How is cell body-based segmentation done

The cell body stain is analyzed as a full 3D stack, but instead of reading it only as a series of XY images, the algorithm examines it from all three orthogonal orientations: XY (the standard imaging plane), XZ, and YZ. Each orientation shows the cell’s shape from a different angle, including its depth profile, which a single XY view can’t capture.

2. Cross-axis consensus
The predictions from the three views are integrated. A cell is only called when it shows continuous convergence across all three axes. A bright patch that appears in one slice but isn’t continuous in Z, or that doesn’t hold up in the other planes, doesn’t become a cell. This acts as a built-in quality filter against debris, out-of-focus signal, and single-slice artifacts.

3. True 3D cell objects
The result is a set of labeled 3D cell volumes, not flattened 2D outlines. Each cell has defined boundaries in X, Y and Z.