancer

Abstract

Purpose: Online adaptive radiation therapy (oART) has shown the ability to diminish interfraction variations. However, oART is a time- and labor-intensive process, and the optimal adaptation frequency remains to be determined for lung cancer oART. The purpose of this study was to quantify and assess dosimetric benefits associated with various adaptive frequencies in patients with lung cancer receiving oART.

Methods and materials: This study included 8 patients with lung cancer receiving oART treated on the Ethos platform in 30 or 33 fractions (n = 7 /1). For a total of 243 fractions, daily contours on cone-beam computed tomography (CT) and adaptive/nonadaptive plans on synthetic CT scan were used to simulate 4 different adaptation frequencies: none, single, weekly, and daily adaptation, resulting in 972 unique dose distributions. Dose-volume-histograms of targets and organs-at-risk (OARs) were compared between adaptation frequencies. Besides Dose-volume-histogram analysis, 3 radiation oncologists reviewed and scored 185 total plans, evenly sampling plans from the various adaptive frequencies. A comprehensive plan scorecard was fine-tuned to correlate with physician reviews and subsequently used for interplan comparison.

Results: Compared with no adaptation, daily adaptation improved the median clinical target volume V100% by 0.2% (IQR, 0.0-1.0) and the planning target volume D98% by 0.5% (IQR, -2.2 to 3.83). It also reduced the planning target volume D0.03cc by 2.1% (IQR, 0.7-3.2), the lungs-internal target volume V20 Gy by 2.5% (IQR, 1.0-4.5), the heart D_mean by 0.9 Gy (IQR, 0.4-2.6), and the esophagus D_mean by 1.6 Gy (IQR, 0.3-4.3). Single and weekly adaptation presented fewer benefits in OAR sparing and led to target undercoverage compared with daily adaptation. The PlanScoreCard effectively quantified plan quality, showing a positive monotonic correlation to physician scores (R = 0.57-0.87). It revealed that daily adaptation significantly improved total plan quality for 7 out of 8 patients, with improvements exceeding 5% of the plan score compared with no adaptation. In contrast, weekly and single adaptations led to improvements in only 2 and 1 patients, respectively.

Conclusions: Online kilovoltage cone-beam CT scan-guided daily adaptation may lead to dosimetric benefits in both target coverage and OAR sparing in patients with lung cancer. Other adaptation frequencies are effective for some patients but tend to lead to target undercoverage compared with daily adaptation.

Figure 1

(a) The study design. (b) The workflow of online adaptive radiation therapy (oART). Patients received an initial plan based on a simulation computed tomography (CT) scan and then received a cone-beam CT (CBCT) scan during each oART session, which was autocontoured and edited as needed. The daily synthetic CT scan was generated using a deformable image registration (DIR) of the simulation CT scan to the daily CBCT scan. Ethos automatically generates standard of care (SOC) plans with no adaptation (recalculating the initial plan onto daily anatomy) and daily adaptive plans (reoptimization given daily CBCT anatomy). Additionally, 1-time adaptation was simulated by optimizing at fraction 21, then recalculating the resulting adaptive plan on all remaining fractions, and weekly adaptation was simulated by optimizing at fractions 1, 6, 11, and so on, and recalculating these resulting plans on daily CBCT scans for the remainder of each week. Dose-washes for fraction 30, scaled to full course dose, of the patient receiving the biggest increase in mean score with daily adaptation (patient 8) are presented, showing a visible reduction in prescription dose and greater target homogeneity index as adaptation frequency increases. (c) Illustration of 4 different adaptation frequencies involved in this study.

Figure 2

Selected scorecard metric examples for a patient.

Figure 3

Comparison of dose-volume histogram (DVH) metrics versus adaptation frequency for 243 fractions analyzed in this work, scaled to full-course dose. Only the results of the 2-sided Wilcoxon paired nonparametric test with Bonferroni corrections comparing the daily-adaptation frequency with other frequencies were displayed, with significance values stratified as follows: ns: P > 0.006; *: .001 < P ≤ 0006; **: .0001 < P ≤ .001; ***: P ≤ .0001.

Figure 4

(a) The relationship between plan score and physician Likert ratings for patients 2, 3, 5, 7, and 8. Data and Spearman's correlation (R) are displayed. The vertical and horizontal lines represent the standard deviation of plan scores and physician Likert ratings for the different adaptation frequencies applied. (b) Mean target (blue) and organs-at-risk (red) comprehensive plan scores for each patient and adaptation frequency. Solid vertical lines and dashed horizontal lines illustrate the score standard deviation and initial plan score, respectively. Total mean scores are annotated for each adaptation frequency. (c) The scores trend as a function of treatment fractions and the adaptation strategy employed.

Figure 5

(a) The mean dice similarity coefficient (DSC) value, 95th percentile Hausdorff distance (HD95) value in centimeters (cm), and volume ratio value of daily online planning target volumes (PTVs) compared with simulation PTV over the treatment course. (b) The color map of daily variations in PTV surface for 8 patients at fractions 1, 15, and 30, compared with the PTV on the simulation computed tomography (CT) scan. Blue represents a surface contraction in the daily PTV segmentation, while red indicates a surface expansion. The shade of color quantitatively provides the degree of variations.