SLIDE 1
Neuroimaging in Multiple Sclerosis
Kristin A. Linn
Post-doctoral researcher, University of Pennsylvania
Personal Background
2004 - 2008 2009 - 2014 2014 - 2017
Recurse Center
Neuroimaging in Multiple Sclerosis Kristin A. Linn Post-doctoral - - PowerPoint PPT Presentation
Neuroimaging in Multiple Sclerosis Kristin A. Linn Post-doctoral - - PowerPoint PPT Presentation
Neuroimaging in Multiple Sclerosis Kristin A. Linn Post-doctoral researcher, University of Pennsylvania Personal Background 2004 - 2008 2009 - 2014 2014 - 2017 Recurse Center Structural MRI 3D array of voxels 2D Grayscale
SLIDE 2
SLIDE 3
SLIDE 4
Structural MRI
Axial slice of a T1 image
- 3D array of voxels
- Grayscale intensities provide
contrast between different tissue types
- Standard size approximately
256 x 256 x 176 voxels
pixel voxel 2D 3D
SLIDE 5
Structural MRI in Multiple Sclerosis
- Used to study lesion accumulation in brain and spinal cord
Inflammation
MS lesion dynamics
SLIDE 6
Structural MRI in Multiple Sclerosis
- Used to study lesion accumulation in brain and spinal cord
Tissue damage and demylenation Inflammation
MS lesion dynamics
SLIDE 7
Structural MRI in Multiple Sclerosis
- Used to study lesion accumulation in brain and spinal cord
Tissue repair and remylenation Tissue damage and demylenation Inflammation
MS lesion dynamics
SLIDE 8
Structural MRI in Multiple Sclerosis
Tissue repair and remylenation Tissue damage and demylenation
MS lesion dynamics
Inflammation
- Used to study lesion accumulation in brain and spinal cord
SLIDE 9
Quantifying MS Treatment Efficacy
- Do chronic lesions show evidence of repair in response to
treatment?
Pre-treatment Post-treatment
SLIDE 10
Quantifying MS Treatment Efficacy
- Do chronic lesions show evidence of repair in response to
treatment?
Pre-treatment Post-treatment
- Initial comparison of average chronic lesion intensity before and
after treatment was highly significant
SLIDE 11
- Do chronic lesions show evidence of repair in response to
treatment?
Pre-treatment Post-treatment
- Statistical challenges: confounding, automatic lesion
segmentation, image preprocessing
Quantifying MS Treatment Efficacy
SLIDE 12
Preprocessing README
- 12 steps before analysis
- Current analysis is a t-test!
- Each induces errors/uncertainty
that is not easy to account for
- Typically, researchers ignore
preprocessing uncertainty
SLIDE 13
Registration Error
- Image alignment sometimes off by one slice
Pre- treatment: Post- treatment:
SLIDE 14
Registration Error
Pre- treatment: Post- treatment: Leads to bias in favor of a treatment effect!
SLIDE 15
Registration Error
Pre- treatment: Post- treatment:
SLIDE 16
Registration Error
Pre- treatment: Post- treatment:
SLIDE 17
Moving Forward
- Many opportunities for statisticians to contribute in meaningful ways
to neuroimaging research
- Need more sophisticated methods to account for preprocessing errors
SLIDE 18
Moving Forward
- Many opportunities for statisticians to contribute in meaningful ways
to neuroimaging research
- Need more sophisticated methods to account for preprocessing errors
- Steep learning curve (anatomy, physics, engineering)
SLIDE 19
Moving Forward
- Many opportunities for statisticians to contribute in meaningful ways
to neuroimaging research
- Need more sophisticated methods to account for preprocessing errors
- Steep learning curve (anatomy, physics, engineering)
- Any methods/models we build must be fast with high-quality, well-
tested software
SLIDE 20
Acknowledgements
Taki Shinohara, Ph.D.
Imaging and clinical collaborators: Post-doctoral Advisor:
Salim Chahin, M.D. (Penn) Gabrial Pilar, M.D. (Penn) Clyde Markowitz, M.D. (Penn) Matthew Schindler, M.D. (NIH) Emily Acton (Penn)
SLIDE 21