home > ebr > winter 2014 > advanced biomarkers
European Biopharmaceutical Review

Advanced Biomarkers

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.


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.

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.

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:

There are no comments in regards to this article.

About the authors

Andrew J Buckler is President, Chief Executive Officer and co-founder of Elucid Bioimaging. He has 30 years of experience in imaging science, medical devices and molecular medicine. Andrew is also the Programme Director for the Quantitative Imaging Biomarker Alliance and is the Scientific Advisor to the Foundation for the National Institutes of Health. He holds an MSCS from the University of Rochester and a BSEE from Lehigh University.

James C Keith Jr. is Senior Director of Alliances and Contract Research at Elucid Bioimaging. He has 31 years of experience in atherosclerosis research, pharmacology and comparative medicine in academia, biotech and pharma. James is the co-author of over 140 peer-reviewed papers and book chapters, as well as an invited reviewer for multiple national and international scientific journals, and an inventor on 31 issued US and international patents. He holds a PhD from The University of Georgia, and BS and DVM, degrees from The University of Tennessee.
Andrew J Buckler
James C Keith
Print this page
Send to a friend
Privacy statement
News and Press Releases

LEO Pharma and ICON enter a strategic partnership to propel clinical trial execution within medical dermatology

BALLERUP, DENMARK and DUBLIN, IRELAND — 10 March, 2023 — LEO Pharma, a global leader in medical dermatology, and ICON plc (NASDAQ: ICLR) today announced a strategic partnership that will enable LEO Pharma to scale clinical trial execution that is patient-centric and cost effective, and which will support the company’s overall ambition of building one of the most effective and efficient clinical portfolio execution organisations in the industry.
More info >>

White Papers

Preparation of Pharmaceutical Samples for Metals Analysis


There has been a lot of discussion regarding the methods of analysis for pharmaceutical products as we await the upcoming changes to both the EP and USP, relating to the analysis of metal impurities.
More info >>




©2000-2011 Samedan Ltd.
Add to favourites

Print this page

Send to a friend
Privacy statement