KAHIBARO
24.4. RT Structure Sets
Table of Contents
Contours
In radiotherapy planning, RT Structure Set files describe anatomical and treatment-related structures by using contours rather than by assigning a material to every voxel directly. For GATE users, these contours are mainly useful to define volumes of interest on top of an existing patient or phantom geometry, for example to extract dose statistics in a target or organ.
An RT Structure Set (DICOM RTSTRUCT) contains one or more structures. Each structure is represented as a set of two dimensional contours, usually drawn on axial CT slices. A contour for one slice is defined by a sequence of points in patient coordinates, typically in millimeters and in the same coordinate system as the CT image and RT plan.
Conceptually, each contour describes the boundary of a region on a given slice. When all slice contours for a given structure are combined, they form a three dimensional surface that encloses the structure volume. In GATE, you usually do not reconstruct that surface yourself. Instead, you rely on a Python or external tool that reads the RTSTRUCT file, interprets the contour points, and either produces a binary mask in the image space or extracts polygonal surfaces that you can later import or use for volume definitions and analysis actors.
Most treatment planning systems store several kinds of contours in the same RT Structure Set. Typical examples include the planning target volume, various organs at risk, external body contour, and auxiliary structures. The RTSTRUCT standard also supports multiple contour types per structure, for example closed planar polygons, which are those mostly used for clinical structures. When you work with GATE you usually focus on closed contours, because only these define a bounded region where you can score quantities like dose.
It is important to understand the relationship between the contour coordinates and the CT image grid that you already imported as a voxelized geometry. The contour points are usually expressed in the DICOM patient coordinate system, using attributes such as Image Position Patient and Image Orientation Patient from the associated CT slices. To use contours in GATE, you must ensure that you use exactly the same mapping between patient coordinates and GATE world coordinates as you used for the CT. Otherwise, the structures will appear shifted or rotated with respect to the patient geometry.
You will typically convert contours into either voxel masks or geometric objects. In the voxel mask approach, every CT voxel is classified as inside or outside the structure by checking its position relative to the contour polygons. The result is a label map that can be used for region based analysis, for example to compute mean dose to a specific structure. In the geometric object approach, contours are used to build a surface or mesh, for instance a polygonal model, which you can then import or use to define specific volumes. For dose evaluation in GATE, the mask approach is often simpler, because actors work naturally with voxel indices when producing structure based dose statistics.
Accuracy of the contour representation is important when you compute quantities such as dose volume histograms. Interpolation between slices is often required, because the spacing between CT slices is usually larger than the in plane pixel size. Many contour handling libraries implement a consistent way to interpolate between slices and to handle slice gaps. When preparing data for GATE analysis, you should use the same library or algorithm for all structures to avoid inconsistencies across different regions.
Finally, contours in an RT Structure Set can be associated with other DICOM objects, such as the CT series and the RT Plan. When you bring them into GATE, it is essential to match each contour to the correct CT image, using identifiers like the Study Instance UID and Series Instance UID. This ensures that contours align with the same anatomical dataset that you turned into voxelized geometry, which is a prerequisite for meaningful dose and activity analysis in later steps.
Regions of interest
Regions of interest, often abbreviated as ROIs, are the named structures defined within the RT Structure Set that you will use in GATE for analysis or for defining parts of the simulation geometry. Each ROI groups together all contours that belong to a particular anatomical or treatment related region. The RTSTRUCT file stores a list of ROIs, each with a unique numerical identifier, a name, and optional additional information such as type or color.
From the perspective of GATE based dosimetry or imaging simulations, ROIs serve two main purposes. First, they provide a way to label anatomical regions or treatment targets, which allows you to compute structure specific quantities, such as mean dose, dose volume histograms, or activity statistics. Second, they can guide you when defining simplified or derived geometries, for example by using an ROI to approximate a specific organ as a simple volume in a separate simulation.
Typical ROIs include target volumes like gross tumor volume, clinical target volume, and planning target volume, as well as multiple organs at risk such as spinal cord, parotid glands, lungs, liver, or heart. There is also often an external body contour ROI that defines the overall patient boundary. When you process an RTSTRUCT file for use with GATE, you usually select a subset of ROIs that are relevant to the scientific question, for instance those organs for which you wish to evaluate dose or those structures where activity is present.
To make ROIs usable inside GATE, you usually convert them into a representation that matches your simulation setup. If you already built a voxelized patient geometry from CT images, the most convenient representation is a voxel based ROI mask. In this mask, every voxel carries an integer label corresponding to the ROI it belongs to, such as 0 for background, 1 for body, 2 for a given organ, and so on. Once such a mask exists, GATE actors can use it to aggregate quantities over specific ROIs. For example, a dose actor can report mean and maximum dose in each ROI if it is configured with the associated label map.
When you create ROI masks, you need to map the contour based ROI definition from the RT Structure Set to the CT voxel grid. This involves checking which voxels fall inside the contours of a given ROI on each slice, taking into account the voxel spacing and slice thickness. The mapping must be consistent with the geometry that you imported into GATE. If you resampled the CT images to a different resolution before creating a voxelized geometry, you must also resample the ROI masks accordingly to keep ROI boundaries aligned with dose or activity grids.
In some workflows, instead of using voxel labels, ROIs are converted into separate geometric volumes. For instance, you might derive an approximate spherical or ellipsoidal volume from a complex contour based ROI and then create a corresponding volume in GATE for a simplified study. In that case, the RT Structure Set serves as a guide to derive the size, position, and orientation of the region of interest, while the actual simulation geometry remains analytic and easier to control.
ROIs are also central when you interpret simulation results. When GATE produces a three dimensional dose map or an activity distribution, those images by themselves do not tell you which voxel belongs to which organ or target. The ROI masks derived from RT Structure Sets allow you to overlay anatomical information on the simulated data. This lets you compute statistics like mean dose to a specific organ, or compare different treatment configurations by inspecting changes in ROI based dose volume histograms.
For a typical beginner workflow, you will usually rely on existing DICOM and image processing tools to read RT Structure Sets, list the available ROIs, select the ones you care about, and generate label maps. Once you have those label maps, you can connect them to GATE actors or external analysis scripts. The detailed steps of reading RTSTRUCT, generating ROI masks, and integrating them with GATE will be addressed through practical examples and data analysis chapters. In the context of RT Structure Sets, what matters is to see ROIs as structured, named regions derived from contours that provide the bridge between detailed clinical planning information and the quantities that you will compute in your simulations.
Views: 11