|
|
|
European Biopharmaceutical Review
|
Vascular disease is a leading cause of death and disability worldwide.
Early identification of plaques most at risk of causing disruption would
enable prompt intervention, and prevent irreversible damage or death
from stroke (1).
Effective biomarkers are needed for the
assessment of atherosclerotic plaque vulnerability, as well as for
prognostic and longitudinal use. The data they provide may also reduce
the cost and time required to develop novel pharmaceuticals and
therapeutics, or serve as important end-points for clinical benefit
(2,3). In all phases of drug development, imaging biomarkers may be used
to select or stratify patients based on disease status, which can
better demonstrate their therapeutic effect.
While considerable
research has been applied to optimise the information content of
imaging, there is a corresponding need and opportunity for more
sophisticated image analysis techniques to exploit this area.
Computer-aided prognosis (CAP) is a new and exciting complement to
computer-aided diagnosis (CAD), involving the development and
application of image analysis and multimodal data fusion to help
physicians.
Diagnostic Techniques
Over the past 20
years, a number of diagnostic techniques for detecting the hallmarks of
vulnerable plaque have been proposed; the goal is generally to identify
signatures for pathogenic mechanisms, which contribute to plaques that
would experience adverse natural histories unless strategic intervention
takes place (4,5).
Carotid angiography is an invasive technique
which uses contrast medium to assess the degree of luminal stenosis and
some features of the artery wall, such as ulceration and mural thrombus
(6). Intravascular ultrasound enables close proximity of the transducer
to the arterial wall, producing high-resolution images from which
calculations of plaque area, extent of arterial remodelling and
differentiation of plaque components can be made (7). Many other
modalities – including the use of special formulations of contrast
agents and specialised hardware approaches in magnetic resonance (MR) –
have also been attempted (8-11). Positron emission tomography has been
proposed, but the requirements of radiolabelled compound, low resolution
and unclear specificity seem to limit this modality to research-only
purposes (12).
There is growing evidence that MR is ideally
suited to the task of discriminating relevant tissue characteristics to
expand on the structural measurements. We further posit that dependence
on specially formulated molecular-targeted contrast agents, or on
special-purpose hardware to achieve high resolution, may limit
widespread adoption and, in fact, be completely unnecessary.
If a
readily available tool, designed to allow a more detailed
characterisation of plaques, were to be developed and installed into the
existing base of imaging systems, the industry could benefit greatly.
We describe our progress towards this objective in this article.
MR Potential
Results
have been published which demonstrate that contrast uptake measurements
reflect inflammation in the tunicae intima, media, adventitia and
neovascularization of the vascular wall (13). Identification of outward
remodelling of the vessel wall – also called positive remodelling – uses
MR images from which blood in the lumen and the vessel wall can be
visualised, thereby allowing quantitative comparison of the luminal and
vessel areas. Positive remodelling is predominant over vulnerable
plaques and negative remodelling – at the expense of the luminal
diameter – in stable plaques (14).
Since vulnerable plaques tend
to exhibit higher permeability of the neovasculature in abundance, a
contrast agent is ideally used. We focus on the kinetics of a contrast
agent entering (wash-in) and exiting (wash-out) the vessel wall, due to
increased endothelial permeability, neovascularisation and necrosis.
Features
that are associated with increased gadolinium diethylenetriamine-penta
acid (Gd-DTPA) uptake include tissue necrosis (increased distribution
volume due to leaky membranes and dead cells); inflammation associated
with increased vessel size and number, as well as increased endothelial
permeability; and fibrosis/edema. Vessel wall pathologic lesions – which
are seen by discernible enhancement patterns after the administration
of Gd-DTPA – measure neo-vascularisation and inflammation in the regions
of vulnerable plaques, as opposed to stable plaques.
Figure 1
illustrates some of the biophysical hallmarks that may be identified
using MR data, provided by sophisticated image processing techniques.
Moreover, the use of these methods in both animal and human models
suggests a unique value in translational studies.
Image Processing
Multivariate
quantitative descriptors are developed using controlled outcome studies
on a specialised model. Localised pixel statistics – as well as second
order co-occurrence textural features, capable of discriminating
functional microvascular networks without requiring resolution beyond
those that are obtained in standard magnetic resonance imaging (MRI) –
are optimised to find signatures consistent with characteristic patterns
of lipids which have been intermixed with extracellular matrix fibres
and necrotic tissue. Neovascularisation and increased tissue
permeability, leading to increased distribution volume, are evaluated by
fitting cubic polynomials to the contrast media uptake signal to
provide effective class separation. This process is described in Figure
2.
Various image filters may be computed on the wall area to
produce texture features. The initial filters interrogate the first
order spatial intensity variations and statistics within small
neighbourhoods. Evaluating first order statistics can allow for the
capture of local, spatially proximal textural changes – for example,
microchanges over time. These may be subtle, local changes in the
contrast uptake that differ between the types of lesions being studied.
The
second two classes of features – Sobel and Kirsch – are gradient/edge
filters which are similar to the first order statistics, but instead
look specifically at changes in particular directions. A further
deterioration in micro-architecture of the vessel may occur with
increasing vulnerability and acuteness of the pathologic process. These
changes may be captured with the Sobel and Kirsch filters.
Haralick
features can capture the co-occurrence of image intensities in higher
order (second order) texture features, which can be proximal to each
other. Steerable Gabor features have been modelled on the patterning of
the human visual cortex and have found widespread application in image
analysis (18,19).
The most relevant feature highlights clusters
relating to biophysical hallmarks of disease (see Figure 3). For
example, certain features relate to spatial heterogeneity in the tissue
that are consistent with lipid patterning, this time intermixed with
inflammatory cells, as well as extracellular matrix fibres and necrotic
tissue. The clusters in the kinetic features are consistent with
enhanced endothelial permeability, neovascularisation, necrosis and
collagen breakdown. The high cluster indices are associated with
morphological measurements, indicating remodelling and wall area
abnormalities.
Per-Voxel Classification
After
evaluating each individual feature’s performance, some may be combined
to build multi-feature classifiers for tissue characterisation. Ensemble
learning methods may be used for classification by constructing
multiple decision trees, based on the training set and outputting of the
class (see Figure 4) (20). It is important to note the clear break
point that occurs when 15 features are used versus fewer in this
particular example. Across the board, we find the classification to be
highly specific for most values of sensitivity. However, adding features
improves the sensitivity at operating points where the test is very
precise.
Conclusion
Having conducted detailed
research, it is now possible to assess the system output using a visual
and quantitative scoring system. In this project, we focused on
considering whether or not a thrombus was formed – identifying plaques
which did form one as vulnerable, and those that did not as stable.
Thus,
to solve this two-sided problem, we are able to present test cases and
obtain, as output, the likelihood that the plaque is in one class or
another. We can then create a score that reflects the biophysical
reality of which there are reasons to believe the outcome, but there is
still a quantitative estimate of which is more likely to follow from the
class that obtains more ‘votes’.
We conclude that automated
registration and segmentation of arterial plaques, as well as the
application of structural measurement and functional feature extraction –
for discriminating between tissue types that are consistent with stable
and vulnerable plaques – is indeed feasible.
Acknowledgement
This
material is based upon work supported by the National Science
Foundation under Grant No. 1248316. Any opinions, findings and
conclusions or recommendations expressed in this material are those of
the authors and do not necessarily reflect the views of the National
Science Foundation.
References
1. Buckler AJ et al, Non-invasive theranostic to predict and assess response to atherosclerotic drugs, Experimental Biology, 2013
2. Woodcock J and Woosley R, The FDA critical path initiative and its influence on new drug development, Annu Rev Med 59: pp1-12, 2008
3. Buckler AJ et al, Quantitative imaging test approval and biomarker qualification: interrelated but distinct activities, Radiology 259(3): pp875-884, 2011
4. Huizenga JT et al, Automated methods and systems for vascular plaque detection and analysis, IP, WO/2005/02090
5. Eijgelaar WJ, Heeneman S and Daemen MJ, The vulnerable patient: refocusing on the plaque? Thromb Haemost 102(2): pp231-239, 2009
6. Schoenhagen P et al, Extent and direction of arterial remodeling in stable versus unstable coronary syndromes: an intravascular ultrasound study, Circulation 101(6): pp598-603, 2000
7. Varghese et al, Characterization of vulnerable plaque using dynamic analysis, IPW, A1
8. Ronald JA et al,
Enzyme-sensitive magnetic resonance imaging targeting myeloperoxidase
identifies active inflammation in experimental rabbit atherosclerotic
plaques, Circulation 120(7): pp592-599, 2009
9. Kim, SnE et al,
Diffusion-weighted imaging of human carotid artery using 2D single-shot
interleaved multi-slice inner volume diffusion-weighted echo planar
imaging (2D ss-IMIVDWEPI) at 3T: diffusion measurement in
atherosclerotic plaque, Journal of Magnetic Resonance Imaging 30(5): pp1,068-1,077, 2009
10. Oikawa M et al,
Carotid magnetic resonance imaging, a window to study atherosclerosis
and identify high-risk plaques, Circ J 73(10): pp1,765-1,773, 2009
11. Li F et al,
Scan-rescan reproducibility of carotid atherosclerotic plaque
morphology and tissue composition measurements using multicontrast MRI
at 3T, J Magn Reson Imaging 31(1): pp168-176, 2010
12. Joshi N et al,
18F-fluoride positron emission tomography for identification of
ruptured and high-risk coronary atherosclerotic plaques: a prospective
clinical trial, The Lancet, 2013
13. Phinikaridou A et al, A robust rabbit model of human atherosclerosis and atherothrombosis, J Lipid Res 50(5): pp787-797, 2009
14. Phinikaridou A, In vivo detection of vulnerable atherosclerotic plaque by MRI in a rabbit model, Circ Cardiovasc Imaging 3(3): pp323-332, 2010
15. Kramer CM, Magnetic resonance imaging identifies the fibrous cap in atherosclerotic abdominal aortic aneurysm, Circulation 109(8): pp1,016-1,012, 2004
16. Maintz D et al, Selective coronary artery plaque visualization and differentiation by contrast-enhanced inversion prepared MRI, Eur Heart J 27(14): pp1,732-1,736, 2006
17. Yeon SB et al,
Delayed-enhancement cardiovascular magnetic resonance coronary artery
wall imaging: comparison with multislice computed tomography and
quantitative coronary angiography, J Am Coll Cardiol 50(5): pp441-447, 2007
18. Jain A and Farrokhnia F, Unsupervised texture segmentation using gabor filters, Pattern Recognition 24(12): pp1,167-1,186, 1991
19. Kruizinga P and Petkov N, Nonlinear operator for oriented texture, IEEE Transactions on Image Processing 8(10): pp1,395-1,407, 1999
20.
Tiwari P, Rosen M and Madabhushi A, A hierarchical spectral clustering
and nonlinear dimensionality reduction scheme for detection of prostate
cancer from magnetic resonance spectroscopy (MRS), Medical Physics 36(9): pp3,927-3,939, 2009 |
Read full article from PDF >>
|
|
 |
|
| Rate this article |
You must be a member of the site to make a vote. |
|
Average rating: |
0 |
| | | | | |
|
|
 |
News and Press Releases |
 |
Catalent Invests $175 Million to Expand Manufacturing Capabilities and Capacity at Winchester, Kentucky
SOMERSET, N.J. – May 19, 2022 — Catalent, the global leader in
enabling biopharma, cell, gene and consumer health partners to optimize
development, launch, and supply of better patient treatments across
multiple modalities, today announced that it has commenced a $175
million project to expand its flagship U.S. manufacturing facility for
large scale oral dose forms in Winchester, Kentucky. Two new buildings,
expected to be completed by January 2024, will add 107,000 square feet
to the site’s manufacturing footprint.
More info >> |
|
 |
White Papers |
 |
|
Translating Cancer Genomics Into A Personalised Approach For Cancer Patients
PRA Health Sciences (PRA)
Over the past few decades, the approach to prevention, diagnosis and treatment of cancer has radically shifted from organ-based to morphology-based and most recently, to genetics-based. Personalized or precision medicine (tailoring a treatment for a patient’s particular disease at a precise time point) is being performed every day at different levels in the clinical setting.
More info >> |
|
|
.jpg)
