Adaptive Stochastic Gradient Descent Optimisation for Image Registration

We present a stochastic gradient descent optimisation method for image registration with adaptive step size prediction. The method is based on the theoretical work by Plakhov and Cruz (J. Math. Sci. 120(1):964–973, 2004). Our main methodological contribution is the derivation of an image-driven mechanism to select proper values for the most important free parameters of the method. The selection mechanism employs general characteristics of the cost functions that commonly occur in intensity-based image registration. Also, the theoretical convergence conditions of the optimisation method are taken into account. The proposed adaptive stochastic gradient descent (ASGD) method is compared to a standard, non-adaptive Robbins-Monro (RM) algorithm. Both ASGD and RM employ a stochastic subsampling technique to accelerate the optimisation process. Registration experiments were performed on 3D CT and MR data of the head, lungs, and prostate, using various similarity measures and transformation models. The results indicate that ASGD is robust to these variations in the registration framework and is less sensitive to the settings of the user-defined parameters than RM. The main disadvantage of RM is the need for a predetermined step size function. The ASGD method provides a solution for that issue.     

Evaluation of optimization methods for nonrigid medical image registration using mutual information and B-splines.

A popular technique for nonrigid registration of medical images is based on the maximization of their mutual information, in combination with a deformation field parameterized by cubic B-splines. The coordinate mapping that relates the two images is found using an iterative optimization procedure. This work compares the performance of eight optimization methods: gradient descent (with two different step size selection algorithms), quasi-Newton, nonlinear conjugate gradient, Kiefer-Wolfowitz, simultaneous perturbation, Robbins-Monro, and evolution strategy. Special attention is paid to computation time reduction by using fewer voxels to calculate the cost function and its derivatives. The optimization methods are tested on manually deformed CT images of the heart, on follow-up CT chest scans, and on MR scans of the prostate acquired using a BFFE, T1, and T2 protocol. Registration accuracy is assessed by computing the overlap of segmented edges. Precision and convergence properties are studied by comparing deformation fields. The results show that the Robbins-Monro method is the best choice in most applications. With this approach, the computation time per iteration can be lowered approximately 500 times without affecting the rate of convergence by using a small subset of the image, randomly selected in every iteration, to compute the derivative of the mutual information. From the other methods the quasi-Newton and the nonlinear conjugate gradient method achieve a slightly higher precision, at the price of larger computation times.     

Physico-Chemical Characterizations of Poly(vinylidene fluoride)/Cu3(BTC)2 Composite Membranes Prepared by In Situ Crystal Growth

In this work, we present a simple and fast method for elaborating hybrid membranes by growing metal–organic framework crystals inside a polymer solution. The solution thus obtained was casted then annealed at 90°C for 5 h. This method was tested with poly(vinylidene fluoride) (PVDF) as a piezoelectric polymer and the Cu₃(BTC)₂, BTC = 1,3,5-benzene tricarboxylate, as a filler. The characterization of the obtained membranes by attenuated total reflectance Fourier transform infrared spectroscopy and X-ray diffraction showed the presence of the characteristic signatures of Cu₃(BTC)₂ and the β-phase of PVDF. Moreover, scanning electron microscopy images reveal that the Cu₃(BTC)₂ crystallites have grown along the PVDF membranes. The effect of the filler on both thermal and mechanical properties of the membranes was also studied. POLYM. ENG. SCI., 60:464–473, 2020. © 2019 Society of Plastics Engineers     

Fourier‐based numerical approximation of the Weertman equation for moving dislocations

This work addresses the numerical approximation of solutions to a dimensionless form of the Weertman equation, which models a steadily moving dislocation and is an important extension (with advection term) of the celebrated Peierls‐Nabarro equation for a static dislocation. It belongs to the class of nonlinear reaction‐advection‐diffusion integro‐differential equations with Cauchy‐type kernel, thus involving an integration over an unbounded domain. In the Weertman problem, the unknowns are the shape of the core of the dislocation and the dislocation velocity. The proposed numerical method rests on a time‐dependent formulation that admits the Weertman equation as its long‐time limit. Key features are (1) time iterations are conducted by means of a new, robust, and inexpensive Preconditioned Collocation Scheme in the Fourier domain, which allows for explicit time evolution but amounts to implicit time integration, thus allowing for large time steps; (2) as the integration over the unbounded domain induces a solution with slowly decaying tails of important influence on the overall dislocation shape, the action of the operators at play is evaluated with exact asymptotic estimates of the tails, combined with discrete Fourier transform operations on a finite computational box of size L; (3) a specific device is developed to compute the moving solution in a comoving frame, to minimize the effects of the finite‐box approximation. Applications illustrate the efficiency of the approach for different types of nonlinearities, with systematic assessment of numerical errors. Converged numerical results are found insensitive to the time step, and scaling laws for the combined dependence of the numerical error with respect to L and to the spatial step size are obtained. The method proves fast and accurate and could be applied to a wide variety of equations with moving fronts as solutions; notably, Weertman‐type equations with the Cauchy‐type kernel replaced by a fractional Laplacian.     

Diagnostic et caractérisation des aléas

France alone has thousands of underground quarries. There are usually old and somewhat diversified workings used for building stones or binding materials, etc. In highly urbanized areas, such as the Paris region, the Nord Pas de Calais, Lower Normandy, Loire and Aquitaine, there is a twofold safety problem: population and rehabilitation of former sites. The dangers presented by the existence of old sites are numerous: instability, subsidence, collapse, caving in, etc. Every problem has its own cause: cracks in walls, abutment walls, upper or lower roofs; and ill-designed roof slopes can cause their own problems. Design and inspection methods have in fact been specified for different types of problems and their effects. The stability rating in question will be on the basis of the degree of instability involved, and will include all risks in order to include all potential major site problems. It has been felt necessary to work out three levels of diagnosis and to define minimal limits and requirements in respect of the measures to be applied. The methods and techniques used for studying stability are defined according to the severity of the problem, and include cavities, site surveys, engineering factors to do with stability, behavioural models and inspection reports. The importance of feedback regarding these surveys is emphasiized and embodied in the symptomatic “risk catchment” concept designed to see precisely what the situation is in respect of each quarry, i.e. Paris-West-Versailles chalk, gypsum from West of Paris, Lutetian limestone from the Oise, etc.).     

Integrated use of fuzzy c-means and fuzzy KNN for GT part family and machine cell formation

This paper presents a new approach for GT part family and machine cell formation. It involves the integrated use of two fuzzy clustering algorithms: fuzzy c-means and fuzzy k-nearest neighbours. It is shown that the proposed approach performs better than using fuzzy c-means alone or FACT (Kamal and Burke) in terms of some commonly used measures such as grouping efficacy, grouping index, number of voids, number of exceptional elements, and number of bottleneck machines. The approach is developed a result of our quest for a better clustering algorithm to process non-binary data and to produce a non-binary classification in the domain of GT applications. These features are deemed important to handle imprecise data and to provide a higher degree of flexibility in the operation stage.     

Photopolymerization of Pollen Based Biosourced Composites and Applications in 3D and 4D Printing

Pollen have the potential to be effective plant-based biorenewable reinforcing fillers for polymers due to their chemical stability and unique micro- or nano-structured surfaces. Pollen-polymer composites can form the basis for a new class of light-weight and sustainable materials if compatible polymer-filler systems can be engineered through photopolymerization, but this idea is previously unexplored. The first demonstration of photopolymerization and 3D printing with the incorporation of pine pollen as filler in poly(ethylene glycol) diacrylate are presented. The filler properties affecting the related depth of cure and the mechanical, thermal, and functional properties are examined in detail. In addition, the lithography technique is applied to the photocomposites for the production of 3D patterns. 4D printing behavior is also possible through the water swelling and dehydration induced actuation of the 3D printed composites with spatial resolution features. This work is expected to provide a new way to a field for photopolymerization reactions in natural material-resin composites and thereby to expand potential applications in 3D and 4D printing.     

Mood reactivity to daily negative events in early adolescence: Relationship to risk for psychopathology.

Emotional responses to negative daily experiences in young adolescents may provide important clues to the development of psychopathology, but research is lacking. This study assessed momentary mood reactivity to daily events as a function of risk profile in a school sample, ages 11-14. High-risk (HR, n = 25) and low-risk (LR, n = 106) subgroups completed frequent self-reports of mood and events for 5 days. HR adolescents reported more negative events involving family and peers. Multilevel modeling results showed that negative events, especially if stressful, were associated with increased negative and decreased positive affects, with heightened responses in HR adolescents. HR adolescents with greater stress over the last 3 months showed additional increases in depressed mood following negative events. Altered reactivity to and dysfunctional appraisals of daily events may link adolescent risk profiles to later mental health problems. (PsycINFO Database Record (c) 2010 APA, all rights reserved)