Deep learning could provide a second-check for prostate pelvic nodal contours
AI pelvic nodal contours achieved a Dice score of 0.79 on withheld prostate RT cases but retained localized boundary disagreements.
Tag
Technical relevance
Heart-base optimization substantially reduced cardiac substructure dose in distal esophageal RT
Heart-base-optimized IMRT/VMAT reduced cardiac avoidance area mean dose by about 6 Gy without compromising target coverage.
Simulation-free VMAT preserved delivered dose for palliative bone radiotherapy
Diagnostic CT-based VMAT preserved PTV D95 at 99.6% of planned dose across 30 palliative bone RT plans.
Proton RBE rose sharply beyond the Bragg peak in pancreatic cancer cells
In two PDAC cell lines, measured proton RBE remained near 1.0 through the Bragg peak but rose as high as 6.1 distally.
Small photon fields reduced detected Cherenkov signal despite stable dose-linked production
Detected Cherenkov signal per dose fell sharply in small fields, driven mainly by optical transport rather than major changes in intrinsic Cherenkov production.
Pretrained deep-learning IMPT planning transferred across centres with comparable plan quality
After two days of local configuration, a pretrained model generated oropharyngeal IMPT plans with overall quality comparable to manual planning.
Consensus QA improved target delineation for perihilar cholangiocarcinoma SBRT
Target Dice improved from 0.57 to 0.74 after consensus QA, although OAR constraints still limited PTV coverage in most plans.
Kidney proton SBRT needed a larger superior–inferior robustness margin
Kidney proton SBRT modeling supported 4-mm superior–inferior and 3-mm lateral/anterior–posterior positioning uncertainty with intrafraction repositioning.
Bayesian heart-dose constraints improved cardiac risk selection in NSCLC
Personalized Bayesian heart-dose limits improved identification of NSCLC patients at low risk of treatment-related cardiac troponin elevation.
Ovary–PTV distance defined where ovarian dose prediction became unreliable
Random forest best predicted ovarian dose after transposition, but all models deteriorated sharply when ovaries lay within approximately 17 mm of the PTV.
FDVH-guided re-optimization improved RapidPlan prostate VMAT plans
Patient-specific feasibility guidance improved target homogeneity and reduced rectal low-dose exposure after automated RapidPlan optimization.
Enhanced MLC modeling did not outperform optimized DLG modeling for single-isocenter SRS
Enhanced leaf and optimized DLG models produced similarly accurate SRS dose calculations, with DLG showing a small QA advantage.
Single-arc VMAT remained comparable to double-arc planning with bilateral hip prostheses
Double-arc VMAT was slightly more robust, but single-arc planning achieved clinically comparable dosimetry with simulated bilateral hip prostheses.
Jaw-partitioned VMAT increased peak-to-valley dose separation in spatially fractionated radiotherapy
Jaw-partitioned VMAT improved dose gradients and peak-to-valley separation in GRID and lattice plans, but approximately doubled monitor units.
Bragg peak FLASH proton plans matched conventional IMPT dosimetry in skull-base tumors
Bragg peak FLASH achieved high ultra-high-dose-rate coverage with similar target coverage to conventional IMPT, but produced higher target and CNS hotspots.