Radiation therapy aims to deliver high radiation dose to tumor target with minimal exposure to
surrounding normal tissue. However, he physics of energy decomposition and mechanical limitations restrict
the achievable dose distribution, necessitating tradeoffs between target coverage and normal tissue sparing.
Treatment planning is the process of obtaining control parameters to yield such a dose pattern. Dose volume
histograms (DVHs) are common tools in radiation therapy treatment planning to characterize plan quality and
used as objective functions. Despite their efficacy as a compact statistical summary of dose pattern, DVHs
provides limited spatial information as a report quantity and insufficient control as optimization objectives.
This limitation is particularly severe for treatment site with complex geometry, as in the case of head and
neck, with the presence of multi-level target volumes and various adjacent organs at risk (OARs). This paper
discussed a method for dose carving, by modifying the optimization objective in the treatment planning system.
Motivated by the emergent compressive sensing techniques, we introduce an objective function whose
minimization leads to more direct tradeoff between target coverage and OAR sparing, sharper dose dropoff
and better target dose homogeneity, achieving a better “carving” in the dose distribution. This principle,
though generally applicable to all sites, is particularly beneficial for cases with complex geometry, such as
head and neck planning.