mapping ventricular expansion and its clinical correlates

26
Mapping Ventricular Expansion and its Clinical Correlates in Alzheimer's Disease and Mild Cognitive Impairment using Multi-Atlas Fluid Image Alignment & Overview Report of the 2009 SPIE Conference EunGyoung Han

Upload: others

Post on 15-Nov-2021

6 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Mapping Ventricular Expansion and its Clinical Correlates

Mapping Ventricular Expansion and its Clinical Correlates in Alzheimer's Disease and Mild Cognitive Impairment using

Multi-Atlas Fluid Image Alignment& Overview Report of the 2009 SPIE Conference

EunGyoung Han

Page 2: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

2009 SPIE Medical Imaging Conference

Page 3: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

2009 SPIE Medical Imaging Conference

Page 4: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

2009 SPIE Medical Imaging Conference

Page 5: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Sessions: Image Processing

1,11. Segmentation I-II

2. Statistical Models

3. Statistical Methods

4, 5, 10. Registration I-III

6. Motion Analysis

7. Vascular Image Processing

8. Atlas-based Methods

9. Keynote and Diffusion Tensor Imaging (Frontiers in D.I.--keynote)

SegmentationRegistrationAtlas-based MethodsStatistical Methods & ModelsDTIMicellaneous

Page 6: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Sessions: Biomedical Applications in Molecular, Structural, and Functional Imaging

1. MR Brain Imaging 2. Keynote and Neuroimaging 3. Lung 4. Blood Flow 5. Tissue Microstructure and Function 6. Motion Analysis 7. Small Animal Imaging 8. Image-based Modeling 9, 10. Mechanics I-II 11. Clinical Applications

Mr Brain and NeuroimagingLung

Blood flow and Tissue Micro-structure and Function

Mechanics

Motion Analys-isMiscellaneous

Page 7: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Sessions: Visualization, Image-guided Procedures and Modeling

1. Neuro 2, 9. Minimally Invasive I-II3. Liver4. CT Guidance5. Cardiac6. Keynote and Modeling7. Robotics and Guidance Systems8. Ultrasound10. Visualization and Geometry11. Registration

Neuro

LiverMinimally Invasive

CT GuidanceCardiac

Robotics and Guidance Sys-temsUltrasoundVisulation and GeometryRegistrationKeynote and Modeling

Page 8: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Mapping Ventricular Expansion and its Clinical Correlates in Alzheimer's Disease and Mild Cognitive Impairment

using Multi-Atlas Fluid Image Alignment(Image Processing : Registration)

Yi-Yu Chou1, Natasha Leporé1, Christina Avedissian1, Sarah K. Madsen1,Xue Hua1, Clifford R. Jack, Jr. 2, Michael W. Weiner3, Arthur W. Toga1,

Paul M. Thompson1, and the Alzheimer's Disease Neuroimaging Initiative

1 Laboratory of Neuro Imaging, UCLA Department of Neurology, Los Angeles, CA, USA

2 Mayo Clinic College of Medicine, Rochester, MN

3 Depts. Radiology, Medicine & Psychiatry, UC San Francisco, San Francisco, CA

Alzheimer's Disease Neuroimaging Initiative (ADNI)

Alzheimer's Disease

Page 9: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Alzheimer's Disease

AD affecting 5~10% over age 6530~40% over age 906~25% of MCI subjects per year transition to AD

Testing subjects - 80 AD patients - 80 individuals with MCI - 80 healthy subjects

Page 10: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Problem Statements and their AnswersDo ventricular measures show a relatively high effect size in

distinguishing disease from normality?• Ventricular expansion appears to provide the greatest sensitivity

as a quantitative marker of disease progression in Alzheimer's Disease (AD) in serial MRI studies.

How can we quantify the factors affecting progression from Mild Cognitive Impairment (MCI) to AD or normal aging to AD?

• By developing automated brain mapping techniques to map and analyze lateral ventricular expansion.

• By discovering which one of these automated techniques is optimal for such a task.

Result is that we should now be able to detect which therapeutic factors may help patients resist neurodegeneration in drug trials.

Page 11: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Automated Lateral Ventricle Segmentation and Shape Modeling

- Lateral ventricular volumes automatically estimated for scans using a “multi-atlas” technique

Method pipeline:1) Map multiple surface-based atlases into each scan via fluid registration2) Combine multiple segmentations of the same scan into a single average

surface mesh3) Randomly choose image samples and manually trace the lateral

ventricles in contiguous coronal brain sections4) Convert lateral ventricular surface into parametric meshes5) Do fluid registration of each atlas and the embedded mesh models to all

other subjects, treating the deforming images as Navier-Stokes viscous fluid, thereby guaranteeing a diffeomorphic mapping.

6) Apply fluid transforms to the manually traced ventricular boundary using tri-linear interpolation, generating a propagated contour on the unlabeled images

7) Match grid-points from corresponding surfaces across subjects to obtain group average parametric meshes

Page 12: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

ParametricSurface

Map

Methods Flowchart

Multiple surface meshes are

mapped into new subjects’

scans via fluid registration. N

Page 13: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Methods Flowchart

Medial curve

- 3D curve traced out by the centroid of the ventricular boundary

The medial curve defined in each individual before averaging the surfaces.

Measure radial ventricular expansion in each individual

Plot the resulting statistics on the average surface

Page 14: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Ventricular Statistical Maps and Analysis Surface contractions and expansions were compared between groups at

equivalent locations using Student's t-tests with 2-tails, and were correlate with different clinical characteristics including diagnosis, cognitive scores, ApoE genotype, clinical scores, and future decline.

CDF plots of p-values determined the method's statistical power for finding links between morphology and different disease measures.

p-value : describes the uncorrected significance of group differences, plotted onto the average surface model as a color-coded map

q-value : gives single overall measure of significance for each p-map. If the q-value DNE, then there is insufficient evidence to reject null hypothesis.

Page 15: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Ventricular Statistical Maps and Analysis

Multi-statistical testCDF plot intersects with the y=20x line ==> the highest values for which at

most 5% false positive are expected in the map.

==> observed p-values are limited to the [0 0.05]

q-value (intersection point CDF and y=20x) ==> single overall measure of significance for each p-map.

no intersection point => no evidence to reject the null hypothesis

To assign overall significance values to each statistical map use false discovery rate (FDR) based on the expected proportions of voxels with intensity above the threshold under the null hypothesis.

Page 16: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Linking Ventricular Morphology and Clinical Characteristics

Significance maps map correlations between local ventricular enlargement and (1) diagnosis (MCI vs. normal, AD vs. normal and AD vs. MCI); (2) cognitive scores (MMSE, Global clinical dementia rate (CDR), and sum of Boxes CDR); (3) clinical depression scores, (4) ApoE genotype and (5) educational level

Page 17: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Linking Ventricular Morphology and Clinical Characteristics

CDFs of significance maps associating ventricular enlargement with diagnosis and clinical measures.

Based on FDR q-values, the AD vs. control and MCI vs. control contrast are significant, as are the links between ventricular dilation (expansion) and (1) MMSE scores, and (2) depression.

Page 18: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Predicting Future Cognitive Change

How do changes detected by brain imaging predict future clinical decline?

=>Their experiments correlated baseline ventricular morphology with subsequent change over 1 year in MMSE, global clinical dementia rate (CDR), sum of boxes CDR scores

Page 19: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Predicting Future Cognitive Change

Significance maps correlate baseline ventricular shape with subsequent decline, over the following year in 3 commonly used clinical scores

FDR analysis of future changes. Correlations were significant between baseline ventricular enlargement and future change in MMSE, Global CDR and Sum of Boxes scores.

Page 20: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Minimal Effective Sample Sizes

How many subjects would suffice to detect statistically significant correlations of ventricular enlargement with diagnosis and with clinical test scores?

==> Randomly throw out subjects until left with a sample of size N Help estimate sample sizes with adequate power to detect

differences between groups, optimizing cost-effectiveness in future trials

40 subjects sufficient to discriminate AD from normal

Page 21: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Effects of Varying the Sample Size

60 subjects required to correlate ventricular enlargement with MMSE

119 subjects required to correlate ventricular enlargement with clinical depression

Page 22: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

ConclusionSurface-based statistical maps => Revels correlations between surface ventricular morphology at baseline and diagnosis, cognitive performance (MMSE scores), depression, and predicted future decline (over a 1year interval) in 3 standard clinical scores (MMSE, global and sum-of boxes CDR)

Surface-based false discovery rate (FDR), along with multi-atlas fluid registration => reduce segmentation error => allow researches to estimate sample sizes with adequate power to detect groups differences => compare the power of mapping methods head-to-head optimizing cost effectiveness for future clinical trials.

Surface averaging within subjects => reduced segmentation error

Page 23: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Conclusion

FDR method ==> estimate minimal sample sizes for several disease tracking hypotheses

Page 24: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Automatic segmentation of cortical vessels in pre- and post-resection laser range images

S. Ding, M. I. Miga, R. C. Thompson, I. Garg, B. M. Dawant,Vanderbilt Univ. (United States)

Measurement of intra-operative cortical brain movement is necessary to drive mechanical models developed to predict sub-cortical shift. At our institution, this is done with a tracked laser range scanner. This device acquires both 3D range data and 2D photographic images. 3D cortical brain movement can be estimated if 2D photographic images acquired over time can be registered. Previously, we have developed a method, which permits this registration using vessels visible in the images. But, vessel segmentation required the localization of starting and ending points for each vessel segment. Here, we propose a method, which automates the process. This method involves several steps: (1) correction of lighting artifacts, (2) vessel enhancement, and (3) vessels’ centerline extraction. Result obtained on 4 images obtained in the operating room show that our method is robust and is able to segment vessels reliably.

Page 25: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Group-wise registration of large image dataset by hierarchical clustering and alignmentQ. Wang, L. Chen, Shanghai Jiao Tong Univ. (China);

D. Shen, The Univ. of North Carolina at Chapel Hill (United States)

Group-wise registration has been proposed recently for consistent registration of all images within a group and producing an unbiased atlas. Since all images are registered with lots of parameters to be optimized simultaneously, the number of images that the existing group-wise registration methods can handle is limited due to CPU and memory. To overcome this limitation, we present a hierarchical group-wise registration method for feasible registration of large image dataset. Our basic idea is to decompose the large-scale group-wise registration problem into a series of small-scale problems, which are easy to solve individually. We particularly use the affinity propagation method to hierarchically cluster images into a pyramid of classes. Then, images in the same class are group-wisely registered and their center images are produced. Those center images of different classes, which represent the corresponding classes, are registered from the lower level of the pyramid to the upper level. A single atlas for the whole image dataset is finally produced when the registration reaches the top level of the pyramid. By using these hierarchical clustering and atlas synthesis steps, we can efficiently and effectively register large image dataset, estimate an unbiased atlas, and map each subject image to the atlas space. We have presented the experimental results on both real and simulated data, and demonstrated that our method can achieve better robustness and registration accuracy than conventional methods.

Page 26: Mapping Ventricular Expansion and its Clinical Correlates

Scientific Computing and Imaging Institute, University of UtahScientific Computing and Imaging Institute, University of Utah

Thank you!

Questions?

Thank you for your attention