Figure 1. C-DoTA architectures. Left: Vanilla DoTA, with a convolutional encoder passing the normalized BEV-CT as a sequence of 2D images to the transformer block, together with the normalised energy E, used for dose prediction (C-DoTA-d). Right: The architecture for predicting α and β parameters is expanded by an additional convolutional encoder and transformer with the predicted dose as an input. The output of both transformers for CT and predicted dose is finally processed by a third transformer block using cross attention.
Read new research from our #AI in #ParticleTherapy special issue which presents the 1st AI dose engine for #CarbonIonTherapy: sciencedirect.com/science/arti... #MedicalPhysics #radonc #DeepLearning #AIinHealthcare #TreatmentPlanning #AdaptiveRadiotherapy #RBE
