A new framework for assessing subject-specific whole brain circulation and perfusion using MRI-based measurements and a multi-scale continuous flow model.

A large variety of severe medical conditions involve alterations in microvascular circulation. Hence, measurements or simulation of circulation and perfusion has considerable clinical value and can be used for diagnostics, evaluation of treatment efficacy, and for surgical planning. However, the accuracy of traditional tracer kinetic one-compartment models is limited due to scale dependency. As a remedy, we propose a scale invariant mathematical framework for simulating whole brain perfusion. The suggested framework is based on a segmentation of anatomical geometry down to imaging voxel resolution. Large vessels in the arterial and venous network are identified from time-of-flight (ToF) and quantitative susceptibility mapping (QSM). Macro-scale flow in the large-vessel-network is accurately modelled using the Hagen-Poiseuille equation, whereas capillary flow is treated as two-compartment porous media flow. Macro-scale flow is coupled with micro-scale flow by a spatially distributing support function in the terminal endings. Perfusion is defined as the transition of fluid from the arterial to the venous compartment. We demonstrate a whole brain simulation of tracer propagation on a realistic geometric model of the human brain, where the model comprises distinct areas of grey and white matter, as well as large vessels in the arterial and venous vascular network. Our proposed framework is an accurate and viable alternative to traditional compartment models, with high relevance for simulation of brain perfusion and also for restoration of field parameters in clinical brain perfusion applications.

Assessment of kidney volumes from MRI: acquisition and segmentation techniques.

OBJECTIVE: The prevalence of chronic kidney disease (CKD) is increasing worldwide. In Europe alone, at least 8% of the population currently has some degree of CKD. CKD is associated with serious comorbidity, reduced life expectancy, and high economic costs; hence, early detection and adequate treatment of kidney disease are important. CONCLUSION: We review state-of-the-art MRI acquisition techniques for CKD, with a special focus on image segmentation methods used for the estimation of kidney volume.

Epistemic uncertainty quantification in deep learning classification by the Delta method.

The Delta method is a classical procedure for quantifying epistemic uncertainty in statistical models, but its direct application to deep neural networks is prevented by the large number of parameters P. We propose a low cost approximation of the Delta method applicable to L2-regularized deep neural networks based on the top K eigenpairs of the Fisher information matrix. We address efficient computation of full-rank approximate eigendecompositions in terms of the exact inverse Hessian, the inverse outer-products of gradients approximation and the so-called Sandwich estimator. Moreover, we provide bounds on the approximation error for the uncertainty of the predictive class probabilities. We show that when the smallest computed eigenvalue of the Fisher information matrix is near the L2-regularization rate, the approximation error will be close to zero even when K≪P. A demonstration of the methodology is presented using a TensorFlow implementation, and we show that meaningful rankings of images based on predictive uncertainty can be obtained for two LeNet and ResNet-based neural networks using the MNIST and CIFAR-10 datasets. Further, we observe that false positives have on average a higher predictive epistemic uncertainty than true positives. This suggests that there is supplementing information in the uncertainty measure not captured by the classification alone.

Automated segmentation of endometrial cancer on MR images using deep learning.

Preoperative MR imaging in endometrial cancer patients provides valuable information on local tumor extent, which routinely guides choice of surgical procedure and adjuvant therapy. Furthermore, whole-volume tumor analyses of MR images may provide radiomic tumor signatures potentially relevant for better individualization and optimization of treatment. We apply a convolutional neural network for automatic tumor segmentation in endometrial cancer patients, enabling automated extraction of tumor texture parameters and tumor volume. The network was trained, validated and tested on a cohort of 139 endometrial cancer patients based on preoperative pelvic imaging. The algorithm was able to retrieve tumor volumes comparable to human expert level (likelihood-ratio test, [Formula: see text]). The network was also able to provide a set of segmentation masks with human agreement not different from inter-rater agreement of human experts (Wilcoxon signed rank test, [Formula: see text], [Formula: see text], and [Formula: see text]). An automatic tool for tumor segmentation in endometrial cancer patients enables automated extraction of tumor volume and whole-volume tumor texture features. This approach represents a promising method for automatic radiomic tumor profiling with potential relevance for better prognostication and individualization of therapeutic strategy in endometrial cancer.

Physical Models for Simulation and Reconstruction of Human Tissue Deformation Fields in Dynamic MRI.

OBJECTIVE: Medical image registration can be formulated as a tissue deformation problem, where parameter estimation methods are used to obtain the inverse deformation. However, there is limited knowledge about the ability to recover an unknown deformation. The main objective of this study is to estimate the quality of a restored deformation field obtained from image registration of dynamic MR sequences. METHODS: We investigate the behavior of forward deformation models of various complexities. Further, we study the accuracy of restored inverse deformations generated by image registration. RESULTS: We found that the choice of 1) heterogeneous tissue parameters and 2) a poroelastic (instead of elastic) model had significant impact on the forward deformation. In the image registration problem, both 1) and 2) were found not to be significant. Here, the presence of image features were dominating the performance. We also found that existing algorithms will align images with high precision while at the same time obtain a deformation field with a relative error of 40%. CONCLUSION: Image registration can only moderately well restore the true deformation field. Still, estimation of volume changes instead of deformation fields can be fairly accurate and may represent a proxy for variations in tissue characteristics. Volume changes remain essentially unchanged under choice of discretization and the prevalence of pronounced image features. SIGNIFICANCE: We suggest that image registration of high-contrast MR images has potential to be used as a tool to produce imaging biomarkers sensitive to pathology affecting tissue stiffness.

Normalized gradient fields for nonlinear motion correction of DCE-MRI time series.

Dynamic MR image recordings (DCE-MRI) of moving organs using bolus injections create two different types of dynamics in the images: (i) spatial motion artifacts due to patient movements, breathing and physiological pulsations that we want to counteract and (ii) signal intensity changes during contrast agent wash-in and wash-out that we want to preserve. Proper image registration is needed to counteract the motion artifacts and for a reliable assessment of physiological parameters. In this work we present a partial differential equation-based method for deformable multimodal image registration using normalized gradients and the Fourier transform to solve the Euler-Lagrange equations in a multilevel hierarchy. This approach is particularly well suited to handle the motion challenges in DCE-MRI time series, being validated on ten DCE-MRI datasets from the moving kidney. We found that both normalized gradients and mutual information work as high-performing cost functionals for motion correction of this type of data. Furthermore, we demonstrated that normalized gradients have improved performance compared to mutual information as assessed by several performance measures. We conclude that normalized gradients can be a viable alternative to mutual information regarding registration accuracy, and with promising clinical applications to DCE-MRI recordings from moving organs.

Segmentation-driven image registration- application to 4D DCE-MRI recordings of the moving kidneys.

Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) of the kidneys requires proper motion correction and segmentation to enable an estimation of glomerular filtration rate through pharmacokinetic modeling. Traditionally, co-registration, segmentation, and pharmacokinetic modeling have been applied sequentially as separate processing steps. In this paper, a combined 4D model for simultaneous registration and segmentation of the whole kidney is presented. To demonstrate the model in numerical experiments, we used normalized gradients as data term in the registration and a Mahalanobis distance from the time courses of the segmented regions to a training set for supervised segmentation. By applying this framework to an input consisting of 4D image time series, we conduct simultaneous motion correction and two-region segmentation into kidney and background. The potential of the new approach is demonstrated on real DCE-MRI data from ten healthy volunteers.