Advanced HyperSight™ imaging for patients with adaptive SBRT of prostate cancer: a longitudinal analysis of tissue demarcation

Abstract

Background: Cone-beam computed tomography (CBCT)-based adaptive radiotherapy (ART) at the Ethos^® linear accelerator (eLinac) allows for daily anatomical and dosimetric adjustments and relies on robust image quality. This study evaluated the longitudinal image quality of the novel HyperSight^TM-CBCT (hCBCT) compared to planning CT (pCT), using phantoms and data of prostate cancer patients undergoing adaptive stereotactic body radiotherapy (SBRT). Building on this, the longitudinal contour sharpness of the organs in fractional hCBCT and their usability for ART workflow across fractions was evaluated.

Methods: Between December 2023 and May 2024, 26 prostate cancer patients receiving ART at the eLinac with hCBCT technology were enrolled. Phantom studies assessed pCT and hCBCT image quality. Patient based analyses of all 156 imaging scans (pCT and each of the fractional hCBCT) analyzed, longitudinally examined firstly image quality, secondly contour sharpness of prostate, seminal vesicles, and rectal wall, and thirdly confidence to delineate the structures for the ART workflow. Time required for the ART based parameters were recorded. Quantitative metrics included CT number changes in the fat adjacent to the prostate and seminal vesicles. Friedman’s test with Bonferroni correction, Spearman and Intraclass Correlation Coefficient (ICC) were used for statistics.

Results: hCBCT scans showed robust image quality parameters in the phantom and patient based analysis across fractions. Inter-observer agreement was moderate, with lower rating score for resident compared to the experienced radiation oncologist (p < 0.001). Patient based analysis showed no significant differences of the contour sharpness of the prostate and seminal vesicles between pCT and initial hCBCT scan, but contour sharpness ratings declined across treatment fractions. Confidence for the delineation of prostate and seminal vesicles during ART was significantly decreased at later fractions (each p_adj<0.05) and this correlated significantly with longer assessment times (p_adj≤0.05). The CT attenuation of the fat tissue adjacent to the prostate and seminal vesicles significantly increased across the fractions (p_adj<0.05).

Conclusions: High-quality imaging for adaptive SBRT in prostate cancer is provided by hCBCT, which offers equivalent tissue visualization compared to pCT. Fraction-dependent decreases in contour sharpness can be detected using longitudinal hCBCT imaging. These decreases are likely related to treatment-induced tissue changes and may impact ART workflow. The rating of the observed effects may potentially be influenced by the observer’s experience.

Keywords: Adaptive radiotherapy; Ethos; HyperSightTM-CBCT; Prostate cancer; Stereotactic radiotherapy.

Fig. 1

HyperSight™ phantom imaging. The imaging protocol used two test objects: (A) an inhomogeneous IMRT 3D Pelvis Phantom (Sun Nuclear) and (B) an Electron Density/Dual Energy Phantom (GAMMEX). All scans were acquired using both conventional planning computed tomography (CT) (Brilliance BigBore, Philips) with and without metal artefact reduction (MAR) and the investigational HyperSight™ conebeam CT (CBCT) system on an Ethos^® linear accelerator (Varian, Siemens Healthineers), with pelvis protocol and iCBCT, and MAR reconstruction. Quantitative image analysis focused on specific regions of interest (ROIs) within the electron density phantom: ROI 1 (yellow) assessed general tissue parameters for contrast and contrast-to-noise ratio calculations, while ROI 2 (red) evaluated liver-equivalent tissue characteristics including Signal-to-Noise-Ratio (SNR) and coefficient of variation. Enlarged views of key regions are highlighted in the green boxed area

Fig. 2

Comparison of Pelvic Imaging. Shown is the comparative analysis of pelvic imaging quality across treatment fractions. The planning computed tomography (pCT) scan (Brilliance BigBore, Philips) serves as the baseline for evaluating HyperSight™ CBCT scans on an Ethos^® linear accelerator (Varian, Siemens Healthineers) from the first (fx1) and fifth (fx5) fractions, with particular focus on the prostate-seminal vesicle complex and anterior rectal wall. Following key findings emerge: First, while pCT and fx1 CBCT show comparable contour sharpness, there is a noticeable decline in contour sharpness with fx5, particularly at the prostate-rectal interface. This trend is quantified by strategically placed regions of interest (ROIs) at tissue interfaces at the height of both target regions prostate and seminal vesicles (ROI1 = green, outer organ tissue/ROI2 = yellow, fat at the organ border/ROI3 = orange, peripheral fat) (see Results Sect. “Quantitative patient based analysis: longitudinal change of CT Attenuation in fractional hCBCT at the target interface”). Structural reference ROIs (turquoise/pink) in pelvic fat and gluteus maximus muscle are used to calculate the quantitative parameters of image quality (see Results Sect. “Quantitative patient based analysis: image quality parameters”)

Fig. 3

Mean Ratings of qualitative tissue differentiation parameters. Shown are Likert-scale ratings of computed tomography (pCT) scans (Brilliance BigBore, Philips) and fractional HyperSight™ CBCT scans on an Ethos^® linear accelerator (Varian, Siemens Healthineers) for adaptive SBRT for prostate cancer with different colours (parameters: pelvic tissue differentiation, contour sharpness of the prostate, rectal wall and seminal vesicles) for Observer R1 (A) and Observer R2 (B) with error bars representing standard deviations. Asterisks indicate statistical significance in the post-hoc comparison of qualitative imaging parameters (p_adj<0.05) between pCT and one of the five fractional HyperSight^TM-CBCT scans (fx1-5) during adaptive SBRT for prostate cancer

Fig. 4

Mean Ratings of qualitative imaging parameters for usability and confidence of delineation for adaptive treatment planning. Shown are Likert-scale ratings of computed tomography (pCT) scans (Brilliance BigBore, Philips) and fractional HyperSight™ CBCT scans on an Ethos^® linear accelerator (Varian, Siemens Healthineers) for adaptive SBRT for prostate cancer (parameters: usability for adaptive treatment planning, confidence of adaptive delineation of prostate/seminal vesicle target structures and the organs at risk) for Observer R1 (A) and Observer R2 (B) with error bars representing standard deviations. A black time curve is overlaid on each of the plots to indicate the mean duration of each of the rating assessments. Blue (rating) and black (time) asterisks indicate statistical significance in the post-hoc comparison of the rated parameters and the time consumption (p_adj<0.05), respectively, between pCT and one of the five fractional HyperSight^TM-CBCT scans (fx1-5) during adaptive SBRT for prostate cancer

Fig. 5

Quantitative Image Quality Parameters in Patient based Analysis of Imaging for adaptive SBRT for prostate cancer. Imaging includes treatment planning computed tomography (pCT) scan (Brilliance BigBore, Philips) with and without metal artefact reduction (MAR) and the five fractional HyperSight^TM-CBCT scans (fx1-5) with pelvic protocol with and without MAR reconstruction mode on the Ethos^® linear accelerator (Varian, Siemens Healthineers). The plots of the quantitative parameters Coefficient of Variation (CV), Signal-to-Noise-Ratio (SNR), contrast (C), contrast-to-noise-ratio (CNR) are given. Asterisks indicate statistical significance in the post-hoc comparison of the quantitative parameters (p_adj<0.05) between pCT and one of the five fractional HyperSight^TM-CBCT scans (fx1-5) during adaptive SBRT for prostate cancer. Each dot in the plot represents the mean value of the parameter values for the given imaging stage with the error bars denoting the interquartile range (IQR). The y axis displays the different parameters’ values (unit for C: Hounsfield Unit (HU); for CV: %)

Fig. 6

Quantitative parameters in patient based analysis of CT/CBCT-based Imaging for adaptive SBRT for prostate cancer. Imaging includes treatment planning CT (pCT) (Brilliance BigBore, Philips) with and without metal artefact reduction (MAR) and the five fractional HyperSight^TM-CBCT scans (fx1-5) with pelvis protocol with and without MAR reconstruction mode at the Ethos^® linear accelerator (Varian, Siemens Healthineers). Mean Hounsfield Units are being displayed on the y-axis. Boxplots comparing the distribution of measured values across three regions of interest (ROIs) at the target interface (ROI1, ROI2, ROI3, see Fig. 2) for each imaging time point (pCT and fx1–5) in two anatomical regions (posterior border of the prostate, top panels; posterior border of the seminal vesicles, bottom panels). Each box represents the interquartile range (IQR) with the median line, whiskers extend to 1.5×IQR, and outliers are shown as individual points. Pairs of ROIs with statistically significant differences (p_adj≤0.05) are marked with asterisks