Ultra-low-field MRI operated during particle delivery without relevant dosimetric disruption
A portable 49 mT MRI system operated during proton and carbon-ion delivery with stable imaging and minimal dosimetric impact.
Tag
Technical relevance
Longitudinal MR radiomics detected distinct liver SBRT response patterns
Per-fraction MR radiomics identified distinct tumor and healthy-liver trajectories during SBRT that were not explained by volume or dose alone.
Fifteen-fraction central-boost planning was feasible for bulky retroperitoneal sarcoma
All 15 replans met target coverage, but one-third exceeded at least one bowel objective and kidney constraints remained challenging.
PET/CT reshaped esophageal radiotherapy targets and reduced lung dose
PET/CT reduced mean PTV by 149 cm³ while identifying substantial disease outside CT-based targets in 20% of patients.
Commercial AI contouring cut head and neck delineation workflow from 44 to 5 minutes
Three commercial AI models produced broadly usable head-and-neck contours, reducing review time to five minutes, but anatomical definitions differed importantly between vendors.
Biological dose calculations remain poorly standardized across radiotherapy practice
Clinicians applying dose-equivalence models to identical scenarios agreed only 25.8%, exposing major variability in treatment gaps, hypofractionation and reirradiation.
AI detected subtle temporal changes in routine prostate MR-Linac images
A deep-learning model ordered first-to-last fraction images with 95% accuracy, while smaller longitudinal changes were associated with later biochemical recurrence.
Online adaptation preserved kidney SABR coverage, but daily anatomical effects were small
Adaptation preserved prescription-level D99 in all 50 fractions, while a 1-mm margin achieved 98% despite minimal daily dose perturbation.
Elevating skin markers simplified proton CT simulation without sacrificing geometric accuracy
A 1.5–3.0 mm foam elevation reduced marker contouring time by 30–37% while keeping average positional error near 1 mm.
Monte Carlo grid size materially changed apparent plan quality in liver SBRT
A 3-mm grid cut Monaco calculation time about threefold but altered target, conformity and dose-falloff metrics despite similar gamma QA.
Synthetic CT from UTE-MRI reproduced lung radiotherapy dose with high accuracy
UTE-MRI-derived synthetic CT produced <1% target-dose differences and a 95.4% gamma pass rate at 2%/2 mm.
Deep learning halved ventricular and brainstem contouring time in pediatric radiotherapy
MRI-based nnU-Net segmentation achieved Dice scores ≥0.90 and reduced clinical contouring workflow time by more than 50%.
Locally fabricated phantoms matched commercial dosimetry for 3DCRT TPS verification
Low-cost acrylic and beeswax thorax phantoms produced dose measurements comparable with a commercial phantom and generally remained within IAEA tolerance.
DIBH workflow changes sharply reduced chest-wall motion during breast radiotherapy
Four quality-improvement measures reduced >3-mm chest-wall errors from 21.2% to 4.7%, but sternal marker placement increased skin reactions.
Fractionation mitigated proton interplay for most free-breathing thoracic treatments
Interplay-related underdosage largely disappeared after five fractions, although two high-motion plans retained clinically relevant target deficits.