+ - 0:00:00
Notes for current slide
Notes for next slide

Data Intensive Brain Science

Joshua T. Vogelstein
Kavli Neuroscience Discovery Institute
questions: jovo@jhu.edu
slides: http://brainx.io/BSI-google
funding: DARPA I2O {GRAPHS, XDATA, D3M, L2M}

1 / 26

Motivation

2 / 26

source: NAMI
3 / 26

The Human Condition

4 / 26

The Grazing Goat Starves

5 / 26

Our Social Cage

6 / 26

Goal

Give each individual the tools she needs to move herself in the desired direction by the desired amount in our high-dimensional experience

7 / 26

Challenges

8 / 26

9 / 26

10 / 26

11 / 26

Enter NeuroData

12 / 26

Our Idea

Multiple measurements from same subject should be more similar to one another than they are to any measurement of any other subject

13 / 26

Our Idea

Multiple measurements from same subject should be more similar to one another than they are to any measurement of any other subject

P[∣∣xij−xij′∣∣<∣∣xij−xi′j′′∣∣]P[|| x_{ij} - x_{ij'}|| < || x_{ij} - x_{i'j''}||]

14 / 26

Our Idea

Multiple measurements from same subject should be more similar to one another than they are to any measurement of any other subject

P[∣∣xij−xij′∣∣<∣∣xij−xi′j′′∣∣]P[|| x_{ij} - x_{ij'}|| < || x_{ij} - x_{i'j''}||]


Without that, how can we trust biomarkers?

15 / 26

Our Idea

Multiple measurements from same subject should be more similar to one another than they are to any measurement of any other subject

P[∣∣xij−xij′∣∣<∣∣xij−xi′j′′∣∣]P[|| x_{ij} - x_{ij'}|| < || x_{ij} - x_{i'j''}||]


Without that, how can we trust biomarkers?

(we prove a bunch of stuff about our estimator)

16 / 26

Our model

17 / 26

Our Pipeline

18 / 26
19 / 26

Our Result

Within a study, subjects are "discriminable"

  • 25+ studies
  • 6000+ scans
  • largest "meganalysis"
  • largest open repo

20 / 26

Our Result

Within a study, subjects are "discriminable"

  • 25+ studies
  • 6000+ scans
  • largest "meganalysis"
  • largest open repo

(we tried 200+ pipelines ⇒ 1M+ compute hours)

21 / 26

But...

  • Across studies, populations are significantly different
  • Conditioning on Phenotype Fails

22 / 26

But...

  • Across studies, populations are significantly different
  • Conditioning on Phenotype Fails

☹

23 / 26

Our Proposed Solutions

  • better pipelines
  • better data (eg, quantitative MRI)
  • deep phenotyping
  • data acquisition harmonization
24 / 26

Acknowledgements

25 / 26

Questions?


♥, 🦁, 👪, 🌎, 🌌

26 / 26

Motivation

2 / 26
Paused

Help

Keyboard shortcuts

↑, ←, Pg Up, k Go to previous slide
↓, →, Pg Dn, Space, j Go to next slide
Home Go to first slide
End Go to last slide
Number + Return Go to specific slide
b / m / f Toggle blackout / mirrored / fullscreen mode
c Clone slideshow
p Toggle presenter mode
t Restart the presentation timer
?, h Toggle this help
Esc Back to slideshow