The Multi-Atlas Segmentation Tool generates ROIs by registering a series of images in a Reference Library to the image that has been loaded in the Data Manager. The Reference Images must each have a corresponding ROI. The tool registers the images in the Reference Library to the loaded data, and applies the same transform to the corresponding ROI. The tool then works on a voxel-by-voxel basis, classifying a voxel to the new ROI if a greater percentage of the reference sets classified that voxel as within the ROI than the threshold percentage set by the user. Important factors that influence performance include image contrast, animal positioning consistency, and the size of the reference library.
The Multi-Atlas Segmentation Tool is accessed by navigating to Advanced Modules -> Multi-Atlas Segmentation. Here, there are three options: Multi-Atlas Segmentation, Add to Library, and Download Sample Library.
Adding to Reference Library
The Reference Library is the basis of the efficacy of the Multi-Atlas Segmentation Tool. The Reference Library is stored Locally, and is built one data set at a time.
The first step in adding a data set to the Reference Library is to open the data and ROI. In order to add to the Library, there must be an active ROI in the 3D ROI Tool. There can be multiple ROIs, but it is important, when building the Reference Library, to keep the naming of ROIs consistent; that is, if the organ of interest is the left kidney, always call it exactly
Left Kidney, otherwise, it will not be recognized as all the same organ (however, capitalization does not have an affect).
With the data and ROI loaded simply navigate to and click Add to Library as described above. And the GUI will appear.
The GUI offers four editable fields:
|Image to Submit||This option allows you to select which of the loaded images to submit to the Reference Library.|
|Patient ID||The Patient ID can be anything that is unique and will help you identify this data set when selecting which scans to use from your Reference Library.|
|Protocol||The Protocol should include information as to how the image was acquired, the field of view, modality, etc. This will be used as a filter to determine which scans from the Reference Library will be used by the tool.|
|Species||The Species dropdown list contains frequently imaged species. If you wish to have additions made to the options provided please contact your Account Manager or email@example.com.|
Segmenting With The Tool
With the image on which the new ROI is to be drawn loaded in the Data Manager, navigate to and click on Multi-Atlas Segmentation Tool. This will open the Multi-Atlas Segmentation Tool GUI, from which the user can apply the segmentation settings and run the tool.
The tool relies on settings selected by the user. There are two levels of settings, the Basic Settings, explained below, and the Advanced Settings, which should not need to be modified.
For most users, the Basic Settings tab contains all the settings that will need to be changed. Here, users can select the appropriate Reference Images, the appropriate Test Image, ROIs to segment, and Probability Threshold to use.
Reference List Selection
Selecting the proper Reference scan is imperative to the proper functionality of the tool. The settings selected by the users that factor into the images used are outlined in the table below:
|Image||This is the image in the Data Manager that will be the basis of the ROIs generated.|
|Species||The species of the animal being segmented. This is based on the available options in the Reference Library.|
|Protocols||This list will display the protocols present in the Reference Library. Changing the protocols included will narrow the Reference List. It is suggested to only include Reference Images with the same protocol as the image loaded in the Data Manager.|
|ROIs||This list will display the possible ROIs to generate based on those present in the Reference Library. Only those selected will be generated by the Tool.|
Other Basic Settings
The other setting that can be edited by the user is the Probability Threshold. Following the registration of each Reference Image to the image in the Data Manager, each voxel is assigned a probability that it belongs in the ROI. Only voxels with a probability greater than the threshold set by the user will be mapped to the ROI. The default for this setting is
0.5 and typically, the ROI will become larger as the threshold is lowered.
The middle pane of the GUI allows the user to select which images to use for the final Reference List. These should already have been narrowed down by the user’s previous selections in the Settings pane. It is up to the user how many Reference Images to use; however, the time the tool takes to run is largely influenced by how many Reference Images are selected. On the other hand, performance is often positively influenced by increasing the number of Reference Images.
Run Tool and View Log
Once the Settings and the Reference Images have been selected, clicking Run in the lower right corner of the GUI will run the tool. The Log will outline the steps being taken by the tool. At its completion, the tool will remain open, but VivoQuant will navigate to the 3D ROI Tool, where the new ROIs can be viewed and modified.
Downloading Sample Library
A sample Reference Library is available depending on the VivoQuant license. This is to be used for testing and learning how to operate the tool. It can be downloaded from Advanced Modules->Multi-Atlas Segmentation->Download Sample Library.