Enhancing the Spatial Resolution of Dose Measurements Using a Super-Resolution Geometry-Informed Neural Network
This paper presents a DICOM RT-Plan-guided super-resolution framework that leverages a UNet++ model augmented with geometric aperture moments to reconstruct high-resolution 1 mm³ dose distributions from coarse 1 cm³ phantom measurements, thereby enabling accurate radiotherapy verification without requiring denser hardware or computationally expensive simulations.