Evaluation of plan quality in radiotherapy planning with an MR-linac

BACKGROUND & PURPOSE

Clinical introduction of magnetic resonance (MR)-guided radiotherapy involves treatment planning while taking into account machine-specific characteristics. Our aim was to investigate the feasibility of high-quality MR-linac treatment planning for an MR-linac and to benchmark MR-linac plan quality (IMRT) against current clinical practice (VMAT).

MATERIALS & METHODS

Data of eight rectal and eight prostate cancer patients, who received radiotherapy on a conventional CBCT-integrated linac, were selected. Clinically acquired CTs and associated delineations of target volumes and organs-at-risk (OARs) were used for MR-linac treatment planning in Monaco. To investigate treatment planning software bias 'quasi MR-linac plans' were generated in Pinnacle3 by mimicking MR-linac specific beam characteristics. MR-linac, quasi MR-linac, and clinical plans were compared and differences in target and OAR doses assessed. Differences in plan complexity were determined by the number of segments and monitor units.

RESULTS

Compared to clinical plans, MR-linac plans showed a statistically significant decrease in plan homogeneity, an increase in PTV Dmean (prostate: 0.6 Gy; rectum: 0.8 Gy) and D1% (prostate: 1.9 Gy; rectum: 2.0 Gy), and increases in OAR dose. Quasi MR-linac plans were comparable to MR-linac plans with respect to OAR dose and plan homogeneity. For rectal cancer an increase was seen in PTV Dmean (0.12 Gy) and D1% (0.5 Gy) compared to regular MR-linac plans. All created plans were clinically equivalent to current clinical practice.

CONCLUSIONS

This study demonstrates the feasibility of creating high-quality MR-linac treatment plans. The results supported the clinical introduction of an MR-linac.

Validation of a 4D-MRI guided liver stereotactic body radiation therapy strategy for implementation on the MR-linac

Purpose. Accurate tumor localization for image-guided liver stereotactic body radiation therapy (SBRT) is challenging due to respiratory motion and poor tumor visibility on conventional x-ray based images. Novel integrated MRI and radiotherapy systems enable direct in-room tumor visualization, potentially increasing treatment accuracy. As these systems currently do not provide a 4D image-guided radiotherapy strategy, we developed a 4D-MRI guided liver SBRT workflow and validated all steps for implementation on the Unity MR-linac.Materials and Methods. The proposed workflow consists of five steps: (1) acquisition of a daily 4D-MRI scan, (2) 4D-MRI to mid-position planning-CT rigid tumor registration, (3) calculation of daily tumor midP misalignment, (4) plan adaptation using adapt-to-position (ATP) with segment-weights optimization and (5) adapted plan delivery. The workflow was first validated in a motion phantom, performing regular motion at different baselines (±5 to ±10 mm) and patient-derived respiratory signals with varying degrees of irregularity. 4D-MRI derived respiratory signals and 4D-MRI to planning CT registrations were compared to the phantom input, and gamma and dose-area-histogram analyses were performed on the delivered dose distributions on film. Additionally, 4D-MRI to CT registration performance was evaluated in patient images using the full-circle method (transitivity analysis). Plan adaption was further analyzedin-silicoby creating adapted treatment plans for 15 patients with oligometastatic liver disease.Results. Phantom trajectories could be reliably extracted from 4D-MRI scans and 4D-MRI to CT registration showed submillimeter accuracy. The DAH-analysis demonstrated excellent coverage of the dose evaluation structures GTV and GTV_TD. The median daily rigid 4D-MRI to midP-CT registration precision in patient images was <2 mm. The ATP strategy restored the target dose without increased exposure to the OARs and plan quality was independent from 3D shift distance in the range of 1-26 mm.Conclusions. The proposed 4D-MRI guided strategy showed excellent performance in all workflow tests in preparation of the clinical introduction on the Unity MR-linac.

Evaluation of Deep Learning Clinical Target Volumes Auto-Contouring for Magnetic Resonance Imaging-Guided Online Adaptive Treatment of Rectal Cancer

Purpose

Segmentation of clinical target volumes (CTV) on medical images can be time-consuming and is prone to interobserver variation (IOV). This is a problem for online adaptive radiation therapy, where CTV segmentation must be performed every treatment fraction, leading to longer treatment times and logistic challenges. Deep learning (DL)-based auto-contouring has the potential to speed up CTV contouring, but its current clinical use is limited. One reason for this is that it can be time-consuming to verify the accuracy of CTV contours produced using auto-contouring, and there is a risk of bias being introduced. To be accepted by clinicians, auto-contouring must be trustworthy. Therefore, there is a need for a comprehensive commissioning framework when introducing DL-based auto-contouring in clinical practice. We present such a framework and apply it to an in-house developed DL model for auto-contouring of the CTV in rectal cancer patients treated with MRI-guided online adaptive radiation therapy.

Methods and Materials

The framework for evaluating DL-based auto-contouring consisted of 3 steps: (1) Quantitative evaluation of the model's performance and comparison with IOV; (2) Expert observations and corrections; and (3) Evaluation of the impact on expected volumetric target coverage. These steps were performed on independent data sets. The framework was applied to an in-house trained nnU-Net model, using the data of 44 rectal cancer patients treated at our institution.

Results

The framework established that the model's performance after expert corrections was comparable to IOV, and although the model introduced a bias, this had no relevant impact on clinical practice. Additionally, we found a substantial time gain without reducing quality as determined by volumetric target coverage.

Conclusions

Our framework provides a comprehensive evaluation of the performance and clinical usability of target auto-contouring models. Based on the results, we conclude that the model is eligible for clinical use.

Incorporating patient-specific prior clinical knowledge to improve clinical target volume auto-segmentation generalisability for online adaptive radiotherapy of rectal cancer: A multicenter validation

BACKGROUND & PURPOSE

Deep learning (DL) based auto-segmentation has shown to be beneficial for online adaptive radiotherapy (OART). However, auto-segmentation of clinical target volumes (CTV) is complex, as clinical interpretations are crucial in their definition. The resulting variation between clinicians and institutes hampers the generalizability of DL networks. In OART the CTV is delineated during treatment preparation which makes the clinician intent explicitly available during treatment. We studied whether multicenter generalisability improves when using this prior clinical knowledge, the pre-treatment delineation, as a patient-specific prior for DL models for online auto-segmentation of the mesorectal CTV.

MATERIAL & METHODS

We included intermediate risk or locally advanced rectal cancer patients from three centers. Patient-specific weight maps were created by combining the patient-specific CTV delineation on the pre-treatment scan with population-based variation of likely inter-fraction mesorectal CTV deformations. We trained two models to auto-segment the mesorectal CTV on in-house data, one with (MRI + prior) and one without (MRI-only) priors. Both models were applied to two external datasets. An external baseline model was trained without priors from scratch for one external center. Performance was evaluated on the DSC, surface Dice, 95HD and MSD.

RESULTS

For both external centers, the MRI + prior model outperformed the MRI-only model significantly on the segmentation metrics (p-values < 0.01). There was no significant difference between the external baseline model and the MRI + prior model.

CONCLUSION

Adding patient-specific weight maps makes the CTV segmentation model more robust to institutional preferences. Performance was comparable to a model trained locally from scratch. This makes this approach suitable for generalization to multiple centers.