in vivo bone characterization from magnetic...

42
In vivo Bone Characterization from Magnetic Resonance Imaging: Morphometry and Mechanical analysis Angel Alberich-Bayarri, PhD [email protected] 15 th october 2014

Upload: buicong

Post on 04-Oct-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

In vivo Bone Characterization from

Magnetic Resonance Imaging:

Morphometry and Mechanical analysis

Angel Alberich-Bayarri, PhD

[email protected]

15th october 2014

La Fe Polytechnics and University Hospital

• Valencia

• East of Spain

• 1000 beds

• Medical Imaging• 7 CT scanners

• 5 MR scanners

• 1 PET/CT

• 1 SPECT/CT

• Research Group & Lab• 1 MR scanner

• 1 angio suite

Introduction to the problem

• Approx. 200 million patients worldwide.

• Prevalence (> 50 y.o.):• Women: 33%

• Men: 8%

• Usually diagnosed in advanced stages.

• Asymptomatic progression.

• First symptom that usually appears: bone

fracture.

- Reginster JY, Burlet N. Bone 2006;38:S4-S9.

- Cummings SR, Melton III JR. Lancet 2002;359:1761–1767.

Introduction to the problem

• Estimation of 2.6 millions of new fractures by 2025.

• NIH definition 2001: “Skeletal disease characterized

by a decreased bone resistance that induces a higher

probability of suffering a fracture in a specific person”.

Fracturas de muñeca

Fracturas

vertebrales

Fracturas de cadera

-Gullberg B, Johnell O, Kanis JA. Osteoporosis International 1997;7:407–413.

Introduction to the problem

Bone resistance

decrease

Bone Mineral

Density

Trabecular

Microarchitecture

Introduction to the problem

Stage DEXA T-score

Healthy > -1.00 SD

Osteopenia -2.50 SD to -1.00 SD

Osteoporosis < -2.50 SD

- WHO technical report series 843, Geneva, Switzerland; 1994.

- Wehrli FW, et al. Topics in Magnetic Resonance Imaging 2002;13:335–355.

• Diagnosis: Dual Energy X-Ray Absorptiometry (DEXA)

• WHO Criteria

• BMD explains around 60% of bone fractures

Introduction to the problem

• There is a need to characterize

bone microarchitecture

• Complement in the diagnosis

process

• Therapy efficacy follow-up

• Quantitative properties of shape,

orientation, topology, tortuosity,

mechanical resistance.

• We must develop and find new

tools to characterize bone quality.

How to obtain bone images

• Radius distal metaphysis (non-dominant)

• 3 Tesla MR (Philips Healthcare)

• Gradients: • 80mT/m (Gmáx), 200mT/m/ms (SR)

• 4-channel surface coil

How to obtain bone images

• Acquisition mode: 3D

• Sequence type: Gradient-Echo T1 (Spoiled)

• TE/TR/α: 5ms/16ms/25º

• 60 axial partitions

• Voxel size: 180 x 180 x 180 µm3

• Parallel Im. Factor: 2/0 (plane/slices)

• Bandwidth: 167 Hz/pix

• Matrix: 512 x 512

• Number of examinations: 3

• Total time: 5’42’’

Processing the images

• Automated• Range filters

• Average filters

• Thresholding

• Adaptative contours

• Semi-automated• Rectangular ROI

• Propagation through slices

Processing the images

• Heterogeneities correction. Bone Volume Fraction

Map (BVF)

• Laplacian based local thresholding of intensities

ILr

~

Original

segmented image

BVF map

- Vasilic B, Wehrli FW. IEEE Transactions on Medical Imaging 2005;24:1574-1585.

Processing the images

• Increase in spatial resolution

• 180 x 180 x 180 µm → 90 x 90 x 90 µm

SuperresolutionBVF map

José V. Manjón, Pierrick Coupe, Antonio Buades, Louis Collins and Montserrat Robles. MRI Superresolution Using Self

Similarity and Image Priors. International Journal of Biomedical Imaging, Article ID 425891, 2010

Processing the images

• Otsu’s thresholding

• Intra-class variance minimization

BinarizationSuperresolution

T

B

tt

2

2

max.arg*

Processing the images

• Pipeline

Binarization Superresolution

Original segmented BVF map

Processing the images

• 3D

• Triangulated surface extraction

• STL exportation routine

Morphometry analysis

• Morphology

Bone volume / total volume percentage BV/TV [%]

Mean Trabecular Thickness [µm]

Mean Trabecular Separation [µm]Mean number of adjacent marrow voxels

Trabecular index

voxelstotaln

voxelsbonenTVBV

__

__/

c

Sk

d

ThTbc

Skp

·

·2

.

min

ThTb

TVBVNTb

.

/.

Morphometry analysis

• Morphology

- Alberich-Bayarri et al. AJR 2008: 191:721-726.

Morphometry analysis

• Irregularities analysis• 2D Fractal Dimension, D2D

• Contour detection

• Box-counting

• Slice-by-slice

• Validation

kDN D )·log()log( 2

Morphometry analysis

• Irregularities analysis• 3D Fractal Dimension, D3D

• Surface detection

• Validation

kDN D )·log()log( 3

- Alberich-Bayarri et al. Medical Physics 2010;37:4930-4937.

Morphometry analysis

• Irregularities analysis• 2D and 3D Fractal Dimension results

- Alberich-Bayarri et al. Medical Physics 2010;37:4930-4937.

Morphometry analysis

• Anisotropy analysis

• Mean Intercept Length

• Tensor Scale

30

210

60

240

90

270

120

300

150

330

180 0

ab

θa

DA=a/b

- Inglis D, Pietruszczak S. International Journal of Solids and Structures 2003;40:1243-1264.

- Saha PK, Wehrli FW. Pattern Recognition 2004;37:1935–1944.

Mechanical analysis

• Meshing algorithm

• Voxelwise

• Hexahedron elements (brick)

• Nodes and elements definition

• Approx: 1 million of elements

Mechanical analysis

• Model generation

• Properties definition• Elastic modulus: Eb=10 GPa (Linear, elastic, isotropic)

• Poisson coefficient: 0.3=ט

• Exportation routines generation• Ansys (Ansys Inc., USA)

• Abaqus (Dassault Systèmes SIMULIA Inc., USA)

• Salome (CAELINUX)

• OpenFOAM

- Ladd AJ, Kinney JH. Journal of Biomechanics 1998;31:941-945.

- Fung YC. Biomechanics. Mechanical properties of living tissues (2nd edition). New York, NY: Springer, 1993.

- Newitt DC, Majumdar S, van Rietbergen B, et al. Osteoporosis International 2002;13:6-17.

Mechanical analysis

• Compression essay• X, Y, Z

• Deformación conocida

• Definición de sistema de ecuaciones

fuK ·K: stiffness matrix

u: displacements vector

f: forces vector

Mechanical analysis

• Compression essay• Solution calculation

• Sparse Gaussian Elimination

• Non-linearities: infinite displacements• Generation of a routine previous to the

generation of an ANSYS exportation file.

fuK ·

eV

Te voldDBBK )(·

2/)1(00

01

01

1 2

v

v

v

v

ED e

ijij KK

FE model

generation

Isolated elements

controlExportable file

Mechanical analysis

• Compression essay• Young’s modulus calculation.

• Homogenization theory

• X, Y, Z

- Hollister SJ, Fyhrie DP, Jepsen KJ, Goldstein SA. Journal of Biomechanics 1991;24:825–839.

- Hollister SJ, Kikuchi N. Computational Mechanics 1992;10:73–95.

n

app FA

E

1

Mechanical analysis

Pa

Pa

• 3D nodal stresses [Pa]

Mechanical analysis

• 3D nodal stresses [Pa]

Pa

Control

Osteoporosis

Trabecular bone analysis

• Influence of the ROI

- Alberich-Bayarri et al. AJR 2008: 191:721-726.

Trabecular bone analysis

• Reproducibility and accuracy

• The goal was to evaluate the accuracy and

reproducibility of the morphometry and mechanical

trabecular bone characterization from MR.

• 5 sheep extremities (tibia proximal metaphysis)

• Reproducibility: 3 MR acquisitions of each sample

(1 per day)

• Accuracy: 1 µCT acquisition per sample (Gold

Standard)

Trabecular bone analysis

• Reproducibility and accuracy• 3 Tesla MR (Philips Healthcare)

• 4-channel surface coil

• Sequence: 3D GE T1 (Spoiled)

• TE/TR/α: 5ms/16ms/25º

• 60 partitions

• Voxel size: 180 x 180 x 180 µm3

Trabecular bone analysis

• Reproducibility and accuracy

• Accuracy: µCT acquisition• eXplore Locus SP, General Electric, USA.

• Spatial resolution (7,5 x 7,5 x 7,5 µm).

• Gold standard

Trabecular bone analysis

• Reproducibility and accuracy

• All parameters showed a high reproducibility,

with variation coefficients below 10%.

BVTV TbTh TbSp TbN D2D D3D Ex Ey Ez

p 0.80 0.60 0.28 0.86 0.54 0.60 0.29 0.36 0.41

RMS

CoV [%]6.1 1.3 4.6 5.1 2.2 1.8 8.01 8.64 8.31

- Alberich-Bayarri A et al.. Titlel Radiologia,

Trabecular bone analysis

• Reproducibility and accuracy

• Only Tb.Th, Tb.Sp y Tb.N showed significant

differences when compared to μCT (p<0.05).

BV/TV

[%]

Tb.Th

[mm]

Tb.Sp

[mm]

Tb.N

[mm-1]D2D D3D

Ex

[MPa]

Ey

[Mpa]

Ez

[Mpa]

MR 0.29 0.19 0.73 1.52 1.69 2.33 244.68 283.94 347.74

µCT 0.30 0.13 0.53 2.42 1.69 2.44 269.70 307.77 408.97

p

value0.83 0.001 0.02 0.01 0.91 0.20 0.94 0.84 0.70

Error

[%]3.3 46.7 37.7 37.2 0.2 4.5 9.3 7.7 14.9

Trabecular bone analysis

• Reproducibility and accuracy

Trabecular bone analysis

• But…

• Too many parameters?

• Clinical confusion due to an excess in the number

of variables

• Solution: Mechanical Competence of Bone (MCP)

unified parameter!

- Alberich-Bayarri et al. IEEE Trans Biomed Engineering 2013: 60(5):1363-70.

Final

Workflow

Report

Bone analysis

• Work in progress: Ultra-Short Echo Time

• Water molecules in small pores have a really

short TE (below ms)

Acknowledgements:

The developments and results shown on bone

characterization are partialy funded by the Ministry

of Economy and Competitiveness of Spain through

Project DPI2011-22413. Design for Osteoporosis

(DFO).

Principal Investigator: María Ángeles Pérez

(University of Zaragoza)

Acknowledgements:

• Enrique Ruiz

• Jose Tomás

• Celia Juan

• Beatriz Dionisio

• Raquel Nombela

• Ana Penadés

• Ángel Alberich-Bayarri

• Luis Martí-Bonmatí

OBRIGADO!

Angel Alberich-Bayarri, PhD

[email protected]

15th october 2014