SLIDE 1
Research area
2
Ninon Burgos CNRS Researcher Brain and Spine Institute Aramis Lab - - PowerPoint PPT Presentation
Ninon Burgos CNRS Researcher Brain and Spine Institute Aramis Lab - - PowerPoint PPT Presentation
Ninon Burgos CNRS Researcher Brain and Spine Institute Aramis Lab Paris, France 2 Research area Decision Image Medical support processing imaging systems Translational research 3 PhD University College London Combination of
SLIDE 2
SLIDE 3
PhD – University College London
Combination of structural and functional imaging modalities
3
PET/MRI PET/CT
CT PET MRI
SLIDE 4
PhD – University College London
4
Image synthesis for the attenuation correction of PET/MR data
Objective
▷ Synthesise CT from MR images to correct for attenuation PET/MR data
- e-
e+
γ (511 keV) γ (511 keV) 180°
MRI CT PET without attenuation correction PET with attenuation correction Attenuation
Clinical motivation
▷ Imperfect attenuation correction on PET/MR scanners
- γ
γ
SLIDE 5
5
PhD – University College London
Method
[Burgos et al., MICCAI, 2013], [Burgos et al., IEEE TMI, 2014], [Burgos et al., EJNMMI, 2015]
▷ Multi-atlas registration, propagation and fusion
... ... ... ...
Affine alignment Resampling Target MRI MRI-CT database Mapped MRIs & CTs
LSIM1 LSIM2 LSIMn
Intensity fusion
W1 W2 Wn
Target CT Inter-subject mapping
Image synthesis for the attenuation correction of PET/MR data
SLIDE 6
PhD – University College London
Evaluation
▷ Validation on FDG and Florbetapir (Aß) PET images: less than 2% difference ▷ Independent multi-centre evaluation study: joint best performance ▷ Evaluation using low resolution MR images in collaboration with University Hospital Zurich and GE Healthcare
6
[Burgos et al., EJNMMI, 2015] [Ladefoged et al., NeuroImage, 2017] [Sekine, Burgos et al., JNM, 2016]
Image synthesis for the attenuation correction of PET/MR data
With real CT Proposed
CT FDG PET Aß PET CT FDG PET Aß PET
SLIDE 7
Technology development
▷ NiftySeg Method part of a software package for image segmentation and synthesis ▷ NiftyWeb Method available online as a web service tool for testing, already used 1900+ times
Transfer to clinical research
▷ Method routinely used on several research studies at the Dementia Research Centre (UCL)
Transfer of technology
▷ Licensing agreement recently signed between Oncovision and UCL
7
https://github.com/KCL- BMEIS/NiftySeg http://niftyweb.cs.ucl.ac.uk/ program.php?p=PCT [Weston et al., Alzheimer’s & Dementia: Diagnosis and Disease Monitoring, 2016] [Lane et al., BMC Neurology, 2017]
PhD – University College London
Image synthesis for the attenuation correction of PET/MR data
SLIDE 8
Current & future work
8
▷Neuroimaging plays a crucial role in the understanding, diagnosis, and treatment of neurological disorders ▷Challenge: joint analysis of multiple imaging modalities at several time points Improve the analysis of multimodal neuroimaging data, and thus improve the understanding and diagnosis of neurodegenerative diseases
Clinical motivation
SLIDE 9
Current work
Automatic identification of abnormality patterns
9
Objective
▷ Locate and characterise patterns of abnormality from multimodal imaging data
... ... MRI-PET control dataset Subject-specific PET model
- f healthy appearance
Subject’s real PET Subject-specific abnormality map Subject’s MRI
Method
[Burgos et al., MICCAI, 2015], [Burgos et al., CMMI, 2017], [Burgos et al., Submitted]
▷ Subject-specific model of healthy appearance ▷ Subject-specific map of abnormality
SLIDE 10
Current work
Automatic identification of abnormality patterns
10
Healthy control
L
Alzheimer’s disease
Z-score SUVR 1 2
- 3
3
Semantic dementia Frontotemporal dementia
PET Abnormality map PET Abnormality map PET Abnormality map PET Abnormality map
SLIDE 11
Current work
Automatic identification of abnormality patterns
11
Visualisation tool Improved interpretability of subsequent analyses
Alzheimer’s disease Semantic dementia Frontotemporal dementia
Alzheimer’s disease