The current status of risk-stratified breast screening.

Apart from high-risk scenarios such as the presence of highly penetrant genetic mutations, breast screening typically comprises mammography or tomosynthesis strategies defined by age. However, age-based screening ignores the range of breast cancer risks that individual women may possess and is antithetical to the ambitions of personalised early detection. Whilst screening mammography reduces breast cancer mortality, this is at the risk of potentially significant harms including overdiagnosis with overtreatment, and psychological morbidity associated with false positives. In risk-stratified screening, individualised risk assessment may inform screening intensity/interval, starting age, imaging modality used, or even decisions not to screen. However, clear evidence for its benefits and harms needs to be established. In this scoping review, the authors summarise the established and emerging evidence regarding several critical dependencies for successful risk-stratified breast screening: risk prediction model performance, epidemiological studies, retrospective clinical evaluations, health economic evaluations and qualitative research on feasibility and acceptability. Family history, breast density or reproductive factors are not on their own suitable for precisely estimating risk and risk prediction models increasingly incorporate combinations of demographic, clinical, genetic and imaging-related parameters. Clinical evaluations of risk-stratified screening are currently limited. Epidemiological evidence is sparse, and randomised trials only began in recent years.

Image-Based Artefact Removal in Laser Scanning Microscopy.

Recent developments in laser scanning microscopy have greatly extended its applicability in cancer imaging beyond the visualization of complex biology, and opened up the possibility of quantitative analysis of inherently dynamic biological processes. However, the physics of image acquisition intrinsically means that image quality is subject to a tradeoff between a number of imaging parameters, including resolution, signal-to-noise ratio, and acquisition speed. We address the problem of geometric distortion, in particular, jaggedness artefacts that are caused by the variable motion of the microscope laser, by using a combination of image processing techniques. Image restoration methods have already shown great potential for post-acquisition image analysis. The performance of our proposed image restoration technique was first quantitatively evaluated using phantom data with different textures, and then qualitatively assessed using in vivo biological imaging data. In both cases, the presented method, comprising a combination of image registration and filtering, is demonstrated to have substantial improvement over state-of-the-art microscopy acquisition methods.

Segmentation of Vasculature From Fluorescently Labeled Endothelial Cells in Multi-Photon Microscopy Images.

Vasculature is known to be of key biological significance, especially in the study of tumors. As such, considerable effort has been focused on the automated segmentation of vasculature in medical and pre-clinical images. The majority of vascular segmentation methods focus on bloodpool labeling methods; however, particularly, in the study of tumors, it is of particular interest to be able to visualize both the perfused and the non-perfused vasculature. Imaging vasculature by highlighting the endothelium provides a way to separate the morphology of vasculature from the potentially confounding factor of perfusion. Here, we present a method for the segmentation of tumor vasculature in 3D fluorescence microscopic images using signals from the endothelial and surrounding cells. We show that our method can provide complete and semantically meaningful segmentations of complex vasculature using a supervoxel-Markov random field approach. We show that in terms of extracting meaningful segmentations of the vasculature, our method outperforms both state-of-the-art method, specific to these data, as well as more classical vasculature segmentation methods.

Oncological image analysis.

Cancer is one of the world's major healthcare challenges and, as such, an important application of medical image analysis. After a brief introduction to cancer, we summarise some of the major developments in oncological image analysis over the past 20 years, but concentrating those in the authors' laboratories, and then outline opportunities and challenges for the next decade.

Pieces-of-parts for supervoxel segmentation with global context: Application to DCE-MRI tumour delineation.

Rectal tumour segmentation in dynamic contrast-enhanced MRI (DCE-MRI) is a challenging task, and an automated and consistent method would be highly desirable to improve the modelling and prediction of patient outcomes from tissue contrast enhancement characteristics - particularly in routine clinical practice. A framework is developed to automate DCE-MRI tumour segmentation, by introducing: perfusion-supervoxels to over-segment and classify DCE-MRI volumes using the dynamic contrast enhancement characteristics; and the pieces-of-parts graphical model, which adds global (anatomic) constraints that further refine the supervoxel components that comprise the tumour. The framework was evaluated on 23 DCE-MRI scans of patients with rectal adenocarcinomas, and achieved a voxelwise area-under the receiver operating characteristic curve (AUC) of 0.97 compared to expert delineations. Creating a binary tumour segmentation, 21 of the 23 cases were segmented correctly with a median Dice similarity coefficient (DSC) of 0.63, which is close to the inter-rater variability of this challenging task. A second study is also included to demonstrate the method's generalisability and achieved a DSC of 0.71. The framework achieves promising results for the underexplored area of rectal tumour segmentation in DCE-MRI, and the methods have potential to be applied to other DCE-MRI and supervoxel segmentation problems.

Deformable image registration by combining uncertainty estimates from supervoxel belief propagation.

Discrete optimisation strategies have a number of advantages over their continuous counterparts for deformable registration of medical images. For example: it is not necessary to compute derivatives of the similarity term; dense sampling of the search space reduces the risk of becoming trapped in local optima; and (in principle) an optimum can be found without resorting to iterative coarse-to-fine warping strategies. However, the large complexity of high-dimensional medical data renders a direct voxel-wise estimation of deformation vectors impractical. For this reason, previous work on medical image registration using graphical models has largely relied on using a parameterised deformation model and on the use of iterative coarse-to-fine optimisation schemes. In this paper, we propose an approach that enables accurate voxel-wise deformable registration of high-resolution 3D images without the need for intermediate image warping or a multi-resolution scheme. This is achieved by representing the image domain as multiple comprehensive supervoxel layers and making use of the full marginal distribution of all probable displacement vectors after inferring regularity of the deformations using belief propagation. The optimisation acts on the coarse scale representation of supervoxels, which provides sufficient spatial context and is robust to noise in low contrast areas. Minimum spanning trees, which connect neighbouring supervoxels, are employed to model pair-wise deformation dependencies. The optimal displacement for each voxel is calculated by considering the probabilities for all displacements over all overlapping supervoxel graphs and subsequently seeking the mode of this distribution. We demonstrate the applicability of this concept for two challenging applications: first, for intra-patient motion estimation in lung CT scans; and second, for atlas-based segmentation propagation of MRI brain scans. For lung registration, the voxel-wise mode of displacements is found using the mean-shift algorithm, which enables us to determine continuous valued sub-voxel motion vectors. Finding the mode of brain segmentation labels is performed using a voxel-wise majority voting weighted by the displacement uncertainty estimates. Our experimental results show significant improvements in registration accuracy when using the additional information provided by the registration uncertainty estimates. The multi-layer approach enables fusion of multiple complementary proposals, extending the popular fusion approaches from multi-image registration to probabilistic one-to-one image registration.

Automated colorectal tumour segmentation in DCE-MRI using supervoxel neighbourhood contrast characteristics.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a powerful protocol for assessing tumour progression from changes in tissue contrast enhancement. Manual colorectal tumour delineation is a challenging and time consuming task due to the complex enhancement patterns in the 4D sequence. There is a need for a consistent approach to colorectal tumour segmentation in DCE-MRI and we propose a novel method based on detection of the tumour from signal enhancement characteristics of homogeneous tumour subregions and their neighbourhoods. Our method successfully detected 20 of 23 cases with a mean Dice score of 0.68 +/- 0.15 compared to expert annotations, which is not significantly different from expert inter-rater variability of 0.73 +/- 0.13 and 0.77 +/- 0.10. In comparison, a standard DCE-MRI tumour segmentation technique, fuzzy c-means, obtained a Dice score of 0.28 +/- 0.17.

Integral invariants for image enhancement.

Medical images pose a major challenge for image analysis: often they have poor signal-to-noise, necessitating smoothing; yet such smoothing needs to preserve the boundaries of regions of interest and small features such as mammogram microcalcifications. We show how circular integral invariants (II) may be adapted for feature-preserving smoothing to facilitate segmentation. Though II is isotropic, we show that it leads to considerably less feature deterioration than Gaussian blurring and it improves segmentation of regions of interest as compared to anisotropic diffusion, particularly for hierarchical contour based segmentation methods.

Towards realtime multimodal fusion for image-guided interventions using self-similarities.

Image-guided interventions often rely on deformable multimodal registration to align pre-treatment and intra-operative scans. There are a number of requirements for automated image registration for this task, such as a robust similarity metric for scans of different modalities with different noise distributions and contrast, an efficient optimisation of the cost function to enable fast registration for this time-sensitive application, and an insensitive choice of registration parameters to avoid delays in practical clinical use. In this work, we build upon the concept of structural image representation for multi-modal similarity. Discriminative descriptors are densely extracted for the multi-modal scans based on the "self-similarity context". An efficient quantised representation is derived that enables very fast computation of point-wise distances between descriptors. A symmetric multi-scale discrete optimisation with diffusion reguIarisation is used to find smooth transformations. The method is evaluated for the registration of 3D ultrasound and MRI brain scans for neurosurgery and demonstrates a significantly reduced registration error (on average 2.1 mm) compared to commonly used similarity metrics and computation times of less than 30 seconds per 3D registration.

The impact of heterogeneity and uncertainty on prediction of response to therapy using dynamic MRI data.

A comprehensive framework for predicting response to therapy on the basis of heterogeneity in dceMRI parameter maps is presented. A motion-correction method for dceMRI sequences is extended to incorporate uncertainties in the pharmacokinetic parameter maps using a variational Bayes framework. Simple measures of heterogeneity (with and without uncertainty) in parameter maps for colorectal cancer tumours imaged before therapy are computed, and tested for their ability to distinguish between responders and non-responders to therapy. The statistical analysis demonstrates the importance of using the spatial distribution of parameters, and their uncertainties, when computing heterogeneity measures and using them to predict response on the basis of the pre-therapy scan. The results also demonstrate the benefits of using the ratio of Ktrans with the bolus arrival time as a biomarker.

Edge- and detail-preserving sparse image representations for deformable registration of chest MRI and CT volumes.

Deformable medical image registration requires the optimisation of a function with a large number of degrees of freedom. Commonly-used approaches to reduce the computational complexity, such as uniform B-splines and Gaussian image pyramids, introduce translation-invariant homogeneous smoothing, and may lead to less accurate registration in particular for motion fields with discontinuities. This paper introduces the concept of sparse image representation based on supervoxels, which are edge-preserving and therefore enable accurate modelling of sliding organ motions frequently seen in respiratory and cardiac scans. Previous shortcomings of using supervoxels in motion estimation, in particular inconsistent clustering in ambiguous regions, are overcome by employing multiple layers of supervoxels. Furthermore, we propose a new similarity criterion based on a binary shape representation of supervoxels, which improves the accuracy of single-modal registration and enables multimodal registration. We validate our findings based on the registration of two challenging clinical applications of volumetric deformable registration: motion estimation between inhale and exhale phase of CT scans for radiotherapy planning, and deformable multi-modal registration of diagnostic MRI and CT chest scans. The experiments demonstrate state-of-the-art registration accuracy, and require no additional anatomical knowledge with greatly reduced computational complexity.

Quantification and normalization of x-ray mammograms.

The analysis of (x-ray) mammograms remains qualitative, relying on the judgement of clinicians. We present a novel method to compute a quantitative, normalized measure of tissue radiodensity traversed by the primary beam incident on each pixel of a mammogram, a measure we term the standard attenuation rate (SAR). SAR enables: the estimation of breast density which is linked to cancer risk; direct comparison between images; the full potential of computer aided diagnosis to be utilized; and a basis for digital breast tomosynthesis reconstruction. It does this by removing the effects of the imaging conditions under which the mammogram is acquired. First, the x-ray spectrum incident upon the breast is calculated, and from this, the energy exiting the breast is calculated. The contribution of scattered radiation is calculated and subtracted. The SAR measure is the scaling factor that must be applied to the reference material in order to match the primary attenuation of the breast. Specifically, this is the scaled reference material attenuation which when traversed by an identical beam to that traversing the breast, and when subsequently detected, results in the primary component of the pixel intensity observed in the breast image. We present results using two tissue equivalent phantoms, as well as a sensitivity analysis to detector response changes over time and possible errors in compressed thickness measurement.

A model of primary and scattered photon fluence for mammographic x-ray image quantification.

We present an efficient method to calculate the primary and scattered x-ray photon fluence component of a mammographic image. This can be used for a range of clinically important purposes, including estimation of breast density, personalized image display, and quantitative mammogram analysis. The method is based on models of: the x-ray tube; the digital detector; and a novel ray tracer which models the diverging beam emanating from the focal spot. The tube model includes consideration of the anode heel effect, and empirical corrections for wear and manufacturing tolerances. The detector model is empirical, being based on a family of transfer functions that cover the range of beam qualities and compressed breast thicknesses which are encountered clinically. The scatter estimation utilizes optimal information sampling and interpolation (to yield a clinical usable computation time) of scatter calculated using fundamental physics relations. A scatter kernel arising around each primary ray is calculated, and these are summed by superposition to form the scatter image. Beam quality, spatial position in the field (in particular that arising at the air-boundary due to the depletion of scatter contribution from the surroundings), and the possible presence of a grid, are considered, as is tissue composition using an iterative refinement procedure. We present numerous validation results that use a purpose designed tissue equivalent step wedge phantom. The average differences between actual acquisitions and modelled pixel intensities observed across the adipose to fibroglandular attenuation range vary between 5% and 7%, depending on beam quality and, for a single beam quality are 2.09% and 3.36% respectively with and without a grid.

MIND: modality independent neighbourhood descriptor for multi-modal deformable registration.

Deformable registration of images obtained from different modalities remains a challenging task in medical image analysis. This paper addresses this important problem and proposes a modality independent neighbourhood descriptor (MIND) for both linear and deformable multi-modal registration. Based on the similarity of small image patches within one image, it aims to extract the distinctive structure in a local neighbourhood, which is preserved across modalities. The descriptor is based on the concept of image self-similarity, which has been introduced for non-local means filtering for image denoising. It is able to distinguish between different types of features such as corners, edges and homogeneously textured regions. MIND is robust to the most considerable differences between modalities: non-functional intensity relations, image noise and non-uniform bias fields. The multi-dimensional descriptor can be efficiently computed in a dense fashion across the whole image and provides point-wise local similarity across modalities based on the absolute or squared difference between descriptors, making it applicable for a wide range of transformation models and optimisation algorithms. We use the sum of squared differences of the MIND representations of the images as a similarity metric within a symmetric non-parametric Gauss-Newton registration framework. In principle, MIND would be applicable to the registration of arbitrary modalities. In this work, we apply and validate it for the registration of clinical 3D thoracic CT scans between inhale and exhale as well as the alignment of 3D CT and MRI scans. Experimental results show the advantages of MIND over state-of-the-art techniques such as conditional mutual information and entropy images, with respect to clinically annotated landmark locations.

Personalization of pictorial structures for anatomical landmark localization.

We propose a method for accurately localizing anatomical landmarks in 3D medical volumes based on dense matching of parts-based graphical models. Our novel approach replaces population mean models by jointly leveraging weighted combinations of labeled exemplars (both spatial and appearance) to obtain personalized models for the localization of arbitrary landmarks in upper body images. We compare the method to a baseline population-mean graphical model and atlas-based deformable registration optimized for CT-CT registration, by measuring the localization accuracy of 22 anatomical landmarks in clinical 3D CT volumes, using a database of 83 lung cancer patients. The average mean localization error across all landmarks is 2.35 voxels. Our proposed method outperforms deformable registration by 73%, 93% for the most improved landmark. Compared to the baseline population-mean graphical model, the average improvement of localization accuracy is 32%; 67% for the most improved landmark.

Task-based performance analysis of FBP, SART and ML for digital breast tomosynthesis using signal CNR and Channelised Hotelling Observers.

We assess the performance of filtered backprojection (FBP), the simultaneous algebraic reconstruction technique (SART) and the maximum likelihood (ML) algorithm for digital breast tomosynthesis (DBT) under variations in key imaging parameters, including the number of iterations, number of projections, angular range, initial guess, and radiation dose. This is the first study to compare these algorithms for the application of DBT. We present a methodology for the evaluation of DBT reconstructions, and use it to conduct preliminary experiments investigating trade-offs between the selected imaging parameters. This investigation includes trade-offs not previously considered in the DBT literature, such as the use of a stationary detector versus a C-arm imaging geometry. A real breast CT volume serves as a ground truth digital phantom from which to simulate X-ray projections under the various acquisition parameters. The reconstructed image quality is measured using task-based metrics, namely signal CNR and the AUC of a Channelised Hotelling Observer with Laguerre-Gauss basis functions. The task at hand is the detection of a simulated mass inserted into the breast CT volume. We find that the image quality in limited view tomography is highly dependent on the particular acquisition and reconstruction parameters used. In particular, we draw the following conclusions. First, we find that optimising the FBP filter design and SART relaxation parameter yields significant improvements in reconstruction quality from the same projection data. Second, we show that the convergence rate of the maximum likelihood algorithm, optimised with paraboloidal surrogates and conjugate gradient ascent (ML-PSCG), can be greatly accelerated using view-by-view updates. Third, we find that the optimal initial guess is algorithm dependent. In particular, we obtained best results with a zero initial guess for SART, and an FBP initial guess for ML-PSCG. Fourth, when the exposure per view is constant, increasing the total number of views within a given angular range improves the reconstruction quality, albeit with diminishing returns. When the total dose of all views combined is constant, there is a trade-off between increased sampling using a larger number of views and increased levels of quantum noise in each view. Fifth, we do not observe significant differences when testing various access ordering schemes, presumably due to the limited angular range of DBT. Sixth, we find that adjusting the z-resolution of the reconstruction can improve image quality, but that this resolution is best adjusted by using post-reconstruction binning, rather than by declaring lower-resolution voxels. Seventh, we find that the C-arm configuration yields higher image quality than a stationary detector geometry, the difference being most outspoken for the FBP algorithm. Lastly, we find that not all prototype systems found in the literature are currently being run under the best possible system or algorithm configurations. In other words, the present study demonstrates the critical importance (and reward) of using optimisation methodologies such as the one presented here to maximise the DBT reconstruction quality from a single scan of the patient.

Robust joint entropy regularization of limited view transmission tomography using Gaussian approximations to the joint histogram.

Information theoretic measures to incorporate anatomical priors have been explored in the field of emission tomography, but not in transmission tomography. In this work, we apply the joint entropy prior to the case of limited angle transmission tomography. Due to the data insufficiency problem, the joint entropy prior is found to be very sensitive to local optima. Two methods for robust joint entropy minimization are proposed. The first approximates the joint probability density function by a single 2D Gaussian, and is found to be appropriate for reconstructions where the ground truth joint histogram is dominated by two clusters, or multiple clusters that are roughly aligned. The second method is an extension to the case of multiple Gaussians. The intended application for the single Gaussian approximation is digital breast tomosynthesis, where reconstructed volumes are approximately bimodal, consisting mainly of fatty and fibroglandular tissues.

Intensity correction with a pair of spoiled gradient recalled echo images.

Intensity inhomogeneities in magnetic resonance images (MRI) are a frequently occurring artefact, and result in the same tissue class to have vastly different intensities within an image. These inhomogeneities can be modelled by a slowly varying field, which is also called the bias field. Previous phantom-, image- or sequence based approaches suffer from long scan times, post-processing times or do not sufficiently remove the intensity variations. These intensity variations cause problems for quantitative image analysis algorithms (segmentation, registration) as well as clinicians (e.g. by complicating the visual assessment). This paper presents a novel technique (COIN, correction of intensity inhomogeneities) that uses two calibration images (fast spoiled gradient echo) to map a parameter containing the bias field, which is specific to the patient during a particular exam. This parametric map can then be used to correct any other images acquired during the same exam, regardless of the sequence employed. By using a short repetition time (less than 5 ms) for the calibration scans, the additional scan time is reduced to 60 s (max). The subsequent post-processing time is approximately 60 s per 20 slices. We successfully validate our approach on simulated brain MRI as well as real liver and spinal images. These images were acquired with a number of different coils, sequences and weightings. A comparison of our method with an existing, commercially available algorithm by radiologists shows that COIN is superior.