Pipeline for the Creation of Surface-basedAveraged Brain Atlases

Anja Kuß

Zuse Institute Berlin

1/30/2007 2/13

Pipeline for the creation of surface-based averaged brain atlases

Course• Motivation and related work• The pipeline• Applications• Conclusion and future work

1/30/2007 3/13

Pipeline for the creation of surface-based averaged brain atlases

Motivation• Goals in neurobiology

– Investigation of anatomical and functional neural network properties

– Analysis of learning and behavior

• Usage of insect brains– Small, easy to access– Already identified structures and neurons

• Population-based brain atlases– Spatial reference for relating data from different experiments– Facilitate visualization of neurobiological data

1/30/2007 4/13

Pipeline for the creation of surface-based averaged brain atlases

Related work• Human brain atlases

– Registration and averaging strategies: Guimond et al. 2000, Toga et al. 2001

– Processing environment: Rex et al. 2003

• Invertebrate brain atlases– Fruitfly: Rein et al. 2002– Honeybee: Brandt et al. 2005

Sequencing and adjustment of methods

Integration into one visualization platform: amira

Averaged brain atlas of the fruitfly, data from Rein et al. 2002

1/30/2007 5/13

Pipeline for the creation of surface-based averaged brain atlases

Pipeline - Overview• Imaging and preprocessing• Segmentation• Averaging• Surface reconstruction

Schematic overview of the pipeline for the generation of surface-based insect brain atlases

1/30/2007 6/13

Pipeline for the creation of surface-based averaged brain atlases

Imaging and preprocessing• Confocal microscopy of stained brains• 2x3 overlapping tiles, 80-100 slices of 512x512• Merging of tiles• First visualization via slicing or volume rendering

1/30/2007 7/13

Pipeline for the creation of surface-based averaged brain atlases

Segmentation• Creation of label fields dividing the brain into already

defined substructures (neuropils)• Support of manual segmentation

Segmentation of neuropils in the honeybee brain

Median calyx

Mushroom body

Lateral calyx

Pedunculus

LobulaMedulla

Cb

1/30/2007 8/13

Pipeline for the creation of surface-based averaged brain atlases

Averaging• Affine registration of all data sets to a chosen template• Non-arithmetic label averaging• Elastic registration of all data sets to the resulting

average and repeated label averaging

Unaligned label fields Result of affine registration Result of first elastic registration

1/30/2007 9/13

Pipeline for the creation of surface-based averaged brain atlases

Surface generation• Generation of non-manifold surfaces using a generalized

reconstruction algorithm• Simplification and remeshing• Tools for editing the surface

~ 100.000 triangles ~ 25.000 triangles ~ 25.000 triangles, remeshed

Surface-based averaged model of the honeybee brain

1/30/2007 10/13

Pipeline for the creation of surface-based averaged brain atlases

Refining the atlas

Bee brain with a selection of integrated neuronsNeuron (red) sticking out the atlas after affine registration, the fitted neuron is shown in green

1/30/2007 11/13

Pipeline for the creation of surface-based averaged brain atlases

Applications• Analysis of the olfactory system in the honeybee brain• Standard atlas of locust and moth brain• Atlas-based segmentation• Extension of the fly brain atlas:

– Averaging of thoracic ganglia– Integration of projections of the legs and wings– Integration of identified neurons

Atlas of fly thoracic ganglion with overlaid individual surface reconstruction

1/30/2007 12/13

Pipeline for the creation of surface-based averaged brain atlases

Conclusion and future work• Methods for the creation and extension of population-

averaged atlases from 3D image data of insect brains• Atlases facilitate scientific investigations

• Hierarchical brain atlases– Anatomical and functional hierarchies– Representation at various length scales

• Generation of statistical shape models

Thanks for your attention

1/30/2007 14/13

Pipeline for the creation of surface-based averaged brain atlases

Future work• Hierarchical brain atlases

– Anatomical and functional hierarchies– Representation at various length scales

• Generation of statistical shape models

Pe1 neuron and surrounding neuropils