Abstract Objective: FLASH radiotherapy requires ultra-high dose rates (>40 Gy/s), yet commercial treatment planning systems are currently unavailable. Monte Carlo (MC) simulation offers high transport accuracy, but full-head modeling for every setup is computationally prohibitive. This work establishes a methodology for generating pulse-width-specific phase-space files for the Mobetron UHDR (9 MeV), accounting for systematic beam quality shifts caused by RF waveguide loading across pulse widths of 1.2-4.0 μs. Approach: Using GAMOS 6.2.0 MC simulations, voxelized dose distributions were post-processed via an iterative script to refine source-level parameters against experimental targets. Mean energy was optimized by matching phantom-measured R50, while energy spread was refined using surface dose and build-up gradients. Relationships derived from a mid-range 6 cm applicator were applied across clinical apertures (2.5-10 cm) to test aperture-independence. The geometric mean of experimental pulse widths (2.28 μs) was evaluated as a single reference condition. Main Results: Phase space files were generated across pulse widths and apertures. For the A6I6 reference, mean energy decreased monotonically from 9.58±0.24 to 9.04±0.24 MeV with pulse width, best described by a logarithmic model (LOO-RMSE=0.033 MeV). Energy spread increased linearly from 0.173±0.020 to 0.287±0.020 MeV (LOO-RMSE=0.005 MeV). Cross-aperture validation confirmed these shifts occur upstream of collimation. Evaluating the 2.28 μs reference predicted 9.30 MeV mean energy and 0.215 MeV energy spread, yielding depth-dose deviations ≤1.3 mm across experimental extremes and satisfying AAPM TG-106 tolerances (≤2.0 mm). Significance: This work provides pulse-width-specific phase-space characterization for Mobetron FLASH commissioning. Cross-validated regression models enable beam parameter prediction within 1.2-4.0 μs, and the 2.28 μs reference phase space streamlines routine planning while maintaining clinical accuracy.
Abstract Objective.Pretrained foundation-model (FM) embeddings are proposed as robust alternatives to handcrafted radiomic features, yet whether they better withstand segmentation-mask variability, the perturbation that radiomics intraclass-correlation (ICC) filtering exists to address, has not been tested head-to-head. We quantified how segmentation-mask perturbation propagates into feature reproducibility and downstream malignancy discrimination for FM embeddings and Image Biomarker Standardization Initiative (IBSI)-compliant radiomics measured on a common cohort. Approach. Two 2D FM encoders (BiomedCLIP; RadImageNet ResNet50) were compared against IBSI-compliant radiomics (3D, 107 features; 2D, 102 features) on the full eligible Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) cohort: 556 lung nodules (300 malignant, 256 benign from radiologist consensus ratings, not pathology) from 399 patients. Up to thirteen perturbed masks per nodule (up to four reader masks, eight synthetic, one Dice-calibrated) were generated, and ICC(2,1) was computed across the ten perturbation conditions available for every nodule. Malignancy classification used L2-regularized logistic regression with patient-grouped five-fold cross-validation and bootstrap 95% confidence intervals on every area under the receiver operating characteristic curve (AUC). Main results. Radiomics outperformed both foundation models on discrimination (AUC 0.940 and 0.930 for radiomics-2D/3D versus 0.833 BiomedCLIP and 0.798 RadImageNet; all radiomics-versus-FM differences significant by DeLong test, p≤10^(-7)). On perturbation stability the foundation models diverged: BiomedCLIP was statistically indistinguishable from radiomics (93.8% versus 90.2–92.5% stable features) and the most stable paradigm, whereas RadImageNet was significantly less stable (75.0%). A pre-specified diameter-stratified analysis showed the radiomics advantage narrows, and partly reverses, within size-matched strata; decision-curve analysis favoured radiomics. Significance. Two-dimensional foundation-model embeddings offered no discrimination advantage over IBSI radiomics under segmentation perturbation, and greater stability did not translate into greater accuracy. Perturbation stability is therefore decoupled from task-informativeness, and comparative benchmarks should report patient-grouped, size-stratified, calibration-aware metrics, not aggregate accuracy alone.
Abstract ObjectiveTo develop a 3D AI framework for detecting and localizing multiple lesions in whole-body FDG PET/CT, and to evaluate whether a pretrained PET/CT foundation model can improve lesion detection performance compared with conventional training from scratch.ApproachThe AutoPET cohort comprised 1,014 studies divided into 700 training, 114 validation, and 200 subject-independent test studies. Tumor masks were converted into 3D bounding boxes. We developed a one-stage 3D lesion detector using a PlainConvUNet backbone to extract multi-scale encoder-decoder features for lesion classification and box regression. To leverage foundation-model representations, the PlainConvUNet backbone was initialized with weights from our previously developed multi-center PET/CT foundation model, pretrained using masked autoencoding, and subsequently fine-tuned for lesion detection. Four configurations were compared: PET+CT trained from scratch, PET only, CT only, and PET+CT initialized from the pretrained PET/CT foundation model. Detection performance was evaluated using free-response receiver operating characteristic (FROC) and alternative FROC (AFROC) analyses, qualitative analysis, and lesion-size-stratified operating curves.Main resultsMultimodal PET+CT consistently outperformed either modality alone across task-based detection metrics, demonstrating the complementary contributions of metabolic and anatomical information. Importantly, initialization from the pretrained PET/CT foundation model further enhanced lesion detection performance compared with training the same PET+CT detector from scratch, increasing absolute FROC sensitivity by 0.034-0.041 across the evaluated 1-4 false-positive-per-scan operating range and improving AFROC AUC by 0.028. Detection performance increased with lesion size, with small lesions remaining the most challenging. The separation between the pretrained and randomly initialized PET+CT configurations was more evident in the smaller-lesion strata, whereas their operating curves increasingly overlapped for larger lesions.SignificanceThis study provides a detection-centered, task-based characterization of multi-lesion analysis in whole-body PET/CT. Multimodal PET+CT improves lesion detection compared with either modality alone, lesion size is a major determinant of detection performance, and initialization from a pretrained PET/CT foundation model provides an additional performance improvement.
Abstract Objective: Linear energy transfer (LET)-informed proton therapy has the potential to improve biological treatment planning beyond the conventional, constant-RBE assumption. Its clinical translation, however, requires experimental tools capable of verifying spatially resolved dose and dose-averaged LET (LET D ) distributions under clinically relevant conditions. We recently introduced an optically stimulated luminescence (OSL) film-based detector system that simultaneously measures two-dimensional (2D) absorbed dose and LET D . This work provides a first assessment of the system for verifying dose and LET D distributions from clinical proton therapy plans. Approach: The system was tested at two proton therapy centers, ranging from simple uniform fields to more complex geometries, including Bragg curve measurements and a clinically realistic anthropomorphic head phantom treatment plan. Measured dose and LET D distributions were compared against commercial treatment planning system (TPS) calculations and, for selected fields, to external Monte Carlo particle transport simulations due to the absence of traceability of LET D . For non-uniform irradiations, agreement was assessed using gamma analysis with 3%/2 mm and 5%/2 mm criteria for dose and LET D , respectively. Main results: For uniform fields, measured dose and LET D distributions generally agreed with TPS and independent Monte Carlo simulations across a range of clinically relevant irradiation conditions. In the Bragg curve measurement, gamma pass rates were 96.6% for dose and 99.7% for LET D , with the largest dose discrepancy observed near the Bragg peak, consistent with a current over-correction of detector quenching in high-LET regions to be improved in future work. In the anthropomorphic phantom, gamma pass rates ranged from 75% to 100% for dose and from 45% to 95% for LET D , with a single layer showing a sub-50% pass rate for LET D . Failing regions were confined to high-dose superior layers for dose and distal-edge, high-LET regions for LET D . Significance: The presented results support the detector system’s use as an experimental benchmarking tool for TPS-calculated dose and LET D maps. Given this and the system’s ability to simultaneously measure 2D dose and LET D maps with high spatial resolution and minimal field perturbation, the system shows promise for future commissioning workflows for LET-informed proton therapy.
Abstract The shift from passive scattering to active pencil beam scanning (PBS) in proton therapy, alongside techniques such as flattening filter-free photon therapy, has increased dose rates in clinical radiotherapy from <0.4 Gy/s to 0.4–35 Gy/s. Meanwhile, ultra-high dose rate radiotherapy (FLASH, ≥40 Gy/s) has emerged as a novel modality that spares normal tissue while maintaining tumor control. Yet, dose rate effects within the 0.4–35 Gy/s range have traditionally been considered negligible. Recent reports of unexpected necroses after proton PBS irradiations challenge this assumption, prompting a reevaluation of potential dose-rate-dependent effects in this intermediate range. This review summarizes current evidence on dose rate effects within 0.4–35 Gy/s, derived from in vivo models (including zebrafish embryos and rodents), in-vitro experiments and water radiolysis studies. Selected experimental and retrospective observations suggest that dose rates in this inter mediate range may modify the biological response and, under certain conditions, increase toxicity relative to conventional reference dose rates (<0.4 Gy/s). The magnitude and even direction of dose rate effects appear to depend on radiation quality, total dose, temporal delivery structure, oxygenation, linear energy transfer, tissue type and biological endpoint. The observations described in this review are hypothesis-generating rather than threshold-defining, highlighting the need for further investigation in experimental conditions, preclinical models, and clinical settings.
