Project topics and materials
Get Complete Project Material File(s) Now! »
EOS: a biplanar low-dose X-ray unit
EOS is a Stereo-Radiographic imaging system based on slot scanning principles that is produced by the French manufacturer EOS imaging. This is the only musculoskeletal system that allows ultra-low dose, simultaneous acquisition of full body frontal and lateral images in weight-bearing position.
Global architecture and properties
EOS acquisition principle is based on the multiwire gaseous detector that has been conceived from Georges Charpak’s1 researches (Charpak et al., 1968). This detector is capable of extracting a significant amount of information from a single photon while being practically unaffected by scattering, which makes it adapted for X-ray image acquisitions. The detector amplifies the X-rays that pass through the body of a scanned patient by means of electronic avalanche in the gas. In medical applications, the amplification of the detector is crucial as it allows obtaining a high quality X-ray image while keeping the dose low. The detector is composed by 1764 strips focused on the X-ray focal spot, i.e. parallel to the incident X-ray photon trajectories. These channels are placed at 254 μm wide intervals, i.e. the distance that corresponds to the spatial resolution within the horizontal plane. Please refer to Damet et al. (2014) for a quantitative evaluation of the detector performances.
EOS system occupies a 2 m2 sized area and measures 2.70 m in height. The patient is located in this area in a weight-bearing standing position, or, less often, in a seated position. The possibility of examining the patient in its natural standing position gives a better insight into possible diseases or deformations of the musculoskeletal system than computed tomography (CT) imaging (Dubousset et al., 2005).
Figure 1.1a shows how the EOS images are acquired. A c-arm scans the entire body of a patient undergoing an examination by covering, at most, a 45 cm wide and 180 cm high field of view in 23 s on average. The X-ray tube emits a thin collimated X-ray fan-beam that is detected line by line. Two couples of co-linked X-ray tubes and detectors are placed orthogonally to each other on the c-arm and, thus, two images are simultaneously acquired: the frontal (AP) and lateral (LAT) view acquisitions.
The two images are then sent to two separated image processing chains.
Given this acquisition procedure, EOS is also known as bi-planar X-ray full-body system.
Figure 1.1b shows that the frontal and lateral view acquisitions can be used to extract a personalized 3D model of the skeleton, which is used to obtain some relevant 2D/3D clinical parameters. This task is achieved by using an EOS imaging proprietary software called sterEOS. It requires the user to identify a limited number of anatomical landmarks. The reconstruction is then completed by exploiting a database of anatomical 3D shapes.
Dosimetric quantities and units
The exposure is a dosimetric measure that refers to ionizing electromagnetic radiations. The amount of energy delivered to a tissue that interacts with ionizing radiations is so high that electrons are liberated from the atoms. Therefore, the tissues can be considerably damaged. The exposure is measured in coulombs per kilogram (Ckg−1) or in roentgens (R).
The exposure is proportional to the absorbed dose that is defined as the energy absorbed per unit mass and is, hence, measured in joules per kilogram (Jkg−1), which are better known as grays (Gy). In medical practice, the absorbed dose depends on the acquisition parameters (see Section 1.1.1), the material used to filter the X-ray beam and the field size. The proportionality factor f between the exposure and the dose depends on the material and on the kV. For example, by considering bone tissues, f(10 kV ) = 3.5 and f(100 kV ) = 1.5. The exposure is often linked to the dose absorbed by the air, which is also called air kerma (kinetic energy released in a material) as in diagnostic radiology the kerma and the absorbed dose coincide. In this manuscript we will use the kerma in the air as measure of dose. The equivalent dose is another important quantity. It is proportional to the absorbed dose as a function of the biological effectiveness of a given radiation and is measured in sieverts (Sv). However, in radiology the proportionality factor among the two quantities is equal to one, i.e. they are identical. The equivalent dose is modified by a factor called detriment, that is the stochastic combination of the probability and extent of harming associated with a given tissue. Therefore, the equivalent dose is more relevant from a clinical point of view.
Table of contents :
Remerciements
Résumé
Synthèse des travaux de thèse
Contexte
Le système EOS
Le principe ALARA en radiologie
Approches choisies et défis
Optimisation de la qualité des images
Quantification automatique de la qualité des images à l’acquisition .
Débruitage et rehaussement de contraste conjoint
État de l’art sur le traitement d’images de radiographie
Filtre par moyennes non locales pour des images à rayons
Carte de bruit
Validation
Conclusion
Points de repère pour l’estimation de l’indice d’exposition
Comparaison de méthodes de définition de ROI et détection
Détection de points de repère
Validation des valeurs d’indice d’exposition automatiques
Conclusion et perspectives
Abstract
Contents
1 Introduction
1.1 EOS: a biplanar low-dose X-ray unit
1.1.1 Global architecture and properties
1.1.2 Examples of applications
1.2 ALARA principle
1.2.1 Dosimetric quantities and units
1.2.2 Dose related risks and ALARA principle
1.2.3 EOS and dose
1.3 Objectives of the thesis and related challenges
1.3.1 Optimization of the image quality
1.3.2 Automatic quantification of the image quality
1.3.3 Organization of the manuscript
2 LowdoseX-rayimages: amatterofhigh noise andlowcontrast
2.1 Image denoising
2.1.1 Introduction to the denoising problem
2.1.2 Filtering in image space
2.1.3 Filtering in transform domain
2.1.4 Filtering in patch space
2.1.5 Beyond the homoscedastic noise model
2.1.6 Reduction of noise in medical applications
2.2 Contrast enhancement techniques
2.2.1 Introduction to the contrast enhancement problem
2.2.2 Multiscale image processing for contrast enhancement
2.2.3 Noise in contrast enhancement
2.3 Evaluation of the image quality in digital radiography
2.3.1 Phantoms
2.3.2 Proposed protocol for clinical images
2.4 Conclusion
3 Jointdenoising andcontrast enhancement ofX-rayimages
3.1 Noise characterization
3.1.1 Noise model
3.1.2 The drawbacks of the Anscombe transform on EOS images .
3.1.3 The percentile method for noise level estimation
3.1.4 Application to clinical EOS images
3.2 X-ray Non Local-means Filter
3.2.1 Formulation
3.2.2 Qualitative evaluation of the X-ray Non Local Means Filter .
3.3 Contrast Enhancement of images affected by noise
3.3.1 Multiscale analysis of the X-ray Non Local Means Filter
3.3.2 Definition of local noise containment maps
3.3.3 Proposed boosting technique
3.3.4 Overview of the proposed framework
3.4 Evaluation of the framework on phantom images
3.4.1 Signal to noise ratio
3.4.2 Contrast to noise ratio
3.4.3 Dynamic
3.4.4 Spatial resolution
3.4.5 Gain in terms of dose reduction
3.5 Evaluation of the framework on clinical images
3.5.1 Quantitative measures for clinical image quality evaluation .
3.5.2 Quantitative feedback from clinicians
3.6 Conclusion
4 Exposure index: overview and anewl and markbased approach
4.1 Measures of exposure level in digital radiography
4.1.1 Motivation
4.1.2 Exposure index algorithm
4.1.3 Proposed measure based on SNR
4.1.4 Link between EI and SNR based measures
4.1.5 Exposure management
4.2 Set of exposure index values on EOS images
4.2.1 Challenges related to full-body X-ray images
4.2.2 Possible applications
4.3 State of the art approaches applied to EOS images
4.3.1 Overview of state of the art approaches
4.3.2 Evaluation of the state of the art methods on EOS images
4.4 Anatomical structure detection applied to the exposure index
4.4.1 Bounding box based approach
4.4.2 Proposed landmark based approach
4.4.3 Robustness to localization errors
4.5 Conclusion
5 Automatic landmarkdetectiontoestimate exposure indices
5.1 State of the art methods for detection of anatomical structures
5.1.1 Atlas registration-based techniques
5.1.2 Classification
5.1.3 Regression
5.1.4 Sequential models
5.1.5 Specificity of the landmark detection on EOS images
5.2 Main notations
5.3 Initialization of the method from spatial relations
5.3.1 Absolute distance from a given control point
5.3.2 Proportions
5.4 Salient points
5.4.1 Saliency as change in density
5.4.2 Applications
5.5 Landmarks detection and recognition on frontal view acquisitions
5.5.1 Detection of control points
5.5.2 Detection of the landmarks in the legs
5.5.3 Detection of the landmarks in the spine
5.5.4 Detection of the landmarks in the lung
5.5.5 Detection of the landmarks in the pelvis
5.6 Landmarks detection and recognition on lateral view acquisitions
5.6.1 Detection of the landmarks in the head, pelvis and legs
5.6.2 Detection of the landmarks in the thoracic spine
5.6.3 Detection of the landmarks in the lumbar spine
5.6.4 Refinement of the control point for the vertebra T12
5.7 Evaluation of automatically computed exposure index values
5.7.1 Quantitative evaluation on frontal view acquisitions
5.7.2 Quantitative evaluation on lateral view acquisitions
5.7.3 Metallic objects
5.7.4 Robustness to particular conditions of the patients
5.7.5 Information extracted from preview images
5.8 Conclusion
6 Conclusion and perspectives
6.1 Contributions and open issues
6.1.1 Optimization of the image quality
6.1.2 Estimation of the image quality
6.2 Perspectives
6.2.1 More on image processing
6.2.2 Applications with preview images
A Computation of patient’s envelop and landmarks
A.1 Method for the extraction of the patient’s envelop
A.2 Landmarks inferred from the envelop
A.3 Examples
B Qualitative evaluation of landmarks detection and recognition
B.1 Detection on adult patients
B.2 Detection on young patients and corresponding preview acquisitions
C List of themain acronyms and notations
Bibliography
