Introduction

The Allen Developing Mouse Brain Atlas Grid Service gives users access to the 3D expression summaries that have been generated to support search and visualization in the Allen Developing Mouse Brain Atlas. The Atlas provides in situ hybridization (ISH) data for ~2000 genes over embryonic and postnatal stages. An informatics data processing pipeline was developed to enable the navigation and analysis of this large and complex data-set to identify gene expression with specific spatial and temporal regulation.

Using this service you can:

Download information about an image-series in xml format, including:
- Gene identifiers
- Specimen related information (i.e. age)
- Meta-information for each image (i.e. path, dimension) that comprises the image-series
- Transform to map each image location to canonical reference space and vice-versa
Download 3D expression summaries of an image-series resampled to a canonical reference space representing:
- Average expression density per voxel
- Average expression intensity per voxel

Usage

Prototype	http://developingmouse.brain-map.org/grid_data/download/[ImageSeriesID]
Return type	A zip file containing a compressed folder
Comments	ImageSeriesID is an integer. These can be found listed in any search returns from on the Atlas web application
Example	http://developingmouse.brain-map.org/grid_data/download/100042562

Compressed folder contents

image_series.xml	XML file containing information about the specimen and images
density.mhd	A simple text header file in MetaImage format describing the volume
density.raw	A raw uncompressed float (32bit) little-endian volume representing average expression density per voxel. Volume dimension is dependent on specimen age (see below). Value of "-1" represents no data.
intensity.mhd	A simple text header file in MetaImage format describing the volume
intensity.raw	A raw uncompressed float (32bit) little-endian volume representing average expression intensity per voxel. Volume dimension is dependent on specimen age (see below). Value of "-1" represents no data.

Volume dimension

age	grid resolution (in micron)	x - voxel dimension (anterior-to-posterior)	y - voxel dimension (superior-to-inferior)	z - voxel dimension	z - offset (in micron)
E11.5	80	70	75	40	-160
E13.5	100	89	109	69	-160
E15.5	120	94	132	65	-160
E18.5	140	67	43	40	-160
P4	160	77	43	50	-160
P14	200	68	40	50	-200
P28	200	73	41	53	-200

Informatics Processing

3D Reference models

The backbone of the automated informatics pipeline is a set of 3D reference models. For each time-point, a specimen is sectioned to span a nearly complete specimen and the slides are either Nissl or Feulgen-HP yellow stained to form one high density image series. The images are reassembled to form a consistent 3D volume. Structural delineation from the 2D reference atlas images are inserted into the 3D model and interpolated to create 3D structural delineations.

Download 3D reference models

Compressed folder content

grid.mhd/.raw	3D reference model per age at grid resolution. The intensity represents a grayscale version of the Nissl or Feulgen-HP yellow stain. The volume is in raw uncompressed uchar (8bit) format.
gridAnnotation.mhd/.raw	3D structural annotation per age at grid resolution. The value represents the <id> of the finest level structure annotated for the voxel. Note: the 3D mask for any structure is compose of all voxel annotated for that structure and all of its descendants. The volume is in raw uncompressed ushort (16bit) format.
ontology.xml	XML document describing the structural ontology. The values in the gridAnnotation volume corresponds to the <id> tag of a <structure>.

Alignment Module

Each image-series is registered to the age-matched canonical reference space based on maximization of image intensity correlation. The module interleaves reassembling a consistent 3D volume for the specimen with co-registration to the 3D reference space.

At the end of the alignment process, each image has a 2D affine transform that maps an image pixel position to a location in the specimen volume. Each image-series has a 3D affine transform that maps the specimen volume to the age-matched reference space. Transform parameters for each image and the image-series is reported in the image-series.xml file.

An affine transform consists of a matrix component (A) and offset component (b).

T(x) = Ax + b

A 2D affine transform is defined by 6 parameters. The first 4 represents the matrix part (A) in row-major order and the last 2 the translation part (b).

A 3D affine transform is defined by 12 parameters. The first 9 represents the matrix part (A) in row-major order and the last 3 the translation part (b)

Terminology

transform	dimension	input	output	notes
<tsv>	2D	image location as pixel index (starting from 0)	specimen volume location in microns	For any image, the z-value in the specimen volume is defined in the xml file with the <z-value> tag
<tvs>	2D	specimen volume location in microns	image location as pixel index (starting from 0)	Inverse of <tsv>
<tvr>	3D	specimen volume location in microns	age-matched reference space location in microns
<trv>	3D	age-matched reference space location in microns	specimen volume location in microns	Inverse of <tvr>

Expression Gridding Module

A detection algorithm is applied to each ISH image to create a mask identifying pixels in the high resolution image which corresponds with gene expression. Based on these mask images, the gridding module create low resolution 3D summary of the gene expression and transform the data to the age-matched canonical reference space. The expression data grids are used for downstream search and analysis, and they can also be viewed directly as 3D volumes in the Brain Explorer application. User can access expression data grids through the Grid Service.

Terminology

data grid	definition
expression density	For each grid voxel, expression density is the number of expressing pixel (as defined by the expression mask) divided by the number image pixels which overlaps the voxel. A value of "-1" indicates no data. In the Brain Explorer application, this value is represented as the size of the sphere at each grid location.
expression intensity	For each grid voxel, expression intensity is the averaged inverted ISH grayscale value at expressing pixels within the span of the grid voxel. A value of "-1" indicates no data. In the Brain Explorer application, this value is represented as the color of the sphere at each grid location.
expression energy	For each grid voxel, expression energy is defined as (expression intensity * expression density). Anatomic search, temporal search, Neuroblast and AGEA features on the Atlas are all based on expression energy. An expression energy volume can be computed from expression intensity and expression density volumes. Note: a "-1" value in either volume indicates no data.