Feature Dis isentanglement to Aid id Im Imaging Biomarker - - PowerPoint PPT Presentation

▶
feature dis isentanglement to aid id im imaging biomarker
SMART_READER_LITE
LIVE PREVIEW

Feature Dis isentanglement to Aid id Im Imaging Biomarker - - PowerPoint PPT Presentation

Feb 20, 2024 238 likes •358 views

Medical Imaging with Deep Learning (MIDL) Conference, July 2020 Feature Dis isentanglement to Aid id Im Imaging Biomarker Characterization for Genetic Mutations Padmaja ja Jo Jonnala lagedda*, Brent Wei einberg, Ja Jason Alle llen, Bir

slide-1
SLIDE 1

Padmaja ja Jo Jonnala lagedda*, Brent Wei einberg, Ja Jason Alle llen, Bir ir Bhanu

*C *Center for Research in in In Intelli lligent Systems Univ iversity of

  • f Cali

alifornia, Ri Riversid ide, CA, A, USA

Feature Dis isentanglement to Aid id Im Imaging Biomarker Characterization for Genetic Mutations

Medical Imaging with Deep Learning (MIDL) Conference, July 2020

slide-2
SLIDE 2

Why? What? How?

Extract visual features of 19/20 co-gain Mutated ⇒ Higher median survival

slide-3
SLIDE 3

Challenges

  • Lack of data
  • High class imbalance
  • High inter-class similarity
  • High intra-class diversity

Training data > 80 samples per class Mutated: 31 patients Control: 135 patients

slide-4
SLIDE 4

Assessment pipeline

  • Do visual indicators

exit?

  • Classification using

multiple state-of- the-art models and validation methods

Presence

  • What are these

features?

  • Isolate and quantify

various macro- features

Characterization

  • Are these features

reproducible?

  • Use GAN to try and

recreate these indicators

Reproducibility

slide-5
SLIDE 5

Reproducibility of Biomarkers

  • If we use biomarkers to generate synthetic images, does it suggest mutation presence?
  • We propose a generative model which can tackle the following problems:
  • Limited data
  • High data diversity
  • Learns unapparent features

FeaD-GAN: Feature Disentanglement GAN

slide-6
SLIDE 6

Shape Texture Noise Shape Texture Latent Space Texture Loss Shape Loss

G

Resampling

slide-7
SLIDE 7

Results: Quantitative

RX: Representation of data (extent of features represented); ACC: Accuracy; SEN: Sensitivity; SPEC: Specificity and DIC: Dice Score. IL: Image Level; PL: Patient Level

slide-8
SLIDE 8

Results: Qualitative

slide-9
SLIDE 9

Conclusions

  • Visual indicators of mutations that correlate to median survival are

present in MRI

  • Location, texture and shape are significant indicative features
  • The features are reproducible
  • FeaD-GAN:
  • Can faithfully generate good quality images from limited dataset
  • Can capture data diversity
slide-10
SLIDE 10

Acknowledgement

This research is supported by the Bourns Endowment Fund at University of California Riverside

Thank You!