PURPOSE: This paper outlines and demonstrates a programmatic method to incorporate spatial information into a dose volume histogram (DVH) by adding vector data on the location of pixels in the dose array relative to structures in the plan to construct a vectorized dose distribution (VDD). With this data the DVH can be subgrouped according to a wide array of vector constraint sets, defining the spatial relationship of pixels to one or several structures to construct a vectorized DVH (VDVH) to reveal vector relationships of dose regions to structures. METHODS: Mathematical models for construction of the VDD and VDVH are described and a dose-vector-histogram (DVctH) is introduced as a means of specifying the location of dose features such as "hot spots." Practical detail on a programmatic approach to implement the methods is provided. A set of tests utilizing phantom and SBRT lung image sets were carried out to demonstrate ability of VDVH and DVctH to reveal clinically relevant spatial detail in dose distributions. RESULTS: The VDVH and DVctH enabled decomposing DVH curves to reveal the relative location of pixels contributing to dose points on the curve. The metrics enabled specificity in defining the location and magnitude of dose features relevant to treatment plan evaluation. The VDD, VDVH, and DVctH differ from other methods described in the literature as a result of using vector based constraints for each pixel, rather than focusing only on distance by construction of a set of shells on around or within a structure and then subgrouping pixels in the overlap region. CONCLUSIONS: The method is an effective means to combine spatial information with DVH metrics and provides a practical means of specifying the location of dose features with respect other structures in the treatment plan.
PURPOSE: This paper outlines and demonstrates a programmatic method to incorporate spatial information into a dose volume histogram (DVH) by adding vector data on the location of pixels in the dose array relative to structures in the plan to construct a vectorized dose distribution (VDD). With this data the DVH can be subgrouped according to a wide array of vector constraint sets, defining the spatial relationship of pixels to one or several structures to construct a vectorized DVH (VDVH) to reveal vector relationships of dose regions to structures. METHODS: Mathematical models for construction of the VDD and VDVH are described and a dose-vector-histogram (DVctH) is introduced as a means of specifying the location of dose features such as "hot spots." Practical detail on a programmatic approach to implement the methods is provided. A set of tests utilizing phantom and SBRT lung image sets were carried out to demonstrate ability of VDVH and DVctH to reveal clinically relevant spatial detail in dose distributions. RESULTS: The VDVH and DVctH enabled decomposing DVH curves to reveal the relative location of pixels contributing to dose points on the curve. The metrics enabled specificity in defining the location and magnitude of dose features relevant to treatment plan evaluation. The VDD, VDVH, and DVctH differ from other methods described in the literature as a result of using vector based constraints for each pixel, rather than focusing only on distance by construction of a set of shells on around or within a structure and then subgrouping pixels in the overlap region. CONCLUSIONS: The method is an effective means to combine spatial information with DVH metrics and provides a practical means of specifying the location of dose features with respect other structures in the treatment plan.