NeuroInfo Glossary

Following are definitions of words that be unfamiliar to you, with a focus on how they are used in NeuroInfo software.

Brain atlas: A reference map of the brain showing anatomic structures and possibly other information like connectivity or function. Atlases are often used to identify parts of the brain or compare information in different experiments. Within NeuroInfo, a brain atlas comprises a reference image used for registration, a segmentation of this image into different anatomic regions, and the hierarchical relationship among brain regions.

Calibration: A NeuroInfo calibration captures basic information about an experimental image, including its anatomic orientation, the atlas to which it will be registered, and its approximate scale relative to the atlas. The same calibration can be used for many experimental images that are prepared in the same manner.

Experiment image: In NeuroInfo, this refers to a digital, microscopy image of a section or brain. This can come from many sources, for example serial section reconstructions of whole slide images, block-face imaging, or image acquisition from intact tissue (e.g., cleared tissue).

Intact image volume: a 3D microscopy image that represents a part or whole brain imaged without physical sectioning, for example from light-sheet microscopy of cleared tissue.

Physical coordinate space: defines the physical properties of an image in real-world units. This establishes a physical representation of the pixel coordinates of an image: the origin, spacing between pixels, and direction of the image axes.

Reference Section: A key section of the experimental image that has been interactively registered to a brain atlas. Reference sections typically have prominent anatomic features that make them suitable for interactive review to ensure accuracy. Reference sections are used in serial section registration to initialize the registration of other sections in the brain.

Registration: a process that finds the geometric mapping between the physical coordinate spaces of the experiment image and the atlas. Registration produces a transform that maps points in the experiment image onto corresponding points in the atlas. Image registration maps points from one image onto corresponding points on another, effectively aligning comparable images to each other. This mapping is achieved by changing the shape and or position (deforming/transforming) of one of the images to match the other; those changes are captured in the transform that is produced during the registration process.

The following types of registration can be applied to serial section images:

  • Single section registration: Each section has a completely independent transform

  • Serial section, Independent registration: Each section has an independent transform with some fixed parameters

  • Atlas-constrained registration:Each section has a transform with some independent parameters and some shared parameters.

For images of unsectioned specimens or intact image volumes, there is a single registration method and single transform applied to the whole volume.

Serial Section Reconstruction: a series of histologic section images that have been assembled into a 3D image volume representing the pre-sectioned brain, for example with the Serial Section Assembler.

Transform: maps geometric points from one physical coordinate space into another. In NeuroInfo, transforms essentially alter the size, shape, and/or sectioning angle of the reference atlas to match those features of the experimental section.

The following are different types of transforms used in NeuroInfo:

Linear Transformation:applies the same operation everywhere in space (e.g., translation, rotation, and scaling

Nonlinear Transformation: applies different operations in different parts of space (e.g., local deformation or warping)

Transform Set: a collection of transforms, one for each section in a serial section registration. Transform Sets are not relevant for image-volume registration.

When registering multiple sections, save sets of transforms in separate folders to avoid overwriting or mixing results from different animals or registration attempts.