Learned synthetic CT enables physically tested lung-radiotherapy dose planning
Source check: Source checked. Not independent scientific confirmation.
Finding: A UNet++ model converted cone-beam scans into synthetic CT images for lung-cancer radiotherapy planning. Plans based on 17 patients’ images were compared with planning-CT references and checked through actual phantom irradiation. Measured dose deviations were 2–7%, comparable to the reference-plan deviations in that experiment. This supports a bounded medical-physics feasibility result, not formal dosimetric equivalence, improved patient outcomes or demonstrated adaptive treatment delivery.
Technology: UNet++ with ResNet50 backbone for synthetic CT generation
Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.
- AI contribution
- UNet++ learns the image-to-image conversion needed to reduce cone-beam CT intensity errors before conventional dose calculation. The generated-image plans, rather than only image-quality scores, are tested against physical dose measurements.
- Test and result
- Model development used 70 patients and image evaluation 15;17 additional patients supplied paired cone-beam/planning CT scans for offline planning comparisons. Generated-image plans underwent physical ArcCHECK and thermoluminescent-dose measurements on phantoms, with 2–7% dose deviations reported.
- Limitation / still unknown
- The clinical images and plan comparison were retrospective; phantom irradiation does not establish actual adaptive patient-treatment benefit. Nonsignificant dose differences do not prove equivalence. The tumor-volume gradient index differed, and some tumor-dose metrics had weaker agreement. The source’s percentage image-MAE improvement is inconsistent with its displayed means and is not reproduced here. Validation is limited to the studied scanner/workflow and geometry.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC11799926/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Exact primary XML front DOI/title/date; full patient/model Methods, generated-plan comparison and actual phantom/TLD measurement Results. Fresh Crossref publication-status metadata; PubMed/EPMC status access was limited..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.