What MRI Sequences are Necessary for Automated Neural Network-Based Metastasis Segmentation - An Ablation Study

PURPOSE/OBJECTIVE(S)

Brain metastasis (BM) delineation is a time-consuming process in both daily clinical practice and research. Automated BM segmentation algorithms can be used to assist in this task. Most approaches to brain tumor segmentation, such as algorithms trained on the BraTS challenge, use four magnetic resonance imaging (MRI) sequences as input, making them susceptible to missing or corrupted sequences and increase the number of sequences necessary for MRI RT planning. The goal of this project is to compare neural networks with different combinations of input sequences for the segmentation of the contrast-enhancing metastasis and the surrounding FLAIR hyperintense edema. All models were tested in a multicenter international external test cohort. This allows us to determine which MRI sequences are needed for effective automated segmentations.

MATERIALS/METHODS

In total, we had T1-weighted sequences without (T1) and with contrast enhancement (T1-CE), T2-weighted sequences (T2), and T2 fluid-attenuated inversion recovery (FLAIR) sequences from 339 patients with at least one brain metastasis from seven centers available. Preprocessing yielded co-registered, skull-stripped sequences with an isotropic resolution of 1 millimeter. The contrast-enhancing metastasis as well as the surrounding FLAIR hyperintense edema were manually segmented to create reference labels. A baseline 3D U-Net with all four sequences as well as six additional U-Nets with different clinically plausible combinations (T1-CE; T1; FLAIR; T1-CE+FLAIR; T1-CE+T1+FLAIR; T1-CE+T1) of input sequences were trained on a cohort of 239 patients from two centers and subsequently tested on an external cohort of 100 patients from the remaining five centers.

RESULTS

All models that included T1-CE in their selected sequences showed similar performance for metastasis segmentation with a median Dice similarity coefficient (DSC) of 0.93-0.96. T1-CE alone likewise achieved a performance of 0.96 (IQR 0.93-0.97). The model trained with only FLAIR performed worse (DSC = 0.73, IQR 0.54-0.84). For edema segmentation, models that included both T1-CE and FLAIR performed best (median DSC = 0.93), while the remaining four models without simultaneous inclusion of these two sequences (T1-CE; T1; FLAIR; T1-CE+T1) reached a median DSC of 0.81-0.89.

CONCLUSION

Automatic segmentation of brain metastases with less than four input sequences is feasible with minimal or no loss of quality. A T1-CE-only protocol suffices for metastasis segmentation. In contrast, for edema segmentation, the combination of T1-CE and FLAIR seems to be important. Missing either T1-CE or FLAIR decreases performance. These findings may improve future imaging routines by omitting unnecessary sequences, thus speeding up procedures in daily clinical practice while allowing for optimal neural network-based target definitions.

Post-Operative Stereotactic Radiotherapy for Resected Brain Metastases: Results of the Multicenter Analysis (AURORA) of the German Working Group "Stereotactic Radiotherapy"

PURPOSE/OBJECTIVE(S)

While the results of prospective studies support the use of postoperative stereotactic radiotherapy (RT) to the resection cavity (RC) as the standard of care after surgery, there are several issues that need to be investigated such as factors for improving local control, risk of leptomeningeal disease and radiation necrosis. Further, the optimal dose and fractionation is still under debate.

MATERIALS/METHODS

The working group "Stereotactic Radiotherapy" of the German Society of Radiation Oncology (DEGRO) analyzed its multi-institutional database with 661 patients who received postoperative stereotactic RT to the RC. Treatment was performed at 13 centers between 2008 and 2021. Patient characteristics, treatment details, and follow-up data including overall survival (OS), local control (LC) were evaluated. Cox Regression and Kaplan-Meier curves with Log-rank Tests were calculated for selected variables.

RESULTS

In this retrospective study, overall survival was 61.5% at 1 year, 47.6% at 2 years, and 35.5% at 3 years, and local control was 84.6% at 1 year, 74.8% at 2 years, and 72.8% at 3 years. 96% of patients were treated with hypofractionated stereotactic radiotherapy (HSRT), only 26 patients received single fraction radiosurgery (4%). Prognostic factors associated with overall survival were Karnofsky Performance Status, RPA and GPA class, controlled primary tumor and absence of extracranial metastases, whereas prognostic factor associated with local control was planning target volume (23 mL or less).

CONCLUSION

HSRT is the most common fractionation form in the treatment of RCs in this multicenter analysis. This approach results in excellent OS and LC outcomes. OS in patients with resected brain metastases is mainly influenced by performance status. In regard to local control, RT of large cavities remain a challenge with significantly worse outcome.

Treatment Planning for Cardiac Radioablation: Multicenter Multiplatform Benchmarking for the XXX Trial

PURPOSE

Cardiac radioablation is a novel treatment option for patients with refractory ventricular tachycardia (VT) unsuitable for catheter ablation. The quality of treatment planning depends on dose specifications, platform capabilities, and experience of the treating staff. To harmonize the treatment planning, benchmarking of this process is necessary for multicenter clinical studies such as the XXX trial.

METHODS AND MATERIALS

Planning computed tomography data and consensus structures from three patients were sent to five academic centers for independent plan development using a variety of platforms and techniques with the XXX study protocol serving as guideline. Three-dimensional dose distributions and treatment plan details were collected and analyzed. In addition, an objective relative plan quality ranking system for VT treatments was established.

RESULTS

For each case, three coplanar volumetric modulated arc (VMAT) plans for C-arm linear accelerators (LINAC) and three non-coplanar treatment plans for robotic arm LINAC were generated. All plans were suitable for clinical applications with minor deviations from study guidelines in most centers. Eleven of 18 treatment plans showed maximal one minor deviation each for target and cardiac substructures. However, dose-volume histograms showed substantial differences: in one case, the PTV≥30Gy ranged from 0.0% to 79.9% and the RIVA V_14Gy ranged from 4.0% to 45.4%. Overall, the VMAT plans had steeper dose gradients in the high dose region, while the plans for the robotic arm LINAC had smaller low dose regions. Thereby, VMAT plans required only about half as many monitor units, resulting in shorter delivery times, possibly an important factor in treatment outcome.

CONCLUSIONS

Cardiac radioablation is feasible with robotic arm and C-arm LINAC systems with comparable plan quality. Although cross-center training and best practice guidelines have been provided, further recommendations, especially for cardiac substructures, and ranking of dose guidelines will be helpful to optimize cardiac radioablation outcomes.

Report on planning comparison of VMAT, IMRT and helical tomotherapy for the ESCALOX-trial pre-study

Background

The ESCALOX trial was designed as a multicenter, randomized prospective dose escalation study for head and neck cancer. Therefore, feasibility of treatment planning via different treatment planning systems (TPS) and radiotherapy (RT) techniques is essential. We hypothesized the comparability of dose distributions for simultaneous integrated boost (SIB) volumes respecting the constraints by different TPS and RT techniques.

Methods

CT data sets of the first six patients (all male, mean age: 61.3 years) of the pre-study (up to 77 Gy) were used for comparison of IMRT, VMAT, and helical tomotherapy (HT). Oropharynx was the primary tumor location. Normalization of the three step SIB (77 Gy, 70 Gy, 56 Gy) was D95% = 77 Gy. Coverage (CVF), healthy tissue conformity index (HTCI), conformation number (CN), and dose homogeneity (HI) were compared for PTVs and conformation index (COIN) for parotids.

Results

All RT techniques achieved good coverage. For SIB77Gy, CVF was best for IMRT and VMAT, HT achieved highest CN followed by VMAT and IMRT. HT reached good HTCI value, and HI compared to both other techniques. For SIB70Gy, CVF was best by IMRT. HTCI favored HT, consequently CN as well. HI was slightly better for HT. For SIB56Gy, CVF resulted comparably. Conformity favors VMAT as seen by HTCI and CN. Dmean of ipsilateral and contralateral parotids favor HT.

Conclusion

Different TPS for dose escalation reliably achieved high plan quality. Despite the very good results of HT planning for coverage, conformity, and homogeneity, the TPS also achieved acceptable results for IMRT and VMAT.

Trial registration ClinicalTrials.gov Identifier: NCT 01212354, EudraCT-No.: 2010-021139-15. ARO: ARO 14-01

Acute radiodermatitis in modern adjuvant 3D conformal radiotherapy for breast cancer - the impact of dose distribution and patient related factors

Purpose

This study was performed to evaluate skin toxicity during modern three-dimensional conformal radiotherapy (3D-CRT) and to evaluate the importance of dose distribution and patient related factors.

Material and methods

This study comprises 255 patients with breast cancer treated with tangential three-dimensional conformal radiotherapy (3D-CRT) after breast conserving surgery between 03/2012 and 05/2017. The median prescribed dose was 50.4 Gy (range 50–50.4) and 92.2% of the patients received a sequential boost of 10–16 Gy. Adverse skin toxicities (according to CTCAE v. 4.03 and the occurrence of moist desquamations) were assessed at the end of treatment. The dose distribution in the skin (5 mm strip from the patient outline) and in the CTV was evaluated and correlated to the CTCAE scores and the occurrence of moist desquamation.

Results

42.4% of the patients developed grade I, 55.7% grade II and 2% grade III skin toxicities. Moist desquamation was observed in 59 cases (23.1%). Dose distribution within the CTV and skin was homogenous with only small areas receiving 107% of the prescribed dose (median: 0.7 cm³) in the CTV and 105% (median 0.5 cm³) in the skin. On univariate analysis breast size as well as V107%(CTV), V105%(skin) and V80%(skin) correlated significantly (p < 0.05) with the incidence of skin toxicity. On multivariate analysis only V80%(skin) was confirmed as independent risk factor.

Conclusion

Modern tangential multi-field 3D-CRT allows a homogeneous dose distribution with similar skin toxicity as compared to studies performing IMRT. Dose distribution within the skin (V80%) might have a relevant impact on the severity of skin toxicity and the occurrence of moist desquamation.