UCL Centre for Medical Image Computing

Camino and DTI-TK: DTI TK: Advanced Diffusion MRI Pipeline for Traumatic Brain Injury Gary Hui Zhang, PhD Microstructure Imaging Group Centre for Medical Image Computing Department of Computer Science University College London

26th of June, 2013

Mi Microstructure t t iimaging i with ith diffusion diff i MRI

Diffusion MRI D

Signal

Predict P

Estimate E

Diffusion MRI quantify water mobility in tissue

Histology y

Tissue Cell size, shape, density Membrane permeability Orientation distribution Axer, J. Neuro. Meth. 1999

Virtual Vi t l Histology Tissue Modeling Model parameters are the tissue microstructure feature themselves!

Pi li ffor advanced Pipeline d d diff diffusion i MRI analysis l i

Imaging g g

Inference

Localization

Normalization

Pi li ffor advanced Pipeline d d diff diffusion i MRI analysis l i

Imaging g g

Inference

Localization

Normalization

Camino: a platform for advanced diffusion MRI analysis l i • Implements a rich hierarchy of analytic models for diffusion MRI • Provides a robust framework for fitting diffusion MRI data to the models • Delivers a sophisticated simulator for validating diffusion MRI models

Monte Carlo Diffusion Simulator ((Hall and Alexander,, IEEE TMI 2009) Displacement PDF

Diffusion MR Signal

Availab ble Substtrates

Simu ulation Pipeline

Diffusion Substrate

Gamma-Distributed Radii

Crossing Cylinders

Permeable Cylinders

Mesh-based substrates

Rich hierarchyy of analytic y models of diffusion MRI (Panagiotaki et al, NeuroImage 2012)

Compartment Models Stick

Cylinder

GDRCylinders

Ball

Zeppelin

Tensor

Multi-Compartment Models

Astrosticks

Astrocylinders

Sphere

Dot

ZeppelinStickAstrosticks

Mapping axon diameter and density in the living h human b brain i with ith A ActiveAx ti A (Alexander et al, NeuroImage 2010)

Fixed tissue: • Vervet monkey • 4.7T; 140mT/m

In vivo: • human volunteer • 3T; 60mT/m

Mapping neurite orientation dispersion and d density it with ith NODDI (Zhang et al, NeuroImage 2012) NODDI

DTI Dominant Orientation

Orientation Dispersion

Fractional Anisotropy 0

1

0

Neurite Density 1

0

CSF 1

0

The acquisition protocol is simple to implement and clinically feasible.

1

Neurite density: a potential imaging marker for b i recovery (Wang et al, PLoS One 2013) brain

NODDI enables the extension of this animal model study to living human subjects.

Pi li ffor advanced Pipeline d d diff diffusion i MRI analysis l i

Imaging g g

Inference

Localization

Normalization

Diffusion MRI supports superior anatomical li t off white hit matter tt structures t t alignment DTI

T1

? Corpus Callosum O ti Radiation Optic R di ti Arcuate Fasciculus

DTI-TK provides the state-of-the-art for aligning diff i MRI d diffusion data t • Ranked the best performing tool of its kind (Wang et al, NeuroImage 2011) • Supports unbiased longitudinal analysis of diffusion MRI data (Keihaninejad et al, al NeuroImage 2013)

The importance of tensor-based alignment for l longitudinal it di l processing i (Keihaninejad et al, NeuroImage 2013)

Tensor-based alignment improves specificity

The importance of tensor-based alignment for l longitudinal it di l processing i (Keihaninejad et al, NeuroImage 2013)

Tensor-based alignment improves sensitivity

Pi li ffor advanced Pipeline d d diff diffusion i MRI analysis l i

Imaging g g

Inference

Localization

Normalization

Tract-specific p analysis y with DTI-TK ((Yushkevich et al,, NeuroImage 2008; Zhang et al, Medical Image Analysis 2010)

• Evaluate specific a priori hypotheses (e.g., ALS impairs only motor tracts) • Reduce confounding effect of neighboring structures • Present findings in the context natural to the structure

Typical Voxelwise Analysis

Tract-Specific Analysis

Summary • Camino provides a rigorous platform for • developing and validating advanced diffusion MRI methods • applying these methods to routine clinical research and practice • DTI-TK supports population-based analysis of diffusion MRI data by • implementing the state-of-the-art spatial normalization tool • delivering a statistical inference tool tailored specifically for white matter • T Together, th they th deliver d li an end-to-end dt d pipeline i li ffor advanced d d diff diffusion i MRI analysis

A k Acknowledgement l d t • Colleagues at • CMIC and MIG (UCL) • Penn Image Computing and Science Laboratory (U Penn)

• Camino funding support • EU CONNECT consortium (www.brain-connect.eu) • MS Society of Great Britain and Northern Ireland • UCLH Biomedical Research Centre funded by NIHR

• DTI-TK funding support • NIH-NIBIB R03-EB009321 • NIH-NINDS R01-NS065347