Liam S. P. Lawrence

Introduction: Glioblastoma (GBM) commonly recurs close to the original gross tumor volume (GTV), suggesting treatment-resistant subregions may be present before radiologically visible progression. This study investigated whether deep learning on apparent diffusion coefficient (ADC) maps could generate pixel-wise probability maps of treatment-resistant regions within the GTV. Methods: The training/validation set included images of 64 in-house patients after radiation fraction 10 (Fx10) and 40 external patients before treatment used as an auxiliary cross-domain training cohort; the held-out test set included 16 in-house Fx10 patients. Resistant regions were defined as the overlap between the GTV from scans acquired at radiation fraction 10 and subsequent recurrence after finishing treatment. Models were trained on 2D GTV-cropped slices with patient-level separation to prevent data leakage. Single- and multimodal MRI inputs were compared, and ADC-based models (including baseline UNet, pretrained ResNet34, BiomedCLIP-assisted UNet, and a BLIP image-text UNet) were later evaluated. Performance was evaluated using conventional metrics, including ROC AUC, PR AUC, F1-score, recall, specificity, and precision. Results: ADC was the most informative single modality, achieving 69% and 80% for Receiver Operating Characteristic (ROC)- and Precision-Recall (PR)-Area Under the Curve (AUC), respectively. In the ADC-based architecture comparison, no model was consistently superior across all metrics. The baseline UNet achieved the highest accuracy, F1-score, and recall, whereas pretrained ResNet34 and BLIP-assisted models showed slightly higher ROC AUC than the baseline UNet (both 70%). Error analysis showed that performance decreased substantially for small or sparse resistant regions, and false-negative errors were enriched near resistant-region and GTV boundaries. Conclusions: ADC-based deep learning showed feasibility for pixel-wise spatial mapping of treatment-resistant glioblastoma regions within the GTV. A UNet trained from scratch provided the strongest sensitivity/F1 trade-off, whereas pretrained and VLM-assisted models provided only metric-specific benefits. Model comparisons showed that BiomedCLIP-assisted and BLIP image-text UNet models did not clearly improve fine spatial segmentation performance over the baseline UNet.

Background/Objectives: Radiotherapy for tumors of the central nervous system (CNS) could be improved by incorporating advanced imaging techniques into treatment planning and response assessment. The objective of this narrative review is to highlight the recent developments in magnetic resonance imaging (MRI) and positron emission tomography (PET) for applications in CNS radiotherapy. Methods: Recent articles were selected for discussion, covering the following topics: advanced imaging on MRI-linear accelerators for early response assessment in glioma; PET for guiding treatment planning and response assessment in glioma; and contrast-enhanced imaging and metabolic imaging for differentiating tumor progression and radiation necrosis for brain metastasis treatment. Where necessary, searches of scholarly databases (e.g., Google Scholar, PubMed) were used to find papers for each topic. The topics were chosen based on the perception of promise in advancing specific applications of CNS radiotherapy and not covered in detail elsewhere. This review is not intended to be comprehensive. Results: Advanced MRI sequences and PET could have a substantial impact on CNS radiotherapy. For gliomas, the tumor response to therapy could be assessed much earlier than using the conventional technique of measuring changes in tumor size. Using advanced imaging on combined imaging/therapy devices like MR-Linacs would enable response monitoring throughout radiotherapy. For brain metastases, radiation necrosis and tumor progression might be reliably differentiated with imaging techniques sensitive to perfusion or metabolism. However, the lack of level 1 evidence supporting specific uses for each imaging technique is an impediment to widespread use. Conclusions: Advanced MRI and PET have great promise to change the standard of care for CNS radiotherapy, but clinical trials validating specific applications are needed.

Background: Direct three‐dimensional imaging of radiation beams could enable more accurate radiation dosimetry. It has been previously reported that changes in T ₁ ‐weighted magnetic resonance imaging (MRI) intensity could be observed during radiation due to radiochemical oxygen depletion. Quantitative T ₁ mapping could increase sensitivity for dosimetry applications. Purpose: We use an MRI linear accelerator (MR‐Linac) to visualize radiation delivery through the real‐time effects of dose on the spin‐lattice magnetic relaxation time ( T ₁ ) of water. We quantify the relationships between dose, spin‐lattice relaxation rates ( R ₁ ) and dissolved oxygen concentration to further investigate the mechanisms of T ₁ change. Methods: An ultrapure water phantom and a 1% agarose gel phantom were irradiated and imaged on a 1.5 T Elekta Unity MR‐Linac. Radiation plans were created using the Monaco treatment planning system. Images were acquired before, during and after radiation. A dual‐echo Look‐Locker inversion recovery pulse sequence was used for simultaneous dynamic T ₁ / B ₀ mapping. The change in R ₁ with respect to dose (∆ R ₁ /∆Dose) and the radiochemical oxygen depletion (ROD = ∆O ₂ /∆Dose) were measured. The relaxivity of oxygen ( r _1,O2 = ∆ R ₁ /∆O ₂ ) in water was also measured in a separate experiment with samples of various dissolved oxygen concentrations. The minimum measurable dose over a 20‐min period was estimated using a single‐tailed 99th quantile Student's t ‐distribution. Results: Changes to R ₁ were found to be spatiotemporally correlated to the predicted delivered radiation dose and persisted for at least 1 h after radiation. A complex dose plan could be imaged in the 1% agarose gel phantom, as the gel limits diffusion and convective mixing. In water, the ∆ R ₁ /∆Dose was found to be −1.0 × 10 ⁻⁴ s ⁻¹ /Gy, the r _1,O2 was found to be 5.4 × 10 ⁻³ s ⁻¹ /(mg/L), and the ROD was found to be −0.010 (mg/L)/Gy. Both r _1,O2 and ROD agree with published values. However, combining these two values yields a predicted ∆ R ₁ /∆Dose of −5.4 × 10 ⁻⁵ s ⁻¹ /Gy, indicating that radiochemical oxygen depletion alone under‐predicts the MRI effect. The detection limit of R ₁ was 1.1 × 10 ⁻³ s ⁻¹ which corresponded to a single‐voxel minimum detectable dose of 11.1 Gy for this specific sequence. Conclusion: Quantitative T ₁ mapping was used to image radiation dose patterns in real‐time in water and agarose gel. Radiochemical oxygen depletion only partially explains the T ₁ changes measured. Agarose gel could be used as a simple system for three‐dimensional patient‐specific quality assurance. Future applications may include in vivo dosimetry for FLASH radiotherapy, though improvements in acquisition methods and hardware are likely needed.

Purpose Treatment response assessment for gliomas currently uses changes in tumour size as measured with T₁- and T₂-weighted MRI. However, changes in tumour size may occur many weeks after therapy completion and are confounded by radiation treatment effects. Advanced MRI techniques sensitive to tumour physiology may provide complementary information to evaluate tumour response at early timepoints during therapy. The objective of this review is to provide a summary of the history and current knowledge regarding advanced MRI techniques for early treatment response evaluation in glioma. Methods The literature survey included perfusion MRI, diffusion-weighted imaging, quantitative magnetization transfer imaging, and chemical exchange transfer MRI. Select articles spanning the history of each technique as applied to treatment response evaluation in glioma were chosen. This report is a narrative review, not formally systematic. Results Chemical exchange saturation transfer imaging potentially offers the earliest method to detect tumour response due to changes in metabolism. Diffusion-weighted imaging is sensitive to changes in tumour cellularity later during radiotherapy and is prognostic for progression-free and overall survival. Substantial evidence suggests that perfusion MRI can differentiate between tumour recurrence and treatment effect, but consensus regarding acquisition, processing, and interpretation is still lacking. Magnetization transfer imaging shows promise for detecting subtle white matter damage which could indicate tumour invasion, but more research in this area is needed. Conclusion Advanced MRI techniques show potential for early treatment response assessment, but each technique alone lacks specificity. Multiparametric imaging may be necessary to aid biological interpretation and enable treatment guidance.

Background: Clinical target volume (CTV) contouring guidelines are frequently developed through studies in which experts contour the CTV for a representative set of cases for a given treatment site and the consensus CTVs are analyzed to generate margin recommendations. Measures of interobserver variability are used to quantify agreement between experts. In cases where an isotropic margin is not appropriate, however, there is no standard method to compute margins in specified directions that represent possible routes of tumor spread. Moreover, interobserver variability metrics are often measures of volume overlap that do not account for the dependence of disagreement on direction. To aid in the development of consensus contouring guidelines, this study demonstrates a novel method of quantifying CTV margins and interobserver variability in clinician-specified directions.Methods: The proposed algorithm was applied to 11 cases of non-spine bone metastases to compute the consensus CTV margin in each direction of intraosseous and extraosseous disease. The median over all cases for each route of spread yielded the recommended margins. The disagreement between experts on the CTV margin was quantified by computing the median of the coefficients of variation for intraosseous and extraosseous margins.Results: The recommended intraosseous and extraosseous margins were 7.0 mm and 8.0 mm, respectively. The median coefficient of variation quantifying the margin disagreement between experts was 0.59 and 0.48 for intraosseous and extraosseous disease.Conclusions: The proposed algorithm permits the generation of margin recommendations in relation to adjacent anatomy and quantifies interobserver variability in specified directions. This method can be applied to future consensus CTV contouring studies.

We consider the problem of designing a feedback controller for a multivariable linear time-invariant system, which regulates a system output to the solution of an equality-constrained convex optimization problem despite unknown constant exogenous disturbances; we term this the linear-convex optimal steady-state (OSS) control problem. We introduce the notion of an optimality model, and show that the existence of an optimality model is sufficient to reduce the OSS control problem to a stabilization problem. This yields a design framework for OSS control that unifies and extends existing design methods in the literature. We illustrate the approach via an application to optimal frequency control of power networks, where our methodology recovers centralized and distributed controllers reported in the recent literature.