OBJECTIVE:This paper proposes a novel approach for automatic left ventricle (LV) quantification using convolutional neural networks (CNN).METHODS:The general framework consists of one CNN for detecting the LV, and another for tissue classification. Also, three new deep learning architectures were proposed for LV quantification. These new CNNs introduce the ideas of sparsity and depthwise separable convolution into the U-net architecture, as well as, a residual learning strategy level-to-level. To this end, we extend the classical U-net architecture and use the generalized Jaccard distance as optimization objective function.RESULTS:The CNNs were trained and evaluated with 140 patients from two public cardiovascular magnetic resonance datasets (Sunnybrook and Cardiac Atlas Project) by using a 5-fold cross-validation strategy. Our results demonstrate a suitable accuracy for myocardial segmentation ( ∼ 0.9 Dice's coefficient), and a strong correlation with the most relevant physiological measures: 0.99 for end-diastolic and end-systolic volume, 0.97 for the left myocardial mass, 0.95 for the ejection fraction and 0.93 for the stroke volume and cardiac output.CONCLUSION:Our simulation and clinical evaluation results demonstrate the capability and merits of the proposed CNN to estimate different structural and functional features such as LV mass and EF which are commonly used for both diagnosis and treatment of different pathologies.SIGNIFICANCE:This paper suggests a new approach for automatic LV quantification based on deep learning where errors are comparable to the inter- and intra-operator ranges for manual contouring.

We re-formulate a classical numerical method for the solution of systems of linear equations to tackle problems with latent data, that is, linear systems of dimension that is a priori unknown. This type of systems appears in the solution of few-body Coulomb problems for Atomic Simulation Physics, in the form of multidimensional partial differential equations (PDEs) that require the numerical solution of a sequence of recurrent dense linear systems of growing scale. The large dimension of these systems, with up to several hundred thousands of unknowns, is tackled in our approach via a task-parallel implementation of a solver based on the QR factorization. This method is parallelized using the OmpSs framework, showing fair strong and weak scalability on a multicore processor equipped with 12 Intel cores.

In this work we present the analysis of the numerical performance and accuracy of the Monte Carlo Event Generator method applied to the double ionization of Neon by fast electrons. This implementation of the code runs in Graphics Processing Units, giving a clear advantage over other computational approaches. The code can provide five hundred cross sections per second with the optimal configuration, an order of magnitude faster than CPU counterparts.

Cardiac function is of paramount importance for both prognosis and treatment of different pathologies such as mitral regurgitation, ischemia, dyssynchrony and myocarditis. Cardiac behavior is determined by structural and functional features. In both cases, the analysis of medical imaging studies requires to detect and segment the myocardium. Nowadays, magnetic resonance imaging (MRI) is one of the most relevant and accurate non-invasive diagnostic tools for cardiac structure and function. In this work we propose to use a deep learning technique to assist the automatization of myocardial segmentation in cardiac MRI. We present several improvements to previous works in this paper: we propose to use the Jaccard distance as optimization objective function, we integrate a residual learning strategy into the code, and we introduce a batch normalization layer to train the fully convolutional neural network. Our results demonstrate that this architecture outperforms previous approaches based on a similar network architecture, and that provides a suitable approach for myocardial segmentation. Our benchmark shows that the automatic myocardial segmentation takes less than 22 seg. for a volume of 128~x~128~x~13 pixels in a 3.1 GHz intel core i7.

An algorithm based on Convolution/Superposition with collapsed cone approximation was developed for radiotherapy dose calculation, reducing numerical complexity and enabling a high accuracy computation in a dense grid. By analyzing the specific integrals and reducing them into a ray tracing problem, we show that both calculation and data evaluations can be mapped to specific and optimized memories types in the GPU. Using constant memory and texture fetches in the algorithm, an 144X speedup is obtained compared to an equivalent multi-threaded CPU code, without precision loss. The developed software is the foundation for a high performance calculation system with a fidelity equivalent to commercial planning systems and with a few seconds of execution.

We study ionization of Beryllium atoms by ion impact with continuum distorted waves models. We calculate cross sections in the intermediate to high energy regime for state resolved and total ionization.

In this work we study the effect of the confinement of H atom in a fullerene cage for the 1s→2p resonant transition of the atom interacting with a laser pulse. We present a novell implementation of the Generalized Sturmian Functions in the time-dependent frame to numerically solve this problem calculating the population of bound states and spectral density.

Continuum Distorted Wave methods are adapted to new heterogeneous computing architectures, exhibiting excellent speed ups in performance.

We compute fully differential cross sections for double ionization of helium by electrons, within the high-impact- energy and low-momentum transfer regimes, using the generalized Sturmian functions approach. Our results are converged relative to the total angular momentum and variable domain size. The method shows very good agreement with convergent close coupling calculations performed by Kheifets et al. [J. Phys. B 32, 5047 (1999)] for all ejection angles for the two electron emission energies considered in the experiments reported in that contribution. Both theoretical methods provide fully differential cross sections that require the same upscaling factors to compare with experimental data and are based on a first-order Born model for the projectile-target interaction. Since that reference was published, there were several theoretical efforts to account for the absolute scale of the experimental results, but agreement in the cross-section magnitude was not achieved even between theories. With the present contribution we conclude that the first-order Born model is now adequately solved, shifting the magnitude controversy towards either the experimental data and/or the addition of higher degrees of projectile-target interaction to the calculation.

The Generalized Sturmian Functions method aims to deal with atomic physics problems. It has seen application to two and three-body problems, and its flexibility enables one to work with bound systems as well as with particles in the continuum. In the present contribution we analyze how the method expands the atomic double continuum in collision problems, using the double ionization of Helium by fast electrons as a showcase. We first test the robustness of the method in a particularly challenging situation, the zero energy case. We then present fully differential cross sections for a scattering problem which after 15 years of continued efforts has not been satisfactorily solved: the double ionization of Helium by electron impact in the fast projectile regime, as measured by the Orsay group.

In this work we study the effect of fullerene confinement on the triple differential cross section (TDCS) for double ionization of atoms. We compare the results corresponding to the free case with single cage C60 and two cage C60@C240 confinement.

In this work we study the double photoionization of helium induced by low intensities laser fields in the regime where only one photon absorption occurs. The method proposed here is based on a Generalized Sturmian Functions (GSF) spectral approach which allows the imposition of outgoing boundary conditions for both ejected electrons. These, in turn, construct an hyperspherical flux characteristic of double continuum wave functions. We compare our calculated cross sections at 20 and 40 eV above threshold with absolute and relative measurements, and with other calculations. Our results definitively demonstrate the applicability of the GSF approach for dealing with break-up Coulomb problems.

We present fully differentiated cross sections for the Helium double ionization by neutronic impact. From our theoretical results we observe the underlying mechanisms that lead to the breakup of the target.

Perturbative driven equation series for single and double ionization by n-photon absorption of atoms are solved in a stationary scattering framework with Generalized Sturmian Function expansions.

In this work we investigate the ionization by proton impact, and photo ionization, involving highly charged ions of W, in particular the Ne-like W64+ and Na-like W63+ which are more likely to appear in ITER plasma. Total cross sections are calculated in the continuum-distorted-wave-eikonal-initial-state (CDW-EIS) approximation for ion-impact, while two completely different methods are used for photoionization: a perturbative dipolar aproach and the recently introduced Sturmian model.

The double ionization of helium by high energy electron impact is studied. The corresponding four-body Schrodinger equation is transformed into a set of driven equations containing successive orders in the projectile-target interaction. The first order driven equation is solved with a generalized Sturmian functions approach. The transition amplitude, extracted from the asymptotic limit of the first order solution, is equivalent to the familiar first Born approximation. Fivefold differential cross sections are calculated for (e, 3e) processes within the high incident energy and small momentum transfer regimes. The results are compared with other numerical methods, and with the only absolute experimental data available. Our cross sections agree in shape and magnitude with those of the convergent close coupling method for the (10+10) eV and (4+4) eV emission energies. To date this had not been achieved by any two different numerical schemes when solving the three-body continuum problem for the fast projectile (e, 3e) process. Though agreement with the experimental data, in particular with respect to the magnitude, is not achieved, our findings partly clarify a long standing puzzle.

We present a computation method to accelerate the calculation of the Hamiltonian of a three-body time independent Schrödinger equation for collisions.The Hamiltonian is constructed with one dimensional (basis overlaps) and two dimensional (interparticle interaction) integrals that are mapped into a computational grid in a Graphics Processing Unit (GPU).We illustrate the method for the case of an electron impact single ionization of a two electron atom.This proposal makes use of a Generalized Sturmian Basis set for each electron, which are obtained numerically on a quadrature grid that is used to compute the integrals in the GPU.The optimal computation is more than twenty times faster in the GPU than the calculation in CPU.The method can be easily scaled to computers with several Graphics Processing Units or clusters.

The double ionization of helium by high energy electron impact is investigated. The pure four-body Coulomb problem may be reduced to a three-body one in accordance with the use of the first Born approximation. Even within this frame, major unexplained discrepancies subsist between several theoretical descriptions, and with available absolute experimental data. In this contribution we discuss an alternative formulation which allows to tackle the problem with a different methodology, the generalized Sturmian approach. We discuss some issues associated to the convergence of the calculated cross sections.

Three aspects of the Generalized Sturmian Function method are discussed, adding to the flexibility of the formulation.