Glossary

precomputed

A chunked storage format for volumetric data that relies on infofiles to keep track of the volume’s metadata, used by neuroglancer. See the documentation.

CloudVolume, cv

A serverless Python client for reading and writing in precomputed format. Supports local, AWS S3, and GCS. Relies on service tokens. See the documentation.

Neuroglancer, ng

WebGL-based viewer for volumetric data in precomputed format. See the documentation.

GCP

Google Cloud Platform. Supported by zetta_utils.mazepa for executing tasks remotely using Kubernetes.

SQS

Amazon’s Simple Queue Service. Supported by zetta_utils.mazepa for keeping track of task executions when executing tasks remotely.

MIP, MIP level

Multum in parvo. Refers to the resolution of volumetric data, with MIP 0 being the highest resolution and successive MIPs being downsampled by a factor of 2 in XY. Used to reduce the data size.

EM

Electron microscopy.

SEM

Scanning Electron Microscopy.

FIB-SEM

Focused Ion Beam Scanning Electron Microscopy.

TEM

Transmission Electron Microscopy.

Section

An individual slice of tissue. A section is reprented by some Z value.

Alignment

The process of warping volumetric data so that the images seem to move smoothly in Z. Each section is warped by a field to produce an alignment.

Field

A layer specifying 2 dimensional vector for each point in 3 dimensional space. Used to represent how the pixels should move for warping images. zetta_utils uses torchfields to handle fields.

Inference

In zetta_utils pipeline context, usually refers to inferring the affinities for a dataset.

Affinities

The probabilities in 3 dimensions for each voxel of a dataset, representing the likelihood that the voxel belongs to the same cell as its neighbour.

Watershed

In zetta_utils pipeline context, using watershed on the affinities to break the volume down into a bunch of small volumes.

Agglomeration

The process of merging the small volumes generated by watershed into full cells.

Segmentation

The process of dividing an image volume into many small 3D objects, or the collection of the 3D objects. Requires running inference to get affinities, watersheding, and agglomeration.

ACED

State-of-the-art alignment pipeline developed by Zetta AI.

Mazepa

An event-driven task queue library used by zetta_utils. See zetta_utils.mazepa for details.

Task

The individual units of work that need to be executed. See zetta_utils.mazepa for details.

Operation

A wrapper for constructing Tasks. See zetta_utils.mazepa for details.

Flow

A definition of the workflow specifying the order of and dependencies between tasks. See zetta_utils.mazepa for details.

FlowSchema

A wrapper for constructing Flows. See zetta_utils.mazepa for details.