-
Notifications
You must be signed in to change notification settings - Fork 8
/
neuronex-3dem.html
230 lines (164 loc) · 6.23 KB
/
neuronex-3dem.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
<!DOCTYPE html>
<html>
<head>
<title>NSF NeuroNex Workshop (3DEM)</title>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8"/>
<link rel="stylesheet" href="fonts/quadon/quadon.css">
<link rel="stylesheet" href="fonts/gentona/gentona.css">
<link rel="stylesheet" href="slides_style.css">
<script type="text/javascript" src="assets/plotly/plotly-latest.min.js"></script>
</head>
<body>
<textarea id="source">
name:opening
<img src="images/neurodata_blue.png" style="height:150px;float:right;"/>
### NeuroData: Embracing Open Source for Big Data Neuroscience
Eric Perlman<br>
<br><br><br><br><br><br><br>
<img src="images/funding/jhu_bme_blue.png" STYLE="HEIGHT:95px;"/>
<img src="images/funding/KNDI.png" STYLE="HEIGHT:95px;"/>
.foot[perlman@jhu.edu | <http://neurodata.io/talks/> ]
---
class: center
### <center>What is NeuroData?</center>
<img src="images/neurodata_venn.png" STYLE="width:50%;"/>
---
### Origin: Hosting public data (2011-)
<img src="images/website_datapage.png" style="width: 100%;"/>
.foot[<http://neurodata.io/data/>]
---
### ... but, it's hard.
- Don’t work alone!
- In 2011, Open Connectome Project set out to revolutionize data management for EM neuroscience with storage and analytics services on our cluster
- And we have had mixed results...
- ...but excited to have seeded a second generation
---
### The New Role of neurodata.io...
Shift from "build your own" to "see what's available; supplement where necessary."
- To facilitate a collaborative, open-source, cloud computing ecosystem for end-to-end neuroscience
- Collaborative: multiple organizations and communities developing and using the same tools
- End-to-end: data management from microscope to publication
- Visualize data at all stages
- Reproducible, extensible analysis
---
### What is our software stack?
<img src="images/nd_pipeline.png" STYLE="width:100%;"/>
.foot[https://neurodata.io/help/]
---
### A brief history of neurodata spatial stores...
- (2011-2013) Open-Connectome Project (OCP)
- First version of a spatial data store for volumetric cutouts and annotation
- (2013-2016) NeurodataStore (ndstore)
- Clustered, scale-out reimplementation of OCP on hosted NoSQL clusters
- Migration to the cloud in 2016
- (2014-present) IARPA MICrONS program drives Boss development
- JHU/APL builds and deploys for MICrONS performers
- (2016-present) neurodata.io adopts BossDB for open-science
- Parallel deployment of cloud services, independently managed
- Contribute features back through pull requests
---
### BossDB Features & Benefits
- ported to AWS for scalability
- random access to arbitrary cutouts
- downsamples
- experimental metadata storage (e.g. JSON)
- authentication
- spatial queries on annotations
<img src="images/bossDB.png" STYLE="width:100%;"/>
.foot[https://github.com/jhuapl-boss/boss]
---
### NDPush & NDPull data to/from cloud
- images: uint8 and uint16
- annotations: uint64
- file formats: png, tiff, jpg, numpy, etc.
- z-slice: flexible naming conventions
- multi-channel
- pip installable
.foot[https://github.com/neurodata/ndpush and https://github.com/neurodata/ndpull]
---
### Is a central (active) data store the right choice?
...not always.
---
### Other storage options
- hdf5
- n5 & Zarr provide chunked N-dimensional data stores
- same api for local or remote and parallel writes
- neuroglander's precomputed chunk & compressed segmentation format
- We are also exploring "Boss direct io", bypassing the service layer
.foot[https://github.com/saalfeldlab/n5/]<br/>
.foot[https://github.com/zarr-developers/zarr/]
---
### NeuroGlancer for Visualization
.pull-left[
Visualization with NeuroGlancer at all stages!
- Acquistion
- Alignment
- Analysis
]
.pull-right[
<img src="images/ndviz2.jpg" style="width: 120%;"/>
]
.foot[https://github.com/google/neuroglancer]
---
### Modifications to NeuroGlancer
- Boss support
- Including authentication
- Color selection
- Color picker & min/max values to generate shaders
- Used heavily on LM datasets
- Other ways to interact with state (export, load remote)
- Static sharing of views w/o python backend.
- Extra visualizations (e.g. render pair matches)
- Ontology info (e.g., display the region name from the ARA)
---
### Render Tools & services
- Web-based version of TrakEM2's XML datastore
- Stores references to raw data & transformations
- RESTful interfaces to load & update data
- Dynamic rendering of data
- Extended to support cloud data sources (e.g., raw images stored in an AWS)
<img src="images/render_stack_list.png" style="width:50%"/>
.foot[https://github.com/saalfeldlab/render/]
---
### Alignment verification...
<img src="images/synaptomes_pointmatch.png" style="width:100%;"/>
---
### Correlated EM-Light
<img src="images/synaptomes_em_light.png" style="width:100%;"/>
---
### Light sheet registration (ndreg)
<br>
<img src="images/ndreg.png" style="width: 100%;"/>
- LDDMM
- Fully automatic (no landmarks)
- Works on iDisco, Clarity, and multiple species...
---
### Other tools ...
- Image analysis
- Statistical machine learning (LOL, RerF, FlashX, knor, ...)
<img src="images/neurodata_tools.png" style="width: 50%"/>
.foot[https://neurodata.io/tools/]
---
### Acknowledgements
- neurodata: Randal Burns, Joshua Vogelstein & Ben Falk
- viz: Alex Baden & Jeremy Maitin-Shepard (Google)
- ndreg: Vikram Chandrashekhar & Daniel Tward
- synaptomes: Jesse Patsolic, Forrest Collman (AIBS), Sharmi Seshamani (AIBS), Stephen Smith (AIBS)
- render: Stephan Saalfeld (Janelia) & Eric Trautman (Janelia)
<img src="images/funding/nsf_fpo.png" STYLE="HEIGHT:95px;"/>
<img src="images/funding/nih_fpo.png" STYLE="HEIGHT:95px;"/>
<img src="images/funding/darpa_fpo.png" STYLE=" HEIGHT:95px;"/>
<img src="images/funding/iarpa_fpo.jpg" STYLE="HEIGHT:95px;"/>
<img src="images/funding/KAVLI.jpg" STYLE="HEIGHT:95px;"/>
---
class:center
<img src="images/neurodata_blue.png" style="width:50%;"/>
</textarea>
<!-- <script src="https://gnab.github.io/remark/downloads/remark-latest.min.js"></script> -->
<script src="remark-latest.min.js"></script>
<script type="text/javascript">
var options = {};
var slideshow = remark.create(options)
</script>
</body>
</html>