Comparison of Safety Margin Generation Concepts in Image Guided Radiotherapy to Account for Daily Head and Neck Pose Variations

Essential parameters needed for a U-Net-based segmentation of individual bones on planning CT images in the head and neck region using limited datasets for radiotherapy application

Objective.The field of radiotherapy is highly marked by the lack of datasets even with the availability of public datasets. Our study uses a very limited dataset to provide insights on essential parameters needed to automatically and accurately segment individual bones on planning CT images of head and neck cancer patients.Approach.The study was conducted using 30 planning CT images of real patients acquired from 5 different cohorts. 15 cases from 4 cohorts were randomly selected as training and validation datasets while the remaining were used as test datasets. Four experimental sets were formulated to explore parameters such as background patch reduction, class-dependent augmentation and incorporation of a weight map on the loss function.Main results.Our best experimental scenario resulted in a mean Dice score of 0.93 ± 0.06 for other bones (skull, mandible, scapulae, clavicles, humeri and hyoid), 0.93 ± 0.02 for ribs and 0.88 ± 0.03 for vertebrae on 7 test cases from the same cohorts as the training datasets. We compared our proposed solution approach to a retrained nnU-Net and obtained comparable results for vertebral bones while outperforming in the correct identification of the left and right instances of ribs, scapulae, humeri and clavicles. Furthermore, we evaluated the generalization capability of our proposed model on a new cohort and the mean Dice score yielded 0.96 ± 0.10 for other bones, 0.95 ± 0.07 for ribs and 0.81 ± 0.19 for vertebrae on 8 test cases.Significance.With these insights, we are challenging the utilization of an automatic and accurate bone segmentation tool into the clinical routine of radiotherapy despite the limited training datasets.

Detecting and Locating the Site of Local Relapse Using 18F-PSMA-1007 Imaging After Primary Treatment of 135 Prostate Cancer Patients-Potential Impact on PSMA-Guided Radiation Therapy

PURPOSE

Due to limited imaging options, the visualization of a local relapse of prostate cancer used to pose a considerable challenge. However, since the integration of ¹⁸F-PSMA-1007-PET/CT into the clinic, a relapsed tumor can now easily be detected by hybrid imaging. The present study aimed to evaluate and map the allocate relapse in a large cohort of prostate cancer patients focusing on individual patient management conclusions for radiation therapy.

PROCEDURES

The current study included 135 men with prostate cancer after primary treatment who underwent ¹⁸F-PSMA-1007-PET/CT due to biochemical relapse detecting a local relapse. Imaging data were reassessed and analyzed with regard to relapse locations. For the correlation of tumor foci with clinical data, we used binary logistic regression models as well as the Kruskal-Wallis test and Mann-Whitney test.

RESULTS

In total, 69.6% of all patients (mean age: 65 years) underwent prostatectomy while 30.4% underwent radiation therapy. PET imaging detected most frequently a unifocal relapse (72.6%). There was a statistically significantly higher rate of ipsilateral cases among the relapsed tumors. Comparing both treatment approaches, tumors relapsed most commonly within the posterior region after surgery and transition/peripheral zone after radiation therapy, respectively.

CONCLUSIONS

The present study confirms that ¹⁸F-PSMA-1007-PET/CT is highly suitable for the localization and allocation of a local relapse in patients with prostate cancer. The data enable further optimizing dose prescriptions and target volume delineations of radiation therapy in the future.

An online adaptive plan library approach for intensity modulated proton therapy for head and neck cancer

BACKGROUND AND PURPOSE

In intensity modulated proton therapy (IMPT), the impact of setup errors and anatomical changes is commonly mitigated by robust optimization with population-based setup robustness (SR) settings and offline replanning. In this study we propose and evaluate an alternative approach based on daily plan selection from patient-specific pre-treatment established plan libraries (PLs). Clinical implementation of the PL strategy would be rather straightforward compared to daily online re-planning.

MATERIALS AND METHODS

For 15 head-and-neck cancer patients, the planning CT was used to generate a PL with 5 plans, robustly optimized for increasing SR: 0, 1, 2, 3, 5 mm, and 3% range robustness. Repeat CTs (rCTs) and realistic setup and range uncertainty distributions were used for simulation of treatment courses for the PL approach, treatments with fixed SR (fSR₃) and a trigger-based offline adaptive schedule for 3 mm SR (fSR₃OfA). Daily plan selection in the PL approach was based only on recomputed dose to the CTV on the rCT.

RESULTS

Compared to using fSR₃ and fSR₃OfA, the risk of xerostomia grade ≥ II & III and dysphagia ≥ grade III were significantly reduced with the PL. For 6/15 patients the risk of xerostomia and/or dysphagia ≥ grade II could be reduced by > 2% by using PL. For the other patients, adherence to target coverage constraints was often improved. fSR₃OfA resulted in significantly improved coverage compared to PL for selected patients.

CONCLUSION

The proposed PL approach resulted in overall reduced NTCPs compared to fSR₃ and fSR₃OfA at limited cost in target coverage.

Application of an automatic, uncertainty model-guided, target-generating algorithm to lung stereotactic body radiotherapy

PURPOSE

This work evaluated a new radiotherapy target-generating framework (the αTarget algorithm) for creating internal target volumes for lung SBRT.

METHODS

Nineteen patients previously treated with definitive intent SBRT to the lung were identified from a clinical database. For each patient's 4DCT simulation scan, deformable image registration was used between phases of the scan in order to generate voxelized models of motion for 35 individual gross tumor volumes. These motion models were then used with a new implementation of a previously described target-generating algorithm to create new internal target volumes (αITVs). The resulting αITVs were analyzed with respect to their volume and the coverage they provided each tumor voxel per that voxel's motion model. The clinically used ITVs were similarly analyzed, and were then compared to the αITVs using paired Student's t-tests. In addition, isotropic margins were added to the αITVs in order to determine the largest margin magnitude that could be added without exceeding the volume of the clinical ITVs.

RESULTS

The αITVs increased the target coverage provided to each tumor's 5th-percentile-most-covered-voxel an average of 50.3% compared to the clinical ITVs (p < 0.0001). At the same time, the αITVs had volumes that were, on average, 31.4% smaller (p < 0.0001). The differences in volume were large enough that, on average, an extra 2 mm isotropic margin could be added to the αITV before it had a volume greater than the clinical ITV.

CONCLUSIONS

The αTarget algorithm can generate more effective lung SBRT internal target volumes that provide greater coverage with smaller volumes. In combination with numerous other advantages of the framework, this effectiveness makes the αTarget algorithm a powerful new method for advanced IGRT or adaptive radiotherapy techniques.

A review of Image Guided Radiation Therapy in head and neck cancer from 2009-201 - Best Practice Recommendations for RTTs in the Clinic

Radiation therapy (RT) is beneficial in Head and Neck Cancer (HNC) in both the definitive and adjuvant setting. Highly complex and conformal planning techniques are becoming standard practice in delivering increased doses in HNC. A sharp falloff in dose outside the high dose area is characteristic of highly complex techniques and geometric uncertainties must be minimised to prevent under dosage of the target volume and possible over dosage of surrounding critical structures. CTV-PTV margins are employed to account for geometric uncertainties such as set up errors and both interfraction and intrafraction motion. Robust immobilisation and Image Guided Radiation Therapy (IGRT) is also essential in this group of patients to minimise discrepancies in patient position during the treatment course. IGRT has evolved with increased 2-Dimensional (2D) and 3-Dimensional (3D) IGRT modalities available for geometric verification. 2D and 3D IGRT modalities are both beneficial in geometric verification while 3D imaging is a valuable tool in assessing volumetric changes that may have dosimetric consequences for this group of patients. IGRT if executed effectively and efficiently provides clinicians with confidence to reduce CTV-PTV margins thus limiting treatment related toxicities in patients. Accumulated exposure dose from IGRT vary considerably and may be incorporated into the treatment plan to avoid excess dose. However, there are considerable variations in the application of IGRT in RT practice. This paper aims to summarise the advances in IGRT in HNC treatment and provide clinics with recommendations for an IGRT strategy for HNC in the clinic.

Generating amorphous target margins in radiation therapy to promote maximal target coverage with minimal target size

BACKGROUND AND SIGNIFICANCE

This work provides proof-of-principle for two versions of a heuristic approach that automatically creates amorphous radiation therapy planning target volume (PTV) margins considering local effects of tumor shape and motion to ensure adequate voxel coverage with while striving to minimize PTV size. The resulting target thereby promotes disease control while minimizing the risk of normal tissue toxicity.

METHODS

This work describes the mixed-PDF algorithm and the independent-PDF algorithm which generate amorphous margins around a radiation therapy target by incorporating user-defined models of target motion. Both algorithms were applied to example targets - one circular and one "cashew-shaped." Target motion was modeled by four probability density functions applied to the target quadrants. The spatially variant motion model illustrates the application of the algorithms even with tissue deformation. Performance of the margins was evaluated in silico with respect to voxelized target coverage and PTV size, and was compared to conventional techniques: a threshold-based probabilistic technique and an (an)isotropic expansion technique. To demonstrate the algorithm's clinical utility, a lung cancer patient was analyzed retrospectively. For this case, 4D CT measurements were combined with setup uncertainty to compare the PTV from the mixed-PDF algorithm with a PTV equivalent to the one used clinically.

RESULTS

For both targets, the mixed-PDF algorithm performed best, followed by the independent-PDF algorithm, the threshold algorithm, and lastly, the (an)isotropic algorithm. Superior coverage was always achieved by the amorphous margin algorithms for a given PTV size. Alternatively, the margin required for a particular level of coverage was always smaller (8-15%) when created with the amorphous algorithms. For the lung cancer patient, the mixed-PDF algorithm resulted in a PTV that was 13% smaller than the clinical PTV while still achieving ≥99.9% coverage.

CONCLUSIONS

The amorphous margin algorithms are better suited for the local effects of target shape and positional uncertainties than conventional margins. As a result, they provide superior target coverage with smaller PTVs, ensuring dose delivered to the target while decreasing the risk of normal tissue toxicity.