Periodic intrafraction image-guided focal dose-escalated stereotactic body radiotherapy for prostate carcinoma on a standard linear accelerator

Graphical abstract

Highlights

• •Linac-based focal dose-escalated stereotactic body radiotherapy for prostate cancer.
• •Treatment delivery was complemented with periodic intrafraction image guidance.
• •Periodic image guidance improved targeting accuracy and delivered dose fidelity.
• •Periodic image guidance reduced the positional (root mean square) error to <1 mm.
• •Target volumes D_98% remained withing 1% of the plan with periodic image guidance.

Abstract

Background and purpose

Intrafraction motion degrades the efficacy of radiation therapy. We reported on the implementation of focal dose-escalated stereotactic body radiation therapy (HypoFocal-SBRT) for prostate carcinoma on standard linear accelerator, and assessed whether periodic intrafraction image guidance (PIF-IG) improved the accuracy of treatment.

Materials and methods

Twenty prostate cancer patients underwent HypoFocal-SBRT. Treatment planning prioritized plan robustness and efficiency accommodating image guidance as-needed for treatment on standard accelerator. A three-dose level planning target volume (PTV) SBRT regime with focal boost was prescribed: PTV30 (seminal vesicles), PTV35 (prostate), and PTV40-42 (intraprostatic tumor lesion) in 5 fractions. Fiducial marker-based PIF-IG, triggered at pre-defined imaging control point(s) (ICPs), was performed using 1 mm (translations)/1° (rotations) action thresholds for online couch corrections. Applied marker registration values were compared to estimated ground-truth, as well as simulated treatment with no PIF-IG and effect on delivered dose estimated.

Results

Interventions occurred at least once in 95% of patients and 83% of fractions. PIF-IG reduced the geometric uncertainty (root-mean-square error) from 1.6 mm, 2.3 mm, and 2.2 mm for the left–right (LR), superior-inferior (SI), and anterior-posterior (AP) directions to 0.6 mm, 0.9 mm, and 0.8 mm, respectively. With PIF-IG, target volumes D_98% doses with 5% less than planned were observed in 2% of fractions for PTV30 and PTV35, and 0% for PTV40-42, compared to 16%, 12%, and 6% without PIF-IG.

Conclusions

Focal dose-escalated SBRT for prostate cancer can be efficiently and accurately performed on standard accelerator. PIF-IG improved the geometric and delivered dose accuracy.

1. Introduction

Radiation therapy (RT) is a curative treatment option for patients with localized prostate cancer [1]. Advances in image guidance and treatment planning have facilitated the widespread of hypofractionated RT over the past decade [2]. Moderate hypofractioned RT trials like the CHHIP [3] and related studies [4] have demonstrated the non-inferiority of 60–62 Gy in 20 fractions over conventional fractionation. Further hypofractionation through stereotactic body radiotherapy (SBRT) have been investigated in the Phase III Hypo-RT [5] and PACE-B [6] trials, which reported non-inferiority of SBRT compared to conventional- or moderate-hypofractionation. The low alpha–beta ratio of prostate cancer, compared to adjacent organs of interest, enables favorable outcomes with extreme hypofractionation due to its increased sensitivity to high radiation doses per fraction [7]. However, recent results suggest SBRT can lead to temporary increased genitourinary side effects, potentially explained by a lower-than-expected urethra’s alpha–beta ratio [8].

In addition to hypofractionation, the concept of focal dose escalation has emerged, given that local recurrences often originate within dominant intraprostatic lesion(s) (IPLs), considered drivers of subsequent nodal and distant metastases. The randomized controlled phase III FLAME trial [9], [10], [11] demonstrated improved biochemical recurrence-free-survival with focal doses escalation based on magnet resonance imaging (MRI). The HypoFocal trials are currently investigating the tolerability and effectiveness of focal boosting based on MRI and prostate membrane specific antigen positron emission tomography/computed tomography (PSMA-PET/CT) to identify IPLs. The HypoFocal Phase II trial demonstrated acceptable acute and chronic side effects with improved quality-of-life, despite substantial boost-volumes [12]. The ongoing HypoFocal-SBRT trial investigates the efficacy of focal doses escalation with SBRT [13].

SBRT with focal boosting requires careful planning and high level of accuracy to ensure adequate boost target volume(s) coverage and steep dose fall-off, outside boost region, to spare healthy tissue. Despite technological improvements in image guidance, side effects are inherent and remain [6], which can be attributed by a host of factors like radiosensitivity and dose delivered to organs of interest [14], [15]. An important source of uncertainty throughout preparation and execution workflow of RT is organ motion. Both inter- and intra-fraction motion can significantly contribute to the overall uncertainty in dose delivery [16], [17], [18], [19]. While interfraction uncertainties are accounted for with routine pre-treatment imaging, intrafraction motion control protocols are uncommon. To date, dedicated robotic- [6], [20] and MRI-based [15], [16], [17], [18], [19] platforms enable on-line inter-/ intra-fraction geometric and/or volumetric adaptation, although these are often expensive and not off-the-shelf. The MIRAGE trial demonstrated reduced side effects resulting from reduced planning target margins with motion compensation [21]. However, geometric/volumetric adaptation workflows are associated with increased quality assurance requirements, labor-intensive tasks, prolonged treatment times, and reduced patient throughput. Moreover, patient discomfort and risk of intrafraction variability augment with time. Therefore, straight-forward workflows are needed. Since the vast bulk of platforms are C-arm based, easy-to-implement workflows will not only improve treatment efficiency and accuracy, but also enable safe and widespread of SBRT for prostate cancer. We developed a workflow for prostate cancer SBRT combining enhanced treatment planning and flattening-filter-free (FFF) volumetric arc therapy (VMAT) delivery with periodic intrafraction image guidance (PIF-IG). In this work, we assessed the efficacy of the workflow and whether PIF-IG improved the accuracy of treatment.

2. Materials and methods

2.1. Patients

Twenty prostate cancer patients treated between January 2023 and December 2024 were included in the study approved by the human research ethics committee (No. EK-21-1183). All patients provided written informed consent. Table 1 summarizes patient characteristics, including median age (74 years, 64–81), median prostate volume (44.9 cm³, 22.4–65.0), prostate-specific antigen (PSA) levels, and risk status.

Table 1.

Patients/fractions characteristics and intra-treatment image guidance information.

Patients/treatments characteristics Patients [n = 20]/Fractions [n = 100]	No.
Age, y (at treatment)
Median	74
Range	64–81
PSA level prior to treatment
<10 ng/ml	14 (70%)
10–20 ng/ml	4 (20%)
<20 ng/ml	2 (10%)
Prostate volume, cm3
Median	44.9
Range	22.4–65
Risk status
Intermediate favorable-risk	14 (70%)
High-risk	6 (30%)
Stereoscopic exposures
Imaging control points (ICP)	300
Verification images	171
Interventions per fraction
At least once	83
More than once	36
Marker motion at ICP (with PIF-IG)
>±1 mm [LR/SI/AP]	12.2/23.5/0.0%
>±2 mm [LR/SI/AP]	1.2/4.1/0.0%
Marker motion at ICP (No PIF-IG)
>±1 mm [LR/SI/AP]	47.9/70.9/55.6%
>±2 mm [LR/SI/AP]	20.9/35.7/28.1%

2.2. Simulation

All patients underwent a 3 Tesla MRI and a PSMA-PET prior to treatment planning. Patients had three gold anchor markers (Naslund Medical AB, Sweden) implanted 5–7 days before simulation. Planning CT was acquired on a Philips Big Bore CT scanner (Forchheim, Erlangen, Germany) with 1 mm slice thickness. Patients were instructed to drink 500–750 ml of water and use Klysma 60–90 min before CT acquisition. Patients were immobilized using a vacuum bag. When possible, MRT and CT were performed on same day. A detailed description of target volumes and organs of interest definition can be found in the Supplementary Material A.

2.3. Treatment planning

Treatment planning process aimed to maximize delivery efficiency and plan robustness, minimizing the on-couch time and probability of intrafraction uncertainty. Prescribed doses were 30 Gy (PTV30) to seminal vesicles, 35 Gy (PTV35) to prostate, and 40–42 Gy (PTV40-42) to ITL(s) in five fractions. Plans were generated for a TrueBeam^TM linear accelerator (linac) (Varian Medical System, Palo Alto, CA, USA) equipped with millennium multileaf-collimator (MLC), PerfectPitch 6-DoF couch, and on-board kilovoltage (kV)/megavoltage (MV) imaging. A surface scanner, Catalyst HD (C-Rad Positioning AB, Uppsala, Sweden) as well as stereoscopic X-ray system, ExacTrac® (BrainLAB AG, Munich, Germany), are also available. Eclipse treatment planning system v15.6 with RapidArc VMAT (Varian Medical System, Palo Alto, USA) was used. Double-arc VMAT plans were generated with ±5° collimator angle, 10 MV-FFF, 2400 monitor (MU)/min dose rate using the Analytical Anisotropic Algorithm (AAA) dose calculation algorithm (until June 2024) and Acuros (from July 2024) with a 1.25 mm calculation grid size. Plan complexity reduction included MU and field aperture constraints. Preliminary results demonstrated efficiency gains of double-arc VMAT and 10 MV-FFF beams [22]. Given that each arc is delivered over ∼1 min, we opted to complement delivery with imaging control points (ICPs) at discrete progress points (section 2.4) for intrafraction motion control. This helped to define, execute and monitor delivery workflow as a series of discrete steps.

2.4. Treatment delivery workflow

Treatment delivery integrated surface and X-ray imaging across three stages: pre-, intra-, and post-treatment (Fig. 1). Pre-treatment involved patient/marker alignment; intra-treatment included dose delivery complemented with PIF-IG through stereoscopic X-ray imaging. Post-treatment served as accuracy assurance.

Fig. 1.

Surface-guided (SG) and stereoscopic X-ray image-guided treatment workflow. Pre-treatment: patient/marker alignment. Intra-treatment: dose delivery assisted with periodic intrafraction image guidance (PIF-IG). Post-treatment: accuracy assurance. Following patient setup, maker alignment is verified through stereoscopic imaging control points (ICP1-3) performed with ExacTrac. A couch correction action is triggered upon detected deviations >1 mm/1°. Each arc is delivered within ∼1 min. During beam-on-time, treatment is monitored through surface guidance (5 mm threshold).

Patients were aligned to in-room lasers, followed by surface imaging. Then, marker-based alignment was performed by stereoscopic X-ray imaging using ExacTrac. Of note, although our TrueBeam platform facilitates orthogonal imaging (e.g. kV-MV image pair), we opted to use stereoscopic imaging in complies with in-house multi-modality assurance protocol. Marker positions in stereoscopic projections were registered to their calculated in reference DRRs, and resultant misalignments processed into couch corrections by ExacTrac; if automatic registration failed, the markers’ pattern was manually registered in 3D (translations only) and deviations corrected. A partial-arc cone-beam computed tomography (CBCT) scan acquisition proceeded for anatomy verification and marker alignment crosscheck. Patients with unsuitable rectal and/or bladder filling were rescheduled. Efforts were made to minimize delays from patient waiting, pre-treatment imaging to start of delivery.

Following patient setup, stereoscopic images were subsequently acquired at three pre-set ICPs (Fig. 1): immediately before Arc1 (ICP1); prior (ICP2) and after (ICP3) Arc2. Marker misalignments >1 mm (translational) and 1° (rotational) were transmitted to the TrueBeam console to trigger online couch corrections; or ignored otherwise. A verification image followed after each correction. This helps discern possible high-frequency, transient- and persistent excursions motion types observed in prostate tumors [23]. In-house ‘cMotion’ surface guidance protocol (5 mm tolerance) was applied to monitor possible gross errors (e.g. movement due to discomfort). As assurance measure, stereoscopic and CBCT acquisitions were repeated post-treatment.

2.5. Analysis

The calculated MUs and beam-on-time (BOT) for each patient were recorded. Treatment efficiency was quantified for the three stages of the execution workflow: i) pre-treatment—time until completion of patient setup, ii) intra-treatment—beam delivery with PIF-IG, and iii) post-treatment—final imaging and accuracy assurance. Total session time spanned from first pre-treatment surface scan to last post-treatment CBCT.

The intrafraction target (markers’ pattern) alignment uncertainty was retrospectively evaluated using stereoscopic images and registration values extracted from ExacTrac. This analysis included data acquired following pre-treatment imaging/setup only. Three-dimensional (3D) marker displacements (translations only) were analysed in the left–right (LR), superior-inferior (SI), and anterior-posterior (AP) directions at isocentre.

To quantify the improvement of PIF-IG in targeting accuracy, applied (actual) registration values were compared to ground-truth (ideal) registration values. Recorded markers’ displacement, $d^{(clin.)}$, in the $\left(i\right)$ direction, corresponding to the LR, SI, or AP, were extracted from ExacTrac. To estimate the ground-truth marker displacement, $d^{(man)}\left(i\right)$, registration was performed offline by manually aligning the projected marker pattern over their (expected) position visible on DRRs. The positional uncertainty, ${\mathit{Err}}_{\mathit{PIF}-IG}$, was defined by point-wise difference between the actual and ground-truth displacements using:

$${\mathit{Err}}_{\mathit{PIF}-IG}(i)=\left(d^{(clin)}\left(i\right)-\langle d^{\left(man\right)}\rangle \left(i\right)\right)$$ (1)

A scenario representing delivery without image guidance was also simulated by omitting couch corrections. The root-mean-square error (RMSE) of absolute position differences was used as a measure of the geometric uncertainty of treatment with and without PIF-IG.

Delivered doses were estimated by means of isocenter shifts based on marker displacement determined at each ICP. Similarly to [24], [25], motion-encoded plans were generated, but at lower (sixty-second) temporal resolution. Delivered doses were reconstructed on a per-fraction basis for two scenarios: i) treatment with PIF-IG (default), and ii) without (simulated). Dose reconstructions were performed on the planning CT, which implies that inter-/intra-fraction volume variations and deformation of target and organs nearby are ignored. However, as stated in [24], had the pre-treatment CBCT scan been used, possible motion occurring following setup would mean that the CBCT was still not representative. Dose-volume indices estimated included D_50% and D_98% for the PTV, D_99% for the CTV; for the rectum, D_0.1cc, D_1cc, D_2cc, and D_20%; bladder, D_0.1cc, D_5cc, D_15%, D_20% and D_50%; urethra the D_0.1cc and D_50%. As suggested by [26], we considered a 5% dose difference between the planned and delivered dose as being clinically relevant.

3. Results

The median MUs was 2052 [1899–2376], and the median BOT was 99.4 s [89.0–109.8]. Table 1 summarizes intrafraction imaging, interventions, and marker displacements. Overall, 471 stereoscopic exposures were recorded across 100 fractions (3 ICPs/fraction), including 171 verification exposures after couch corrections. Intrafraction interventions were performed in at least 1 fraction for 95% (19/20) of patients and 83% (83/100) of fractions. In 36% (36/100) fractions, intra-treatment couch shifts were performed more than one time. Across all patients, the mean treatment session time was 14.6 min. Improvements in treatment efficiency over the two-year period were recorded, with mean session times of 16.5 min and 12.6 min for treatments performed in 2023 and 2024, respectively (Fig. 2). Pre-treatment imaging was the longest at 7.4 min on average, while intra-treatment phase accounted for 4.8 min (32.9%) of the whole session time. Generally, PIF-IG increased treatment times by 35 s to 1.5 min. Percentage of time prostate motion > ±1 mm/2 mm were 12.2/1.2% (LR), 23.5/4.1% (SI), and 0.0/0.0% (AP), respectively, compared to 47.9/20.9% (LR), 70.9/35.7% (SI), and 55.6/28.1% (AP) with no PIF-IG.

Fig. 2.

Pie chart illustration of the measured efficiency across pre-, intra-, and post-treatment surface-guided (SG) and X-ray image-guided workflow for both 2023 and 2024 cohort groups.

Overall, PIF-IG improved target alignment across all patients, fractions, and arcs compared with treatment without PIF-IG (Fig. 3). The 90th percentile of target (marker) alignment distribution for treatment with PIF-IG was 2.2 mm (LR), 2.6 mm (SI), and 2.3 mm (AP), whereas without PIF-IG was 5.5 mm (LR), 6.8 mm (SI), and 7.2 mm (AP), respectively. The RMSE with PIF-IG was 0.6 mm (LR), 0.9 mm (SI), and 0.8 mm (AP), whereas without PIF-IG the positional error would have been 1.6 mm (LR), 2.3 mm (SI), and 2.2 mm (AP), respectively. The skew in no PIF-IG distributions, more pronounced in the SI and AP directions, reflects the predominant displacement of prostate superiorly and posteriorly.

Fig. 3.

Summary over all patients and fractions indicating the distribution of target (markers) alignment error for treatment with (black) and without (red) periodic intrafraction image guidance (PIF-IG), where positive denotes right, inferior and anterior. The 90th percentile for each density distribution as well as the determined root-mean-square error (RMSE) are also shown. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

With PIF-IG, target volume coverage deficits and degradation in organs of interest indices relative to the planned doses were marginal compared with treatment without PIF-IG. Averaged across all patients and fractions, the estimated D_98% for treatment with PIF-IG, was 99.1% [86.7–100.8%], 99.1% [90.5–100.0], and 99.4% [97.8–100.9] for PTV30, PTV35, and PTV40-42, respectively, whereas for treatment without PIF-IG was 97.5% [77.7–101.2%], 96.8% [84.5–99.8%], and 98.4% [90.9–100.6%] for PTV30, PTV35, and PTV40-42. PIF-IG resulted in lower percentage differences in rectum, bladder, and urethra dose-volume indices compared with treatment without PIF-IG (Supplementary Table S1). Percentage of treatments with dose-volume index differences ±2 and 5% of planned doses are also summarized. With PIF-IG the percentage of fractions with D_98% dose ±2% than of the planned dose was 18%, 10%, and 4% for PTV30, PTV35 and PTV40-42, respectively. Without PIF-IG, the number of fractions with D_98% dose ±2% than the planned would have been 34% (PTV30), 52% (PTV35), and 24% (PTV40-42), respectively. Clinically relevant (±5%) D_98% dose difference were observed in 16% (PTV30), 12% (PTV35), and 6% (PTV40-42) of the treatments without PIF-IG, compared to only 2% (PTV30), 0% (PTV35), and 0% (PTV40-42) with PIF-IG.

As for the rectum D_0.1cc, D_1cc, D_2cc and D_20%, as well as bladder D_0.1cc, D_5cc, D_15%, D_20%, and D_50% indices the percentage of fractions with 5% dose more/less than planned was 0, 0, 0, and 20% (rectum) and 0, 2, 18, 26 and 24% with PIF-IG. Conversely, without PIF-IG, these volume indices would have been 4, 4, 0, and 44% (rectum), and 0, 14, 68, 74, and 62% (bladder), respectively. Minor deviations in urethra D_0.1cc was observed, with a mean percentage difference of 0% [-1.3–1.2%] for treatment with PIF-IG compared to −1% [−0.4–2.4%] without PIF-IG.

4. Discussion

This study presented an efficient, easy-to-implement workflow for prostate cancer SBRT on standard linac. Our approach aimed to enhance plan robustness and delivery efficiency complementing treatment with PIF-IG. Retrospectively, we showed that PIF-IG improved targeting accuracy and delivered dose fidelity compared with treatments that would have been delivered without any intrafraction image guidance.

The probability of intrafraction prostate motion increases with time [27]. The median BOT was 1.7 min. Of note, due to safety measures the gantry maximal rotation speed is restricted to 1 rotation/min by most C-arm accelerator vendors. The mean treatment session time was 14.6 min, which constitutes an upper limit given our multi-modality/–vendor setup. Nevertheless, it is noteworthy that the intra-treatment phase, comprising BOT with PIF-IG, averaged 4.8 min only. Previous studies have reported efficiency gains of advanced planning using single- [28], double-arc [29] VMAT and FFF-beam for prostate SBRT. Recently, Mastella et al. [28] compared a single-arc strategy with the conventional double-arc and reported reductions in mean BOTs of 46.2% (1.6 min) and 56.2% (1.3 min) for 6 MV-FFF and 10 MV-FFF qualities, respectively. In the current trial, given the three-dose level prescription with focal doses tailored to IPL(s) and steep dose fall-off toward healthy tissue, a double-arc arrangement was used. This increased optimizer flexibility, minimizing plan complexity.

The percentage of time motion exceeding the 2 mm (GTV-union to PTV40-42) margin occurred rarely with PIF-IG (0.0% LR, 4.1% SI, and 0.0% AP) but more frequently without (24.9% LR, 35.7% SI, and 28.1% AP). Curtis et al. [30] suggested additional 2 mm margins to compensate intrafraction motion at sixty-second control imaging intervals. Considerable deviations (>5 mm) were discerned despite rigorous rectum/bladder protocols and idealized (∼1 rotation/min) BOTs, highlighting the need for more frequent or continuous monitoring. The increased RMSE in clinical registration can be attributed to workflow- and system-specific factors. For instance, efficiency demands (e.g. on-couch time), anatomy changes (e.g. passive bladder filling), and system accuracy (e.g. marker detection errors) act at treatment, increasing workload and reliance on operator. Often trade-offs are required, resulting in increased (residual) errors. Conversely, manual registration was performed off-line, minimizing strategy-/operator-related errors. Overall, PIF-IG improved the correspondence between planned and the delivered dose. Faccenda et al. [25] reported (mean) relative dose differences (PTV D_95%) of −2.4% [−11.9–0.9%] in reconstructed motion-inclusive plans and 8 Gy/fraction delivery with and without motion control on standard linac. In this study, mean relative differences (D_98%) was −0.9% [−9.5–0.9%], −0.9% [−2.2–0.9%], and −0.6% for PTV30, PTV35 and PTV40-42, respectively. The increased mean difference in [25] may be attributed to the high-temporal resolution of real-time motion tracking and 2 mm action threshold, in contrast to 1 mm (translational) used in this study. These results improve confidence and give more realistic hint on true clinical outcomes; it minimizes bias when assessing clinical failures and side effects from unaccounted intrafraction motion or marginal misses.

Most published data concerning prostate cancer SBRT and motion control are attributed to robotic-based approaches [6], [20], with MRI-guided [15], [18], [19], [21] and CT-guided [31] adaptive workflows becoming more widespread. Although robotic-based systems enable increased degrees-of-freedom targeting, major limitations include, increased quality assurance requirements, and prolonged treatment times. Similarly, on-line adaptive platforms enable superior soft-tissue contrast and (daily) volumetric changes adaptation, although the associated workflows are labor- and time-intensive to the point that intrafraction anatomy/positional changes become a concern [32]. Reduced dose rates (600–800 MU/min) constitute further drawbacks when considering ultra-hypofractioned regimens. While the technology and expertise continuous to mature, caution is recommended when reducing margins and applying high single doses. Electromagnetic-guided solution have also become available for standard linac [24], [25], some requiring temporary insertion of catheters [25]. Real-time kV [33] or kV-MV [34] intrafraction monitoring have been proposed. However, these are not commercially available. Ultrasound-based motion control has also become available [35]. Unlike marker- and electromagnetic-based approaches, ultrasound guidance provides volumetric information; however, its clinical use has been discontinued. Standard linacs are technically equipped to potentially match or even outperform robotic- and adaptive-based workflows, provided that structured planning and image guidance strategy are in place.

The presented methodology has limitations. Along those inherent to the use of implanted fiducials (e.g., marker migration, localization errors), target position information was available approx. every 1 min. This may be a source of bias and affect actual intrafraction motion and thus accumulated dose distribution estimates, leading to under- or overestimation of dose-volume indices. Therefore, these should be considered solely as estimates. Rotations and deformation, which can substantially affect dose distributions [32], [36], were not addressed. Unlike [35], we assumed rigid prostate and organs, despite well-documented daily volume changes [32], [35]. It is questionable whether geometric adaptation solely would suffice. Ideally, both geometric and volumetric adaptation, and in real-time, would provide the ultimate means towards prostate cancer management [37]. Extending our approach on to include volume adaptation and explore their joint benefit would be the next step.

In conclusion, enhanced planning and coupling of dose delivery with periodic intrafraction imaging enabled robust and efficient treatment delivery, as well as improved geometric and delivered dose accuracy of prostate cancer focal dose-escalated SBRT.

CRediT authorship contribution statement

Marco Serpa: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Rolf Wiehle: Methodology, Investigation. Thomas Rothe: Methodology, Investigation. Sophie Küchler: Methodology, Investigation. Tobias Fechter: Software, Investigation. Ilias Sachpazidis: Software, Investigation. Carlotta Rizzo: Methodology, Investigation. Sonja Adebahr: . Siegmar Kuhn: Investigation. Paolina Toncheva: Investigation. Dimos Baltas: Supervision, Investigation. Costantinos Zamboglou: Writing – original draft, Supervision, Methodology, Investigation, Formal analysis, Conceptualization. Anca-L. Grosu: Writing – original draft, Supervision, Resources, Project administration, Funding acquisition, Formal analysis, Conceptualization. Simon K.B. Spohn: Writing – review & editing, Writing – original draft, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The HypoFocal-SBRT trial (NCT06330909) is funded by the Federal Ministry of Research, Technology and Space (BMFTR) under the program “National Decade against Cancer”. No 01KD2201.

Appendix A. Supplementary data

The following are the Supplementary data to this article: