News in oncology.

Artificial Intelligence-based Synthetic Simulation Computed Tomography Generation from Diagnostic Computed Tomography for Simulation-Free Workflow of Spinal Palliative Radiotherapy Dedicated simulation computed tomography (sCT) on a flat treatment couch is the standard for radiotherapy (RT) planning but creates significant delays—especially critical in palliative spinal RT where timely pain relief and neurologic protection are essential. Diagnostic computed tomography (dCT) is readily available yet unsuitable for direct planning due to differences in patient positioning and curved diagnostic table geometry. This study (medRxiv, 5Sep2025; DOI: 10.1101/2025.09.02.25334595) introduces a lightweight Artificial Intelligence framework that converts standard dCT into synthetic simulation computed tomography (ssCT) by correcting spine position and table curvature, enabling accurate, simulation-free palliative workflows. Two simple 3-layer neural networks (ReLU activation, Adam optimizer, MSE loss) were developed: Spine position adjustment:Uses heart center as reference; inputs heart point + dCT spinal-cord vector + slice z; outputs 3D shift to match sCT spinal-cord position [Figure 1].Table curvature adjustment: The table curvature adjustment module of the AI framework corrects the lower body contour on the dCT to match the flat tabletop geometry used in sCT acquisition [Figure 2]. Figure 1: Spine position adjustmentFigure 2: Table curvature adjustmentData: 30 spinal palliative RT patients from a safety-net hospital (22 train, 8 test); external validation on seven academic medical center (AMC) cases. Paired dCT/sCT rigidly registered on bone windows. Same clinical plan (with sCT contours) recalculated on dCT, ssCT, and reference sCT. Evaluation: Geometric alignment, PTV DVH metrics (Dmean, Dmax, D95, D99, V100, V107, RMS difference), and blinded 4-physician Likert scoring (1–5 scale) vs. sCT gold standard. Statistics: one-sided Wilcoxon signed-rank test; Gwet’s AC² for inter-observer agreement. Key Results Geometric correction ssCT spinal-cord trajectory and body contour aligned closely with sCT [visual overlays in Figures 3 and 4], eliminating dCT offsets.Figure 3: Two examples of spinal cord position adjustment resultsFigure 4: Table curvature adjustment resultDosimetric accuracy Table 1 presents the average differences in dosimetric parameters between dCT and ssCT compared to sCT as the reference, using the same treatment plan.Table 1: Average differences in dosimetric parameters between dCT and ssCT compared to sCT as the reference, using the same treatment planPhysician evaluation This presents the physician scoring of dCT and ssCT as treatment planning images, using a scale from 1 to 5 (5: Perfect, 4: Good, 3: Acceptable, 2: Unacceptable, 1: Poor). The result is shown with the standard deviation.Clinical example ssCT restored proper PTV/spinal-cord alignment and eliminated cold spots seen on dCT dose distribution. In summary, this study presents a lightweight, CPU-only AI framework that requires only 22 training cases to generate ssCT directly from routine dCT. The resulting ssCT achieves geometric and dosimetric accuracy comparable to dedicated sCT for spinal palliative RT, while substantially outperforming direct use of dCT. Physician confidence improved markedly—from “Acceptable” to “Good to Perfect”—particularly in resource-limited environments where optimal patient positioning is challenging. By algorithmically standardizing spine position and table curvature, the method eliminates reliance on technologist-dependent alignment and removes the need for a separate simulation scan. This offers a practical, low-barrier pathway to fully simulation-free palliative spinal RT workflows. Future work will focus on multi-institutional validation across diverse scanners and patient populations, development of probabilistic soft-tissue envelopes to address residual mobile organ uncertainties, and extension of the framework to other disease sites such as pelvic, hepatic, and thoracic malignancies. This innovation holds strong potential to reduce treatment initiation delays, lower resource demands, and improve timely access to high-quality RT worldwide, especially in safety-net and underserved settings. Proton versus Photon Radiotherapy for Patients with Oropharyngeal Cancer in the USA: A Multicenter, Randomized, Open – Label, Non- Inferior Phase 3 Trial Oropharyngeal cancer is on the rise, its management today revolves around multimodality treatment combining radiation, systemic therapy, and surgery. Photon-based intensity-modulated radiation therapy (IMRT) has been the backbone of this approach. But there is a problem: IMRT concurrent with chemotherapy carries a heavy toxicity burden. Across multiple phase 3 cooperative group trials, rates of severe malnutrition and gastrostomy tube dependence at end of treatment have consistently exceeded 60%. It represents a major quality-of-life crisis for patients. Intensity-modulated proton therapy (IMPT) offers a physically different approach. The Bragg peak of charged particles, protons deposit their dose at the tumor and stop, sparing the oral cavity, larynx, brainstem, and skull base from unnecessary radiation. Case-matched studies had hinted at better toxicity outcomes with IMPT.This trial (Frank SJ, et al. Lancet 2026; 407:174–84. DOI: https://doi.org/10.1016/S0140-6736(25)01962-2) was a multicenter, open-label, randomized non-inferiority phase 3 study run across 21 cancers centers and universities in the USA. They included 444 adult patients with stage III or IV oropharyngeal cancer (AJCC 7th edition) who were planned for concurrent systemic therapy with radiation to the primary site and bilateral neck. Radiation dose was 70 Gy in 33 fractions to the primary tumor and involved nodes with simultaneous integrated boost. Randomization was 1:1 to IMPT (221) or IMRT (219). The primary endpoint was progression-free survival (PFS). Most were HPV/p16 positive (95%). Most patients (86–89%) presented with stage IVA disease. Concurrent therapy, platinum-based chemotherapy was given to 91%, cetuximab in the remaining 9%. The trial met its primary endpoint. IMPT was non-inferior to IMRT for PFS (HR 0.88, 95% CI 0.57–1.35; P = 0.005 for non-inferiority). At 3 years, PFS was virtually identical 82.5% with IMPT versus 83.0% with IMRT. IMPT was associated with a 42% reduction in the hazard of death compared to IMRT (HR 0.58, P = 0.045). Subgroup analyses showed the protective effect of IMPT on overall survival was consistent. The strongest signals were seen in patients under 65, never-smokers, ECOG 0, base-of-tongue sub-site, no induction chemotherapy, and those receiving concurrent cisplatin. The toxicity story is where proton therapy really earns its place. At one year, not a single patient on the IMPT arm was still tube-dependent. Four patients on the IMRT arm were. Oral mucositis rates were similar in both groups (44% vs 41%) and late chronic grade 3+ toxicities were uncommon in both arms. This is the first phase 3 randomized trial to show that proton therapy can improve overall survival compared to photon IMRT while causing less harm. However, it is worthwhile to keep in mind that radiation delivery technology used in photon arm was still evolving over the 2013–2022 accrual period, and techniques varied across the 21 participating sites. This is a real limitation when comparing a technology-dependent modality like IMPT. Also, systemic therapy was not standardized, each center followed NCCN guidelines but made their own choices. Only 14% received induction chemotherapy and concurrent regimens differed. It should also be noted that as many as 23% of patients crossed over between arms after randomization but before treatment, because proton therapy insurance was denied post-randomization. This reflects the real-world access challenges around proton therapy, particularly in insurance-driven healthcare systems.