Overview of Scanner Invariant Representations: Moyer et al. 2020, - - PowerPoint PPT Presentation

▶
overview of scanner invariant representations moyer et al
SMART_READER_LITE
LIVE PREVIEW

Overview of Scanner Invariant Representations: Moyer et al. 2020, - - PowerPoint PPT Presentation

Dec 22, 2022 215 likes •296 views

Overview of Scanner Invariant Representations: Moyer et al. 2020, Magn. Reson. Med. Daniel Moyer , Greg ver Steeg, Paul M Thompson MIDL 2020 USC ISI / MIT CSAIL The Scanner Problem Multi-site analyses have varying site signals. Site signals

slide-1
SLIDE 1

Overview of Scanner Invariant Representations: Moyer et al. 2020, Magn. Reson. Med.

Daniel Moyer, Greg ver Steeg, Paul M Thompson MIDL 2020

USC ISI / MIT CSAIL

slide-2
SLIDE 2

The Scanner Problem

Multi-site analyses have varying site signals. Site signals don’t generalize.

1

slide-3
SLIDE 3

The Scanner Problem

q(z|x) ⊥ s Data x z

{ } , , ,

Latent factors z such that z ⊥ s

Inv Representation

  • 1. Remove just the info

about s from x.

  • 2. Then free to use z

without s-bias.

2

slide-4
SLIDE 4

Architecture

Conditional Auto Encoder:

q(z|x) Encoder z s Conditional Decoder p(x|z, s) ˆ x x inputs x s site id. ˆ x recon.

Bound from Moyer et al. 2018: I(z, s) ≤ −Ex,s,z∼q[log p(x|z, s)]

  • Cond. Reconstruction

+ Ex[ KL[ q(z|x) ∥ q(z) ] ]

  • Compression

− H(x|s)

Const

.

3

slide-5
SLIDE 5

Comparison

MICCAI CDMRI Challenge 2018 dataset, Mirzaalian et al. 2018 Baseline

4

slide-6
SLIDE 6

End

Links:

  • Paper: arxiv:1904.05375
  • NeurIPS Paper: arxiv:1805.09458 arxiv:1904.07199
  • Inv. Code: https://github.com/dcmoyer/
  • Questions: dmoyer@csail.mit.edu

Funding: NIH Grants P41 EB015922, R01 MH116147, R56 AG058854, RF1 AG041915, and U54 EB020403, DARPA grant W911NF-16-1-0575, NSF Grant Number DGE-1418060

5