[Amira 3D Pro / Avizo 3D Pro]
Context
While grayscale intensity-based methods remain a staple for detecting and segmenting data of interest in an acquired image, it is sometimes insufficient to properly identify key items that are similar in brightness but differ in shape or orientation. This article presents some potential approaches to segment data based on such features.
Process
1. Texture Classification
1.1. Load your data.
1.2. Open the Segmentation+ workroom.
1.3. Select a label image that contains all materials of interest (minimum 2 materials).
1.4. Set the Texture Classification parameters in the Advanced Segmentation Tools panel.
1.5. Click Compute Preview or select Auto compute to see the 2D preview of classification in the viewer window.
1.6. Click Compute Label to train a classifier to identify the materials in the current label image. This generates a new label image that indicates the most likely material for each pixel.
2. AI-Assisted Segmentation
2.1. Load your data.
2.2. Open the Segmentation+ workroom.
2.3. Create a representative label image (structure annotation) using other Segmentation+ workroom tools.
2.4. Create a Patch Set for the training materials, then add patches containing labeled materials and counter-patches with no materials (labeled or unlabeled).
2.5. In the Advanced Segmentation Tools panel, click to select the tool. The button is only available after you have created a Patch Set.
2.6. Select which material(s) to predict.
2.7. Select whether to overwrite the existing label image or create a new image.
2.8. Click Compute Label to run the AI-Assisted Segmentation Tool.
3. Filter by Measure Range
3.1. Load your data.
3.2. Segment your data using any other methods, including intensity-based approaches. Each element of your data should be in its own distinct label/material.
3.3. Open the Project workroom.
3.4. Right-click on your labeled data and create a Measure and Analyze -> Individual Measures -> Filter by Measure Range module.
3.5. Set the Measure that will be used for this filtering to "Shape_VA3d".
3.6. Modify the Range of values that will be kept in the parameters. The closer the values are to 1, the closer they are to a perfect sphere.
3.7. Click Apply to generate your result data.
Note
A complete tutorial on using the AI Assisted Segmentation tool is available in the software User's Guide. We recommend reviewing this tutorial to become familiar with the tool's features and workflow.