name:opening #### Ailey in an Hour:
(A "Soup-to-Nuts" Pipeline for Analysis of Whole Cleared Brain Data)
.foot[[Joshua T. Vogelstein](https://neurodata.io) | [jovo@jhu.edu](mailto:jovo@jhu.edu) | [neurodata.io](http://neurodata.io) | [@neuro_data](https://twitter.com/neuro_data)] ---
--- ### [Reggie](https://neurodata.io/reg/) Pipeline for analysis of CLARITY/COLM/iDisco/etc. data - Downloads 2D tiles from cloud - Cleans artifacts - Stitch 2D tiles into 3D volume using Terastitcher - Downsample to atlas resolution - Nonlinear registration atlas to data - Inverts registration to atlas - Upsample registration to native resolution - .gold[Detect all cells (works ok)] - .r[Track all processes (nothing works yet out of the box for us)] - Upload registered brain and atlas to cloud - Visualize registered brain in cloud - Democratize access --- ### [Reggie](https://neurodata.io/reg/)
- [Open source](https://github.com/neurodata/ardent) - [Documented with tutorials](https://ardent.neurodata.io/demo/demo.html) - [pip installable](https://ardent.neurodata.io/install.html) - GPU and distributed support - Mostly automatic - .r[Runs in a 1 hour] on LaVision data - .r[Runs in a 9 hours] on 1 TB data (COLM) --- ### LaVision → Overlay ← Allen Atlas
.footnote[ (from Ailey Crow at Karl Deisseroth)] --- ### Rat COLM → Overlay ← Waxholm
.footnote[ (1 TB image from Ailey Crow at Karl Deisseroth)] --- ### LaVision iDisco Rat ← Waxholm Atlas
.footnote[ .w[(from Audrey Branch at Michela Gallagher)]] --- ### Human MRI → Overlay ← Desikan
--- .headnote[Allen Atlas + CLARITY + AI Cell Detection (from Ailey Crow at Karl Deisseroth)]
--- ### Data Visualization & Democratization - From local to web viz with atlas overlays ([Neuroglancer](https://github.com/google/neuroglancer)) - From spatial database to flat files ([zarr](https://zarr.readthedocs.io/en/stable/)+[n5](https://github.com/saalfeldlab/n5) coming soon) - From centralized to distributed storage (no more [BOSS](https://github.com/jhuapl-boss/boss)) - From blosc to brotli compression ([brotli](https://github.com/google/brotli)) --- ### Discussion - LDDMM previously implemented in [ITK](https://github.com/InsightSoftwareConsortium/ITKNDReg/), but lacked GPU support - Nonlinear registration works well across modalities and taxa - Integrating 2D to 3D ([paper here](https://link.springer.com/chapter/10.1007/978-3-030-33226-6_18)) - Integrating missing data and artifacts [paper here](https://doi.org/10.1101/494005) - Applying now to Janelia [MouseLight](https://www.janelia.org/project-team/mouselight) project - Working on BICCN with Brain Architecture and Pavel Osten - Working with Sam Wang to integrate atlas with full cerebellum - Integrating into [scikit-image](https://scikit-image.org) (widely used image library) - Active development, community participation encouraged! -- #### Lessons learned - New cloud technology enables more scalable data access - From developer to integrator - Open-source, community-developed, widely-used, general tools - Highly collaborative: Google, Janelia, CSHL, APL, Stanford, JHU... --- ### Papers - A. Charles, et al. [Towards Community-Driven Big Open Brain Science](mailto:jovo@jhu.edu). Annual Reviews in Neuroscience, 2020. - J. T. Vogelstein, et al. [A Community-Developed Open-Source Computational Ecosystem for Big Neuro Data](https://www.nature.com/articles/s41592-018-0181-1). Nature Methods, 2018. - A. Branch, et al. [An optimized protocol for iDISCO+ rat brain clearing, imaging, and analysis](https://doi.org/10.1101/639674). bioRxiv, 2019. - D. J. Tward, et al. [Diffeomorphic registration with intensity transformation and missing data: Application to 3D digital pathology of Alzheimer's disease](https://doi.org/10.1101/494005). bioRxiv, 2019. - D. J. Tward, et al. [3D Mapping of Serial Histology Sections with Anomalies Using a Novel Robust Deformable Registration Algorithm](https://link.springer.com/chapter/10.1007/978-3-030-33226-6_18). International Workshop on Multimodal Brain Image Analysis, 2019 - V. Chandrashekhar, Nonlinear Registration for all. In prep. --- ### Code - [NeuroData Cloud](http://neurodata.io/nd_cloud) - [Reggie](https://neurodata.io/reg/) - [Neuroglancer visualization](https://github.com/google/neuroglancer) (not us) - [Zarr file format](https://zarr.readthedocs.io/en/stable/) (not us) - [n5](https://github.com/saalfeldlab/n5) (not us) - [Brotli compression](https://github.com/google/brotli) (not us) - [Scikit-Image](https://scikit-image.org) (not us) --- ### Website: [neurodata](https://neurodata.io)
- $\approx$ 100,000 unique visitors to date - $>$130 countries & $>$3,000 cities - $>$5,600 twitter followers - Support data from $>$30 labs (including several NeuroNex PIs) - Modalities/scales: EM/nano, LM/micro, MRI/milli - Taxa: worms, flies, fish, mice, rats, sea lions, coyotes, non-human primates, humans --- ### Acknowledgements
yummy
lion
baby girl
family
earth
milkyway
Randal Burns
Michael Miller
Daniel Tward
Vikram Chandrashekhar
Jesse Patsolic
Benjamin Falk
Kwame Kutten
Devin Crowley
--- class:center
.headnote[Questions?] --- class: middle, inverse ## .center[Extra Slides] ---
.footnote[Precomputed + Brotli + EM (from [Forrest Collman](https://alleninstitute.org/what-we-do/brain-science/about/team/staff-profiles/forrest-collman/) at Allen Institute)] --- ### Centralized vs. Distributed | Property | Centralized | Distributed |:--- |:--- |:--- | | Developer | Us → APL | Community | Where are data | cloud | cloud | Access rates | ~100 mbps | ~100 mbps | Who pays | me | you | Who is in control | me | you | Who has access | up to me | up to you | Costs | more | less | Who manages | all me | me + you + new software | Our current | yes | yes | Our plan | no | yes --- ### File Formats | Property | [BOSS](https://github.com/jhuapl-boss/boss) | CloudVolume + [Precomputed](https://github.com/google/neuroglancer/tree/master/src/neuroglancer/datasource/precomputed) | [Zarr](https://zarr.readthedocs.io/en/stable/) + [n5](https://github.com/saalfeldlab/n5) |:--- |:--- |:--- | | Developer | Us → [Will Gray Roncal](https://bossdb.org/about) | [Seung Lab](http://seunglab.org/) + [Google](https://github.com/google/neuroglancer/graphs/contributors) | [Biologists](https://www.cggh.org/) + [Saalfeld Lab](https://www.janelia.org/lab/saalfeld-lab) | Last commit | Aug 9 | Sep 26 | yesterday | Used by (github) | 0 | 18 | 121 | User Auth | yes | no | no | Data Management | yes | no | no | Our current | most | some | none | Our plan | no | maybe | probably --- ### Compression | Property | [Blosc](http://blosc.org/) | [Gzip](https://www.gzip.org) | [Brotli](https://github.com/google/brotli) |:--- |:--- |:--- | | Developer | [some guy](http://blosc.org/pages/francesc-alted-resume/)| [GNU](https://www.gnu.org/home.en.html) | [Google]() | Lat commit | Sep 30 | 2018 | Oct 2 | Used by (github) | ? | $\gg 0$ | 502 | Browser support | none | all | all | Compression size | ok | good | better | Decompression speed | great | good | good | Compression speed | great | good | ok | Our current | most | some | none | Our plan | none | none | all --- ### Visualization | Property | [Fiji/ImageJ](https://imagej.net/Fiji) | [CATMAID](https://catmaid.readthedocs.io/en/stable/) | [Neuroglancer](https://github.com/google/neuroglancer) |:--- |:--- |:--- | | Developer | [scientists](https://github.com/imagej/imagej/graphs/contributors)| [Janelia](https://github.com/catmaid/CATMAID/graphs/contributors) | [Google](https://github.com/google/neuroglancer/graphs/contributors)+us | Lat commit | Oct 8 | yesterday | yesterday | Used by (github) | ? | ? | 18 | Web | no | yes | yes | skeletons | yes | yes | yes | volume annotations | yes | recently | yes | analysis | lots | skeletons | none | atlas overlay | yes | maybe | yes | Our current | local | none | all | Our plan | local | none | all --- ### Registration | Property | [Reggie](https://neurodata.io/reg/) | [NDReg](https://github.com/InsightSoftwareConsortium/ITKNDReg) | [ANTS](https://github.com/ANTsX/ANTs) | [skimage](https://scikit-image.org) |:--- |:--- |:--- | | Developer | [Us](https://github.com/neurodata/ardent/graphs/contributors)| [ITK](https://github.com/InsightSoftwareConsortium/ITKNDReg/graphs/contributors) | [Brian Avants](http://stnava.github.io/) | [300+](https://github.com/scikit-image/scikit-image/graphs/contributors) | Lat commit | yesterday | Aug 21 | yesterday | yesterday | Used by | just us | nobody? | MRI people | millions? | Used by (github) | 0 | 0 | ? | 26,500 | Linear registration | yes | yes | yes | no | LDDMM | yes | yes | approximate | no | GPU Support | yes | no | no | no | cleared brains | yes | yes | no | no | Our current | yes | no | no | no | Our plan | yes | no | no | yes --- ### Access Rates
--- ### [Benchmarks](https://www.opencpu.org/posts/brotli-benchmarks/)
Size
Decompress
Compress
brotli = red, gzip = blue