Segmentation is an important structural processing step for many applications in BrainVoyager. It is an essential prerequisite for volume rendering as well as surface rendering of the head and brain. While it is very easy to segment and render the skin surface of the head, the successful segmentation of the cerebral cortex is a very difficult task requiring the application of several 3D image processing tools. While these tools are used in sequence during automatic cortex segmentation, they can also be used in isolation allowing to segment and render other anatomical structures such as the ventricles or the cerebellum. Furthermore, advanced cortex segmentation tools and deep learning based segmentation tools are provided that expect sub-millimeter resolution data with the goal to produce precise segmentations of the inner (white/grey matter) as well as the outer (pial or grey/CSF) boundary of the cortex enabling advanced applications.
The next topics describe the following major segmentation and mesh reconstruction tools:
- Head (skin) reconstruction using a sphere shrink-wrapping approach.
- The intensity inhomogeneity correction tool that is an important feature to improve the quality of any segmentation.
- Automatic cortex segmentation for conventional resolution (1 mm) 3D anatomical data.
- Advanced cortex segmentation for high-resolution high-quality 3D anatomical data.
- Deep neural network (DNN) based precise cortex segmentation of sub-millimeter data.
The advanced cortex segmentation "conventional" pipeline and the DNN based segmentation are both suited to prepare advanced applications such as cortical thickness analysis and laminar and columnar fMRI analysis. The DNN approach is recommended in case that the T1-weighted data fulfils the specified requirements for its application.
Copyright © 2020 Rainer Goebel. All rights reserved.