Machine Learning Segmentation Interface
The Machine Learning Segmentation interface contains of a number of panels — Model, Input, Method, and Result — that you access to complete a classification workflow. As shown in the screen capture below, the training inputs — datasets, multi-ROIs, and masks — are available on the Data Properties and Settings panel.
Choose Artificial Intelligence > Machine Learning Segmentation on the menu bar to open the Machine Learning Segmentation dialog.
Machine Learning Segmentation dialog and inputs
Machine Learning Segmentation Panels
You select the settings for training classifiers and exporting results on the following panels.
Model… Provides options to create new classifiers, as well as information about all the trained classifiers that are available (see Model Panel).
Input… Lets you select the inputs required to train a classifier, including the classifier engine, datasets, segmentation labels, and masks (see Input Panel).
Method… Lets you select the working area and the features that will be extracted from the input datasets to train the classifier (see Method Panel).
Result… Provides options to threshold inference results and to export segmentations and confidence maps (see Result Panel).
Preview Rendering
The Preview rendering options — Show segmentation and Show regions — can be used to control the way previews are displayed. You can show or hide segmented classes and the regions generated for region-based and majority-based training. You should note that the Preview rending box appears on multiple panels of the Machine Learning Segmentation dialog.
Preview rendering box
| Description | |
|---|---|
| Show segmentation |
If checked, the segmentation preview will appear in the current 2D view. You can adjust the opacity of the highlight applied to the segmented classes with the Intensity slider. |
| Show regions |
If checked, the regions computed for 'Region' and 'Majority Vote' training will appear at the selected Intensity setting. Additional options for viewing regions include applying a Look-Up Table (LUT) function, which determine how regions are highlighted and how colors are applied. The available LUTs can be edited interactively to optimize displays (see Using the LUT Editor).
Note Whenever a LUT is applied, you can select the Uniform color option to view the regions in a single color. |
Train, Preview, and Segment Controls
Controls to train models, generate previews, and segment datasets, which appear on multiple panels of the Machine Learning Segmentation dialog, are shown below and described in the following table.
Train, Preview, and Segment controls
| Description | |
|---|---|
| Lock | If selected or applied, the parameters of the trained classifier cannot be changed.
You should note that trained classifiers are always initially locked. If you unlock a trained classifier and make any changes, you will have to retrain the classifier. |
| Train | Available only after the required inputs have been added to the classifier and the features tree has been defined.
Click Train to train the selected classifier. |
| Preview | Available only after the selected classifier has been trained.
Click Preview to view a segmentation computed for the visible portion of the input dataset in the current 2D view. |
| Segment | Available only for trained classifiers when the required dataset inputs are available.
Click Segment to segment the currently selected input dataset or to export a confidence map. The segment outputs are selectable on the Result panel (see Export Options). |
