Evaluating an optical-flow-based registration algorithm for contrast-enhanced magnetic resonance imaging of the breast

Development and validation of a four-dimensional registration technique for DCE breast MRI

BACKGROUND

Patient motion can degrade image quality of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) due to subtraction artifacts. By objectively and subjectively assessing the impact of principal component analysis (PCA)-based registration on pretreatment DCE-MRIs of breast cancer patients, we aim to validate four-dimensional registration for DCE breast MRI.

RESULTS

After applying a four-dimensional, PCA-based registration algorithm to 154 pretreatment DCE-MRIs of histopathologically well-described breast cancer patients, we quantitatively determined image quality in unregistered and registered images. For subjective assessment, we ranked motion severity in a clinical reading setting according to four motion categories (0: no motion, 1: mild motion, 2: moderate motion, 3: severe motion with nondiagnostic image quality). The median of images with either moderate or severe motion (median category 2, IQR 0) was reassigned to motion category 1 (IQR 0) after registration. Motion category and motion reduction by registration were correlated (Spearman's rho: 0.83, p < 0.001). For objective assessment, we performed perfusion model fitting using the extended Tofts model and calculated its volume transfer coefficient K^trans as surrogate parameter for motion artifacts. Mean K^trans decreased from 0.103 (± 0.077) before registration to 0.097 (± 0.070) after registration (p < 0.001). Uncertainty in perfusion quantification was reduced by 7.4% after registration (± 15.5, p < 0.001).

CONCLUSIONS

Four-dimensional, PCA-based image registration improves image quality of breast DCE-MRI by correcting for motion artifacts in subtraction images and reduces uncertainty in quantitative perfusion modeling. The improvement is most pronounced when moderate-to-severe motion artifacts are present.

Motion correction of free-breathing magnetic resonance renography using model-driven registration

INTRODUCTION

Model-driven registration (MDR) is a general approach to remove patient motion in quantitative imaging. In this study, we investigate whether MDR can effectively correct the motion in free-breathing MR renography (MRR).

MATERIALS AND METHODS

MDR was generalised to linear tracer-kinetic models and implemented using 2D or 3D free-form deformations (FFD) with multi-resolution and gradient descent optimization. MDR was evaluated using a kidney-mimicking digital reference object (DRO) and free-breathing patient data acquired at high temporal resolution in multi-slice 2D (5 patients) and 3D acquisitions (8 patients). Registration accuracy was assessed using comparison to ground truth DRO, calculating the Hausdorff distance (HD) between ground truth masks with segmentations and visual evaluation of dynamic images, signal-time courses and parametric maps (all data).

RESULTS

DRO data showed that the bias and precision of parameter maps after MDR are indistinguishable from motion-free data. MDR led to reduction in HD (HDunregistered = 9.98 ± 9.76, HDregistered = 1.63 ± 0.49). Visual inspection showed that MDR effectively removed motion effects in the dynamic data, leading to a clear improvement in anatomical delineation on parametric maps and a reduction in motion-induced oscillations on signal-time courses.

DISCUSSION

MDR provides effective motion correction of MRR in synthetic and patient data. Future work is needed to compare the performance against other more established methods.

The Effect of Registration on Voxel-Wise Tofts Model Parameters and Uncertainties from DCE-MRI of Early-Stage Breast Cancer Patients Using 3DSlicer

We quantitatively investigate the influence of image registration, using open-source software (3DSlicer), on kinetic analysis (Tofts model) of dynamic contrast enhanced MRI of early-stage breast cancer patients. We also show that registration computation time can be reduced by reducing the percent sampling (PS) of voxels used for estimation of the cost function. DCE-MRI breast images were acquired on a 3T-PET/MRI system in 13 patients with early-stage breast cancer who were scanned in a prone radiotherapy position. Images were registered using a BSpline transformation with a 2 cm isotropic grid at 100, 20, 5, 1, and 0.5PS (BRAINSFit in 3DSlicer). Signal enhancement curves were analyzed voxel-by-voxel using the Tofts kinetic model. Comparing unregistered with registered groups, we found a significant change in the 90th percentile of the voxel-wise distribution of K^trans. We also found a significant reduction in the following: (1) in the standard error (uncertainty) of the parameter value estimation, (2) the number of voxel fits providing unphysical values for the extracellular-extravascular volume fraction (v_e > 1), and (3) goodness of fit. We found no significant differences in the median of parameter value distributions (K^trans, v_e) between unregistered and registered images. Differences between parameters and uncertainties obtained using 100PS versus 20PS were small and statistically insignificant. As such, computation time can be reduced by a factor of 2, on average, by using 20PS while not affecting the kinetic fit. The methods outlined here are important for studies including a large number of post-contrast images or number of patient images.

An investigation of the effect of fat suppression and dimensionality on the accuracy of breast MRI segmentation using U-nets

PURPOSE

Accurate segmentation of the breast is required for breast density estimation and the assessment of background parenchymal enhancement, both of which have been shown to be related to breast cancer risk. The MRI breast segmentation task is challenging, and recent work has demonstrated that convolutional neural networks perform well for this task. In this study, we have investigated the performance of several two-dimensional (2D) U-Net and three-dimensional (3D) U-Net configurations using both fat-suppressed and nonfat-suppressed images. We have also assessed the effect of changing the number and quality of the ground truth segmentations.

MATERIALS AND METHODS

We designed eight studies to investigate the effect of input types and the dimensionality of the U-Net operations for the breast MRI segmentation. Our training data contained 70 whole breast volumes of T1-weighted sequences without fat suppression (WOFS) and with fat suppression (FS). For each subject, we registered the WOFS and FS volumes together before manually segmenting the breast to generate ground truth. We compared four different input types to the U-nets: WOFS, FS, MIXED (WOFS and FS images treated as separate samples), and MULTI (WOFS and FS images combined into a single multichannel image). We trained 2D U-Nets and 3D U-Nets with these data, which resulted in our eight studies (2D-WOFS, 3D-WOFS, 2D-FS, 3D-FS, 2D-MIXED, 3D-MIXED, 2D-MULTI, and 3D-MULT). For each of these studies, we performed a systematic grid search to tune the hyperparameters of the U-Nets. A separate validation set with 15 whole breast volumes was used for hyperparameter tuning. We performed Kruskal-Walis test on the results of our hyperparameter tuning and did not find a statistically significant difference in the ten top models of each study. For this reason, we chose the best model as the model with the highest mean dice similarity coefficient (DSC) value on the validation set. The reported test results are the results of the top model of each study on our test set which contained 19 whole breast volumes annotated by three readers fused with the STAPLE algorithm. We also investigated the effect of the quality of the training annotations and the number of training samples for this task.

RESULTS

The study with the highest average DSC result was 3D-MULTI with 0.96 ± 0.02. The second highest average is 2D WOFS (0.96 ± 0.03), and the third is 2D MULTI (0.96 ± 0.03). We performed the Kruskal-Wallis one-way ANOVA test with Dunn's multiple comparison tests using Bonferroni P-value correction on the results of the selected model of each study and found that 3D-MULTI, 2D-MULTI, 3D-WOFS, 2D-WOFS, 2D-FS, and 3D-FS were not statistically different in their distributions, which indicates that comparable results could be obtained in fat-suppressed and nonfat-suppressed volumes and that there is no significant difference between the 3D and 2D approach. Our results also suggested that the networks trained on single sequence images or multiple sequence images organized in multichannel images perform better than the models trained on a mixture of volumes from different sequences. Our investigation of the size of the training set revealed that training a U-Net in this domain only requires a modest amount of training data and results obtained with 49 and 70 training datasets were not significantly different.

CONCLUSIONS

To summarize, we investigated the use of 2D U-Nets and 3D U-Nets for breast volume segmentation in T1 fat-suppressed and without fat-suppressed volumes. Although our highest score was obtained in the 3D MULTI study, when we took advantage of information in both fat-suppressed and nonfat-suppressed volumes and their 3D structure, all of the methods we explored gave accurate segmentations with an average DSC on >94% demonstrating that the U-Net is a robust segmentation method for breast MRI volumes.

A graph-based lesion characterization and deep embedding approach for improved computer-aided diagnosis of nonmass breast MRI lesions

Nonmass-like enhancements are a common but diagnostically challenging finding in breast MRI. Nonmass-like lesions can be described as clusters of spatially and temporally inter-connected regions of enhancements, so they can be modeled as networks and their properties characterized via network-based connectivity. In this work, we represented nonmass lesions as graphs using a link formation energy model that favors linkages between regions of similar enhancement and closer spatial proximity. However, adding graph features to an existing computer-aided diagnosis (CAD) pipeline incurs an increase of feature space dimensionality, which poses additional challenges to traditional supervised machine learning techniques due to the inability to increase accordingly the number of training datasets. We propose the combination of unsupervised dimensionality reduction and embedded space clustering followed by a supervised classifier to improve the performance of a CAD system for nonmass-like lesions in breast MRI. Our work extends a previoulsy proposed framework for deep embedded unsupervised clustering (DEC) to embedding space classification, with the joint optimization of objective functions for DEC and supervised multi-layered perceptron (MLP) classification. The strength of the method lies in the ability to learn and further optimize an embedded feature representation of lower dimensionality that maximizes the diagnostic accuracy of a CAD lesion classifier to discriminate between benign and malignant lesions. We identified 792 nonmass-like enhancements (267 benign, 110 malignant and 415 unknown) in 411 patients undergoing breast MRI at our institution. The diagnostic performance of the proposed method was evaluated and compared to the performance of a conventional supervised MLP classifier in original feature space. A statistically significant increase in diagnostic area under the ROC curve (AUC) was achieved. Generalization AUC increased from 0.67 ± 0.08 to 0.81 ± 0.10 (21% increase, p-value=4.2×10^-8) with the proposed graph-based lesion characterization and deep embedding framework.

Anatomically plausible models and quality assurance criteria for online mono- and multi-modal medical image registration

Medical imaging is currently employed in the diagnosis, planning, delivery and response monitoring of cancer treatments. Due to physiological motion and/or treatment response, the shape and location of the pathology and organs-at-risk may change over time. Establishing their location within the acquired images is therefore paramount for an accurate treatment delivery and monitoring. A feasible solution for tracking anatomical changes during an image-guided cancer treatment is provided by image registration algorithms. Such methods are, however, often built upon elements originating from the computer vision/graphics domain. Since the original design of such elements did not take into consideration the material properties of particular biological tissues, the anatomical plausibility of the estimated deformations may not be guaranteed. In the current work we adapt two existing variational registration algorithms, namely Horn-Schunck and EVolution, to online soft tissue tracking. This is achieved by enforcing an incompressibility constraint on the estimated deformations during the registration process. The existing and the modified registration methods were comparatively tested against several quality assurance criteria on abdominal in vivo MR and CT data. These criteria included: the Dice similarity coefficient (DSC), the Jaccard index, the target registration error (TRE) and three additional criteria evaluating the anatomical plausibility of the estimated deformations. Results demonstrated that both the original and the modified registration methods have similar registration capabilities in high-contrast areas, with DSC and Jaccard index values predominantly in the 0.8-0.9 range and an average TRE of 1.6-2.0 mm. In contrast-devoid regions of the liver and kidneys, however, the three additional quality assurance criteria have indicated a considerable improvement of the anatomical plausibility of the deformations estimated by the incompressibility-constrained methods. Moreover, the proposed registration models maintain the potential of the original methods for online image-based guidance of cancer treatments.

Use of image registration and fusion algorithms and techniques in radiotherapy: Report of the AAPM Radiation Therapy Committee Task Group No. 132

Image registration and fusion algorithms exist in almost every software system that creates or uses images in radiotherapy. Most treatment planning systems support some form of image registration and fusion to allow the use of multimodality and time-series image data and even anatomical atlases to assist in target volume and normal tissue delineation. Treatment delivery systems perform registration and fusion between the planning images and the in-room images acquired during the treatment to assist patient positioning. Advanced applications are beginning to support daily dose assessment and enable adaptive radiotherapy using image registration and fusion to propagate contours and accumulate dose between image data taken over the course of therapy to provide up-to-date estimates of anatomical changes and delivered dose. This information aids in the detection of anatomical and functional changes that might elicit changes in the treatment plan or prescription. As the output of the image registration process is always used as the input of another process for planning or delivery, it is important to understand and communicate the uncertainty associated with the software in general and the result of a specific registration. Unfortunately, there is no standard mathematical formalism to perform this for real-world situations where noise, distortion, and complex anatomical variations can occur. Validation of the software systems performance is also complicated by the lack of documentation available from commercial systems leading to use of these systems in undesirable 'black-box' fashion. In view of this situation and the central role that image registration and fusion play in treatment planning and delivery, the Therapy Physics Committee of the American Association of Physicists in Medicine commissioned Task Group 132 to review current approaches and solutions for image registration (both rigid and deformable) in radiotherapy and to provide recommendations for quality assurance and quality control of these clinical processes.

Non-Parametric Bayesian Registration (NParBR) of Body Tumors in DCE-MRI Data

The identification of tumors in the internal organs of chest, abdomen, and pelvis anatomic regions can be performed with the analysis of Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) data. The contrast agent is accumulated differently by pathologic and healthy tissues and that results in a temporally varying contrast in an image series. The internal organs are also subject to potentially extensive movements mainly due to breathing, heart beat, and peristalsis. This contributes to making the analysis of DCE-MRI datasets challenging as well as time consuming. To address this problem we propose a novel pairwise non-rigid registration method with a Non-Parametric Bayesian Registration (NParBR) formulation. The NParBR method uses a Bayesian formulation that assumes a model for the effect of the distortion on the joint intensity statistics, a non-parametric prior for the restored statistics, and also applies a spatial regularization for the estimated registration with Gaussian filtering. A minimally biased intra-dataset atlas is computed for each dataset and used as reference for the registration of the time series. The time series registration method has been tested with 20 datasets of liver, lungs, intestines, and prostate. It has been compared to the B-Splines and to the SyN methods with results that demonstrate that the proposed method improves both accuracy and efficiency.

Registration of abdominal tumor DCE-MRI data based on deconvolution of joint statistics

The analysis of Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) data of body tumors presents several challenges. The accumulation of contrast agent in tissues results in a temporally varying contrast in an image series. At the same time, the body regions are subject to potentially extensive motion mainly due to breathing, heart beat, and peristalsis. This complicates any further automated analysis of a DCE-MRI time series such as for tumor lesion segmentation and volumetry. To address this problem we propose a novel effective non-rigid registration method based on the restoration of the joint statistics of pairs of images in the time series. Every image in the time series is registered to a reference one from the contrast enhanced phase. The pairwise registration is performed with deconvolution of the joint statistics, forcing the results back to the spatial domain and regularizing them with Gaussian spatial smoothing. The registration method has been validated with both a simulated phantom as well as real datasets with improved results for both its accuracy and efficiency.

A Metric for Reducing False Positives in the Computer-Aided Detection of Breast Cancer from Dynamic Contrast-Enhanced Magnetic Resonance Imaging Based Screening Examinations of High-Risk Women

Magnetic resonance imaging (MRI)-enabled cancer screening has been shown to be a highly sensitive method for the early detection of breast cancer. Computer-aided detection systems have the potential to improve the screening process by standardizing radiologists to a high level of diagnostic accuracy. This retrospective study was approved by the institutional review board of Sunnybrook Health Sciences Centre. This study compares the performance of a proposed method for computer-aided detection (based on the second-order spatial derivative of the relative signal intensity) with the signal enhancement ratio (SER) on MRI-based breast screening examinations. Comparison is performed using receiver operating characteristic (ROC) curve analysis as well as free-response receiver operating characteristic (FROC) curve analysis. A modified computer-aided detection system combining the proposed approach with the SER method is also presented. The proposed method provides improvements in the rates of false positive markings over the SER method in the detection of breast cancer (as assessed by FROC analysis). The modified computer-aided detection system that incorporates both the proposed method and the SER method yields ROC results equal to that produced by SER while simultaneously providing improvements over the SER method in terms of false positives per noncancerous exam. The proposed method for identifying malignancies outperforms the SER method in terms of false positives on a challenging dataset containing many small lesions and may play a useful role in breast cancer screening by MRI as part of a computer-aided detection system.

Semi-automatic region-of-interest segmentation based computer-aided diagnosis of mass lesions from dynamic contrast-enhanced magnetic resonance imaging based breast cancer screening

Cancer screening with magnetic resonance imaging (MRI) is currently recommended for very high risk women. The high variability in the diagnostic accuracy of radiologists analyzing screening MRI examinations of the breast is due, at least in part, to the large amounts of data acquired. This has motivated substantial research towards the development of computer-aided diagnosis (CAD) systems for breast MRI which can assist in the diagnostic process by acting as a second reader of the examinations. This retrospective study was performed on 184 benign and 49 malignant lesions detected in a prospective MRI screening study of high risk women at Sunnybrook Health Sciences Centre. A method for performing semi-automatic lesion segmentation based on a supervised learning formulation was compared with the enhancement threshold based segmentation method in the context of a computer-aided diagnostic system. The results demonstrate that the proposed method can assist in providing increased separation between malignant and radiologically suspicious benign lesions. Separation between malignant and benign lesions based on margin measures improved from a receiver operating characteristic (ROC) curve area of 0.63 to 0.73 when the proposed segmentation method was compared with the enhancement threshold, representing a statistically significant improvement. Separation between malignant and benign lesions based on dynamic measures improved from a ROC curve area of 0.75 to 0.79 when the proposed segmentation method was compared to the enhancement threshold, also representing a statistically significant improvement. The proposed method has potential as a component of a computer-aided diagnostic system.

A vector machine formulation with application to the computer-aided diagnosis of breast cancer from DCE-MRI screening examinations

This study investigates the use of a proposed vector machine formulation with application to dynamic contrast-enhanced magnetic resonance imaging examinations in the context of the computer-aided diagnosis of breast cancer. This paper describes a method for generating feature measurements that characterize a lesion's vascular heterogeneity as well as a supervised learning formulation that represents an improvement over the conventional support vector machine in this application. Spatially varying signal-intensity measures were extracted from the examinations using principal components analysis and the machine learning technique known as the support vector machine (SVM) was used to classify the results. An alternative vector machine formulation was found to improve on the results produced by the established SVM in randomized bootstrap validation trials, yielding a receiver-operating characteristic curve area of 0.82 which represents a statistically significant improvement over the SVM technique in this application.

Bi-modal Non-rigid Registration of Brain MRI Data with Deconvolution of Joint Statistics

A brain MRI protocol typically includes several imaging contrasts that can provide complementary information by highlighting different tissue properties. The acquired datasets often need to be co-registered or placed in a standard anatomic space before any further processing. Current registration methods particularly for multicontrast data are computationally very intensive, their resolution is lower than that of the images, and their distance metric and its optimization can be limiting. In this work a novel and effective non-rigid registration method is proposed that is based on the restoration of the joint statistics of pairs of such images. The registration is performed with the deconvolution of the joint statistics with an adaptive Wiener filter. The deconvolved statistics are forced back to the spatial domain to estimate a preliminary registration. The spatial transformation is also regularized with Gaussian spatial smoothing. The registration method has been compared with the B-Splines method implemented in 3DSlicer and with the SyN method implemented in the ANTs toolkit. The validation has been performed with a simulated Shepp-Logan phantom, a BrainWeb phantom, the real data of the NIREP database, and real multi-contrast datasets of healthy volunteers. The proposed method has shown improved comparative accuracy as well as analytical efficiency.

Deformable registration for quantifying longitudinal tumor changes during neoadjuvant chemotherapy

PURPOSE

To evaluate DRAMMS, an attribute-based deformable registration algorithm, compared to other intensity-based algorithms, for longitudinal breast MRI registration, and to show its applicability in quantifying tumor changes over the course of neoadjuvant chemotherapy.

METHODS

Breast magnetic resonance images from 14 women undergoing neoadjuvant chemotherapy were analyzed. The accuracy of DRAMMS versus five intensity-based deformable registration methods was evaluated based on 2,380 landmarks independently annotated by two experts, for the entire image volume, different image subregions, and patient subgroups. The registration method with the smallest landmark error was used to quantify tumor changes, by calculating the Jacobian determinant maps of the registration deformation.

RESULTS

DRAMMS had the smallest landmark errors (6.05 ± 4.86 mm), followed by the intensity-based methods CC-FFD (8.07 ± 3.86 mm), NMI-FFD (8.21 ± 3.81 mm), SSD-FFD (9.46 ± 4.55 mm), Demons (10.76 ± 6.01 mm), and Diffeomorphic Demons (10.82 ± 6.11 mm). Results show that registration accuracy also depends on tumor versus normal tissue regions and different patient subgroups.

CONCLUSIONS

The DRAMMS deformable registration method, driven by attribute-matching and mutual-saliency, can register longitudinal breast magnetic resonance images with a higher accuracy than several intensity-matching methods included in this article. As such, it could be valuable for more accurately quantifying heterogeneous tumor changes as a marker of response to treatment.

Image registration for quantitative parametric response mapping of cancer treatment response

Imaging biomarkers capable of early quantification of tumor response to therapy would provide an opportunity to individualize patient care. Image registration of longitudinal scans provides a method of detecting treatment associated changes within heterogeneous tumors by monitoring alterations in the quantitative value of individual voxels over time, which is unattainable by traditional volumetric-based histogram methods. The concepts involved in the use of image registration for tracking and quantifying breast cancer treatment response using parametric response mapping (PRM), a voxel-based analysis of diffusion-weighted magnetic resonance imaging (DW-MRI) scans, are presented. Application of PRM to breast tumor response detection is described, wherein robust registration solutions for tracking small changes in water diffusivity in breast tumors during therapy are required. Methodologies that employ simulations are presented for measuring expected statistical accuracy of PRM for response assessment. Test-retest clinical scans are used to yield estimates of system noise to indicate significant changes in voxel-based changes in water diffusivity. Overall, registration-based PRM image analysis provides significant opportunities for voxel-based image analysis to provide the required accuracy for early assessment of response to treatment in breast cancer patients receiving neoadjuvant chemotherapy.

A Gauss-Newton approach to joint image registration and intensity correction

We develop a new efficient numerical methodology for automated simultaneous registration and intensity correction of images. The approach separates the intensity correction term from the images being registered in a regularized expression. Our formulation is consistent with the existing non-parametric image registration techniques, however, an extra additive intensity correction term is carried throughout. An objective functional is formed for which the corresponding Hessian and Jacobian is computed and employed in a multi-level Gauss-Newton minimization approach. In this paper, our experiments are based on elastic regularization on the transformation and total variation on the intensity correction. Validations on dynamic contrast enhanced MR abdominal images for both real and simulated data verified the efficacy of the model. The pursued approach is flexible in which we can exploit various forms of regularization on the transformation and the intensity correction.

Feature selection in computer-aided breast cancer diagnosis via dynamic contrast-enhanced magnetic resonance images

The accuracy of computer-aided diagnosis (CAD) for early detection and classification of breast cancer in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is dependent upon the features used by the CAD classifier. Here, we show that fast orthogonal search (FOS), which provides a more efficient iterative manner of computing stepwise regression feature selection, can select features with predictive value from a set of kinetic and texture candidate features computed from dynamic contrast-enhanced magnetic resonance images. FOS can in minutes search candidate feature sets of millions of terms, which may include cross-products of features up to second-, third- or fourth-order. This method is tested on a set of 83 DCE-MRI images, of which 20 are for cancerous and 63 for benign cases, using a leave-one-out trial. The features selected by FOS were used in a FOS predictor and nearest-neighbour predictor and had an area under the receiver operating curve (AUC) of 0.889 and 0.791 respectively. The FOS predictor AUC is significantly improved over the signal enhancement ratio predictor with an AUC of 0.706 (p = 0.0035 for the difference in the AUCs). Moreover, using FOS-selected features in a support vector machine increased the AUC over that resulting when the features were manually selected.

A measure to evaluate deformable registration fields in clinical settings

Deformable registration has migrated from a research topic to a widely used clinical tool that can improve radiotherapeutic treatment accuracy by tracking anatomical changes. Although various mathematical formulations have been reported in the literature and implemented in commercial software, we lack a straightforward method to verify a given solution in routine clinical use. We propose a metric using concepts derived from vector analysis that complements the standard evaluation tools to identify unrealistic wrappings in a displacement field. At the heart of the proposed procedure is identification of vortexes in the displacement field that do not correspond to underlying anatomical changes. Vortexes are detected and their intensity quantified using the CURL operator and presented as a vortex map overlaid on the original anatomy for rapid identification of problematic regions. We show application of the proposed metric on clinical scenarios of adaptive radiotherapy and treatment response assessment, where the CURL operator quantitatively detected errors in the displacement field and identified problematic regions that were invisible to classical voxel‐based evaluation methods. Unrealistic warping not visible to standard voxel‐based solution assessment can produce erroneous results when the deformable solution is applied on a secondary dataset, such as dose matrix in adaptive therapy or PET data for treatment response assessment. The proposed metric for evaluating deformable registration provides increased usability and accuracy of detecting unrealistic deformable registration solutions when compared to standard intensity‐based approaches. It is computationally efficient and provides a valuable platform for the clinical acceptance of image‐guided radiotherapy.

PACS numbers: 87.57.nj; 87.55.Qr; 87.57.cp

Clinical application of pharmacokinetic analysis as a biomarker of solitary pulmonary nodules: dynamic contrast-enhanced MR imaging

The purpose of this study is to evaluate perfusion indices and pharmacokinetic parameters in solitary pulmonary nodules (SPNs). Thirty patients of 34 enrolled with SPNs (15-30 mm) were evaluated in this study. T1 and T2-weighted structural images and 2D turbo FLASH perfusion images were acquired with shallow free breathing. B-spline nonrigid image registration and optimization by χ² test against pharmacokinetic model curve were performed on dynamic contrast-enhanced MRI. This allowed voxel-by-voxel calculation of k(ep) , the rate constant for tracer transport to and from plasma and the extravascular extracellular space. Mean transit time, time-to-peak, initial slope, and maximum enhancement (E(max) ) were calculated from time-intensity curves fitted to a gamma variate function. After blinded data analysis, correlation with tissue histology from surgical resection or biopsy samples was performed. Histologic evaluation revealed 25 malignant and five benign SPNs. All benign SPNs had k(ep) < 1.0 min⁻¹. Nineteen of 25 (76%) malignant SPNs showed k(ep) > 1.0 min⁻¹. Sensitivity to diagnose malignant SPNs at a cutoff of k(ep) = 1.0 min⁻¹ was 76%, specificity was 100%, positive predictive value was 100%, negative predictive value was 45%, and accuracy was 80%. Of all indices studied, k(ep) was the most significant in differentiating malignant from benign SPNs.

A margin sharpness measurement for the diagnosis of breast cancer from magnetic resonance imaging examinations

RATIONALE AND OBJECTIVES

Cancer screening by magnetic resonance imaging (MRI) has been shown to be one of the most sensitive methods available for the early detection of breast cancer. There is high variability in the diagnostic accuracy of radiologists analyzing the large amounts of data acquired in a breast MRI examination, and this has motivated substantial research toward the development of computer-aided detection and diagnosis systems. Most computer-aided diagnosis systems for breast MRI focus on dynamic information (how a lesion's brightness changes over the course of an examination after the injection of a contrast agent). The inclusion of lesion margin measurements is much less common. One characteristic of malignant tumors is that they grow into neighboring tissues. This growth creates tumor margins that are variably fuzzy or diffuse (ie, they are not sharp).

MATERIALS AND METHODS

In this short report, the authors present a new method for measuring a tumor's margin from breast MRI examinations and compare it with an existing mathematical technique for margin measurements.

RESULTS

The proposed method can yield a test with sensitivity of 77% (specificity, 65%) on screening data, outperforming existing mathematical lesion margin measurement methods. Furthermore, when the presented margin measurement is combined with existing dynamic features, there is a statistically significant improvement in computer-aided diagnosis test performance (P < .0014).

CONCLUSIONS

The proposed method for measuring a tumor's margin outperforms existing mathematical methods on an extremely challenging data set containing many small lesions. The technique presented may be useful in discriminating between malignant and benign lesions in the context of the computer-aided diagnosis of breast cancer from MRI.

The effect of motion correction on pharmacokinetic parameter estimation in dynamic-contrast-enhanced MRI

A dynamic-contrast-enhanced magnetic resonance imaging (DCE-MRI) dataset consists of many imaging frames, often acquired both before and after contrast injection. Due to the length of time spent acquiring images, patient motion is likely and image re-alignment or registration is required before further analysis such as pharmacokinetic model fitting. Non-rigid image registration procedures may be used to correct motion artefacts; however, a careful choice of registration strategy is required to reduce misregistration artefacts associated with enhancing features. This work investigates the effect of registration on the results of model-fitting algorithms for 52 DCE-MR mammography cases for 14 patients. Results are divided into two sections: a comparison of registration strategies in which a DCE-MRI-specific algorithm is preferred in 50% of cases, followed by an investigation of parameter changes with known applied deformations, inspecting the effect of magnitude and timing of motion artefacts. Increased motion magnitude correlates with increased model-fit residual and is seen to have a strong influence on the visibility of strongly enhancing features. Motion artefacts in images close to the contrast agent arrival have a disproportionate effect on discrepancies in parameter estimation. The choice of algorithm, magnitude of motion and timing of the motion are each shown to influence estimated pharmacokinetic parameters even when motion magnitude is small.

Supine breast MRI

PURPOSE

To achieve high-quality unilateral supine breast magnetic resonance imaging (MRI) as a step to facilitate image aiding of clinical applications, which are often performed in the supine position. Contrast-enhanced breast MRI is a powerful tool for the diagnosis of cancer. However, prone patient positioning typically used for breast MRI hinders its use for image aiding.

MATERIALS AND METHODS

A fixture and a flexible four-element receive coil were designed for patient-specific shaping and placement of the coil in close conformity to the supine breast. A 3D spoiled gradient sequence was modified to incorporate compensation of respiratory motion. The entire setup was tested in volunteer experiments and in a pilot patient study.

RESULTS

The flexible coil design and the motion compensation produced supine breast MR images of high diagnostic value. Variations in breast shape and in tissue morphology within the breast were observed between a supine and a diagnostic prone MRI of a patient.

CONCLUSION

The presented supine breast MRI achieved an image quality comparable to diagnostic breast MRI. Since supine positioning is common in many clinical applications such as ultrasound-guided breast biopsy or breast-conserving surgery, the registration of the supine images will aid these applications.

Impact of nonrigid motion correction technique on pixel-wise pharmacokinetic analysis of free-breathing pulmonary dynamic contrast-enhanced MR imaging

PURPOSE

To investigates the impact of nonrigid motion correction on pixel-wise pharmacokinetic analysis of free-breathing DCE-MRI in patients with solitary pulmonary nodules (SPNs). Misalignment of focal lesions due to respiratory motion in free-breathing dynamic contrast-enhanced MRI (DCE-MRI) precludes obtaining reliable time-intensity curves, which are crucial for pharmacokinetic analysis for tissue characterization.

MATERIALS AND METHODS

Single-slice 2D DCE-MRI was obtained in 15 patients. Misalignments of SPNs were corrected using nonrigid B-spline image registration. Pixel-wise pharmacokinetic parameters K(trans) , v(e) , and k(ep) were estimated from both original and motion-corrected DCE-MRI by fitting the two-compartment pharmacokinetic model to the time-intensity curve obtained in each pixel. The "goodness-of-fit" was tested with χ(2) -test in pixel-by-pixel basis to evaluate the reliability of the parameters. The percentages of reliable pixels within the SPNs were compared between the original and motion-corrected DCE-MRI. In addition, the parameters obtained from benign and malignant SPNs were compared.

RESULTS

The percentage of reliable pixels in the motion-corrected DCE-MRI was significantly larger than the original DCE-MRI (P = 4 × 10(-7) ). Both K(trans) and k(ep) derived from the motion-corrected DCE-MRI showed significant differences between benign and malignant SPNs (P = 0.024, 0.015).

CONCLUSION

The study demonstrated the impact of nonrigid motion correction technique on pixel-wise pharmacokinetic analysis of free-breathing DCE-MRI in SPNs.

Computer-aided diagnosis of breast cancer from magnetic resonance imaging examinations by custom radial basis function vector machine

This paper presents a new method for performing supervised learning (classification) and demonstrates the technique by applying it to the detection of breast cancer from the dynamic information obtained in magnetic resonance imaging examinations. The proposed method is a vector machine similar to the established support vector machine (SVM) method, however, our method involves a reformulation of the classification/prediction process. The proposed classification methodology is compared with the SVM, with both methods using the established radial basis function kernel. The proposed vector machine formulation applies test biasing in a new manner and is demonstrated to produce robust solutions as measured by the receiver operating characteristic (ROC) curve area. The technique is compared with SVMs and yields test improvements up to an additional 9.8% sensitivity or 7.2% specificity.

A general PDE-framework for registration of contrast enhanced images

This paper presents a general PDE-framework for registration of contrast enhanced images. The approach directly applies the idea of separating the contrast enhancement term from the images in the regularization terms. In our formulation, we stay consistent with existing non-parametric image registration techniques, however, we carry an additional contrast enhancement term throughout. A mathematically rigorous approach is pursued which can exploit various forms of regularization. In this paper, our experiments are built based on diffusion regularization for both contrast enhancement and the deformation field.

Effect of the enhancement threshold on the computer-aided detection of breast cancer using MRI

RATIONALE AND OBJECTIVES

To evaluate the effect that variations in the enhancement threshold have on the diagnostic accuracy of two computer-aided detection (CAD) systems for magnetic resonance based breast cancer screening.

MATERIALS AND METHODS

Informed consent was obtained from all patients participating in cancer screening and this study was approved by the participating institution's review board. This retrospective study was nested in a prospective, single-institution, high-risk, breast screening study involving dynamic contrast-enhanced magnetic resonance imaging. Only those screening examinations (n = 223) for which a histopathological diagnosis was available were included. Two CAD methods were performed: the signal enhancement ratio (SER) and support vector machines (SVMs). Statistical analysis was performed by tracking changes in each CAD test's diagnostic accuracy (eg, receiver-operating characteristic [ROC] curve area, maximum possible sensitivity) with changes in the enhancement threshold.

RESULTS

The enhancement threshold plays a significant role in affecting a CAD test's potential sensitivity, ROC curve area, and number of assumed true and false-positive predictions per cancerous examination. A high threshold can also limit the CAD-based detection of the full size of a lesion.

CONCLUSIONS

Enhancement thresholds can limit a CAD test's ability to diagnose a lesion's full size and as such should not be raised above 60%. The clinically used SER method exhibits a high rate of false positives at low enhancement thresholds and as such the threshold should not be set lower than 50%. The SVM method yielded better results in our study than the SER method at clinically realistic enhancement thresholds.

Classification of dynamic contrast-enhanced magnetic resonance breast lesions by support vector machines

Early detection of breast cancer is one of the most important factors in determining prognosis for women with malignant tumors. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has been shown to be the most sensitive modality for screening high-risk women. Computer-aided diagnosis (CAD) systems have the potential to assist radiologists in the early detection of cancer. A key component of the development of such a CAD system will be the selection of an appropriate classification function responsible for separating malignant and benign lesions. The purpose of this study is to evaluate the effects of variations in temporal feature vectors and kernel functions on the separation of malignant and benign DCE-MRI breast lesions by support vector machines (SVMs). We also propose and demonstrate a classifier visualization and evaluation technique. We show that SVMs provide an effective and flexible framework from which to base CAD techniques for breast MRI, and that the proposed classifier visualization technique has potential as a mechanism for the evaluation of classification solutions.

Efficient computational fluid dynamics mesh generation by image registration

Most implementations of computational fluid dynamics (CFD) solutions require a discretisation or meshing of the solution domain. The production from a medical image of a computationally efficient mesh representing the structures of interest can be time consuming and labour-intensive, and remains a major bottleneck in the clinical application of CFD. This paper presents a method for deriving a patient-specific mesh from a medical image. The method uses volumetric registration of a pseudo-image, produced from an idealised template mesh, with the medical image. The registration algorithm used is robust and computationally efficient. The accuracy of the new algorithm is measured in terms of the distance between a registered surface and a known surface, for image data derived from casts of the lumen of two different vessels. The true surface is identified by laser profiling. The average distance between the surface points measured by the laser profiler and the surface of the mapped mesh is better than 0.2 mm. For the images analysed, the new algorithm is shown to be 2-3 times more accurate than a standard published algorithm based on maximising normalised mutual information. Computation times are approximately 18 times faster for the new algorithm than the standard algorithm. Examples of the use of the algorithm on two clinical examples are also given. The registration methodology lends itself immediately to the construction of dynamic mesh models in which vessel wall motion is obtained directly using registration.