Connectome Classification: Statistical Connectomics for Analysis of Connectome Data
Upcoming SlideShare
Loading in...5
×
 

Like this? Share it with your network

Share

Connectome Classification: Statistical Connectomics for Analysis of Connectome Data

on

  • 1,088 views

Talk at HBM11 (duration: 10 minutes)

Talk at HBM11 (duration: 10 minutes)

Statistics

Views

Total Views
1,088
Views on SlideShare
1,081
Embed Views
7

Actions

Likes
1
Downloads
10
Comments
0

1 Embed 7

http://twitter.com 7

Accessibility

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

Connectome Classification: Statistical Connectomics for Analysis of Connectome Data Presentation Transcript

  • 1. Connectome Classification: Statistical Connectomics for Analysis of Connectome Data Joshua T. Vogelstein, PhD d: Applied Math. & Stats u: Johns Hopkins w: jovo.me e: joshuav@jhu.edu
  • 2. Statistical Connectomics Statistics “the art of data collection and analysis”Connectomics “the study of connectomes” Statistical “the art of connectome data collectionConnectomics and analysis”
  • 3. Contributors Stats Carey E. Priebe Data Collection Glen A. Coppersmith Susan Resnick Mark DredzeConnectome Inference Will R. Gray Wisdom John Bogovic R. Jacob Vogelstein Jerry Prince Support: various grants
  • 4. Simplest. Example. Ever.Blind People Deaf People V1 V1 A1 M1 A1 M1
  • 5. Simplest. Example. Ever.Blind People Deaf People V1 No possible classifier V1 based on graph invariants can perform this insanely simple classification A1 M1 A1 M1 problem!!!
  • 6. Realest. Example. Ever.MR Connectome Gender Classification statistical graph model graph invariants > 83% accuracy < 75% accuracy
  • 7. Statistical Connectomics1. Collect Data Multi-Modal MR Imaging2. Preprocess Data MR Connectome Pipeline3. Assumptions Signal Subgraph4. Construct a Decision Rule Robust Bayes Plugin Classifier5. Evaluate Performance Leave-One-Out X-Validation6. Check Assumptions Synthetic Data Analysis7. Extensions Relax assumptions
  • 8. Statistical Connectomics1. Collect Data Multi-Modal MR Imaging2. Preprocess Data MR Connectome Pipeline3. Assumptions Signal Subgraph4. Construct a Decision Rule Robust Bayes Plugin Classifier5. Evaluate Performance Leave-One-Out X-Validation6. Check Assumptions Synthetic Data Analysis7. Extensions Relax assumptions
  • 9. 1. Collect Data: Multi-Modal MR Imaging• 49 senior individuals; 25 male, 24 female • diffusion: standard DTI protocol • structural: standard MPRAGE protocol
  • 10. 2. Preprocess Data:MR Connectome Automated Pipeline• coherent collection of code• fully automatic and modular• about 12 hrs/subject/core• yields 70 vertex graph/subject http://www.nitrc.org/projects/mrcap/
  • 11. 3. Data Assumptions: Signal Subgraph
  • 12. 4. Construct a Decision Rule:Robust Bayes Plugin Classifier• asymptotically optimal and robust• finite sample niceness ￿ auv 1−auv y= ˆ puv|y (1 ˆ − puv|y ) ˆ πy ˆ ˆ (u,v)∈S
  • 13. 5. Evaluate Performance: Leave-One-Out X-Validation incoherent estimator coherent estimator 0.5misclassification rate # signal−vertices 0.5 L π ˆˆ = 0. 5 ˆ L n b = 0. 41 ˆ L c o h= 0. 16 10 0.4 0.25 20 0.3 ˆ L i n c= 0. 27 30 0 0 1 2 3 0.16 10 10 10 10 200 400 600 800 1000 log size of signal subgraph size of signal subgraph some coherent estimators zoomed in coherent estimator 0.5lassification rate 0.5 star−vertices 15 0.4 18 0.25 0.3 0.16 21
  • 14. 6. Check Assumptions:Synthetic Data Analysis Correlation Matrix 1 100 0.5 vertex 0 200 −0.5 300 −1 100 200 300 vertex
  • 15. 7. Extensions• relax the independent edge assumption• relax binary edge assumption
  • 16. Discussion• 83% > 75%• yay statistical modeling!
  • 17. Q(&A)• anything?
  • 18. 4. Construct a Decision Rule: Signal Subgraph Estimation • for each edge, we compute the significance of the difference between the two classes using Fisher’s exact test • the incoherent signal subgraph estimator finds the s edges that are most significant • the coherent signal subgraph estimator finds the s edges that are most significant incident to m vertices
  • 19. 4. Construct a Decision Rule: Signal Subgraph Estimation negative log incoherent coherent significance matrix estimate estimate # correct = 15 # correct = 7 20 vertex n=64 40 60 20 40 60 −4.4 −1. vertex
  • 20. 6. Check Assumptions: incoherent estimator coherent estimator 1misclassification rate # star−vertices 0.75 0.7 10 0.25 0.5 Synthetic Data Analysis 20 0.5 30 0.3 0 0.18 0 1 2 3 200 400 600 800 1000 10 10 10 10 log size of signal subgraph size of signal subgraph 1 0.5 misclassification rate missed−edge rate coh 0.4 inc 0.3 nb 0.5 0.2 0.1 0 0 20 40 60 80 100 0 20 40 60 80 100 # training samples # training samples