The first four items mirror the main AI page. Use the archive controls to reveal more in batches of four without leaving the page.
4 of 8 items visibleGE HealthCare announced FDA 510(k) clearance for MIM Contour ProtegeAI+ 2.0, an AI-enabled auto-contouring tool for radiation therapy planning. This is still the sort of announcement that needs local validation, contour review, and commissioning before any department treats it as routine clinical infrastructure.
Read sourceA Cureus review surveyed AI use in prognostic modelling, adaptive monitoring, response assessment, and toxicity prediction. Its main value is not hype but restraint: the paper repeatedly points back to inconsistent methods, limited external validation, and the need for governance before routine use.
Read sourceThis arXiv preprint describes a workflow-integrated large-language-model system for physician summaries and potential trial identification. It is interesting as an operations signal, but it remains early-stage evidence rather than proof of clinical accuracy or downstream patient benefit.
Read sourceSiemens Healthineers announced CE marking for AI contouring capabilities integrated with the Eclipse treatment planning system. Like any vendor milestone, it is relevant for workflow direction, but it does not remove the need for local commissioning, site-specific QA, and physician review.
Read sourceA perspective in npj Digital Medicine argues for human-centric intelligent treatment planning, with AI operating under clinician oversight rather than replacing planning judgment. It is useful because it focuses on workflow design, safety, and planner interaction instead of treating automation as a goal by itself.
Read sourceA Frontiers in Oncology paper reported daily AI-assisted adaptive small-margin radiotherapy during trimodal therapy for muscle-invasive bladder cancer. This is closer to practice-facing evidence than a generic AI claim, because it sits inside an actual adaptive workflow rather than a purely technical benchmark.
Read sourceA Radiation Oncology paper used longitudinal delta-radiomics and explainable machine learning to predict radiation-induced xerostomia during head-and-neck treatment. It matters because the model is not only predictive but interpretable, which is much more clinically useful than a black-box score alone.
Read sourceThe AIRT arXiv preprint presents an end-to-end deep-learning framework for prostate VMAT planning in under one second. The speed claim is striking, but as with any preprint, the clinical takeaway should stay cautious until independent validation and real-world workflow testing catch up.
Read source