Search

Empirical studies in agricultural economics usually involve policy implications. In many cases, such studies rely on proprietary or confidential data that cannot be published along with the article, challenging the replicability and credibility of the results. To overcome this problem, the use of synthetic data—that is, data that do not contain a single unit of the original data—has been proposed. In this note, we illustrate the utility of synthetic data generation methods for replication purposes using a range of methods from agricultural production analysis. More specifically, we compare input elasticities and technical efficiency scores based on different farm-level production data between original data and synthetic data. We generate synthetic data using a non-parametric method of classification and regression trees (CART) and parametric linear regressions. We find synthetic data result in elasticities and technical efficiency distributions that are very similar to the original data, especially when generated with CART, and conclude with implications for the research community., Journal of Agricultural Economics, 74 (1), ISSN:0021-857X, ISSN:1477-9552

PURPOSE: To compare different optimization approaches for choosing the spin-lock times (TSLs), in spin-lattice relaxation time in the rotating frame (T(1ρ)) mapping. METHODS: Optimization criteria for TSLs based on Cramér-Rao lower bounds (CRLB) are compared with matched sampling-fitting (MSF) approaches for T(1ρ) mapping on synthetic data, model phantoms, and knee cartilage. The MSF approaches are optimized using robust methods for noisy cost functions. The MSF approaches assume that optimal TSLs depend on the chosen fitting method. An iterative non-linear least squares (NLS) and artificial neural networks (ANN) are tested as two possible T(1ρ) fitting methods for MSF approaches. RESULTS: All optimized criteria were better than non-optimized ones. However, we observe that a modified CRLB and an MSF based on the mean of the normalized absolute error (MNAE) were more robust optimization approaches, performing well in all tested cases. The optimized TSLs obtained the best performance with synthetic data (3.5–8.0% error), model phantoms (1.5–2.8% error), and healthy volunteers (7.7–21.1% error), showing stable and improved quality results, comparing to non-optimized approaches (4.2–13.3% error on synthetic data, 2.1–6.2% error on model phantoms, 9.8–27.8% error on healthy volunteers). CONCLUSION: A modified CRLB and the MSF based on MNAE are robust optimization approaches for choosing TSLs in T(1ρ) mapping. All optimized criteria allowed good results even using rapid scans with 2 TSLs when a complex-valued fitting is done with iterative NLS or ANN.

Developing high‐performing machine learning algorithms requires large amounts of annotated data. Manual annotation of data is labour‐intensive, and the cost and effort needed are an important obstacle to the development and deployment of automated analysis. In a previous work, we have shown that deep learning classifiers can successfully be trained on synthetic images and annotations. Here, we provide a curated set of fish image data and backgrounds, the necessary software tools to generate synthetic images and annotations, and annotated real datasets to test classifier performance. The dataset is constructed from images collected using the Deep Vision system during two surveys from 2017 and 2018 that targeted economically important pelagic species in the Northeast Atlantic Ocean. We annotated a total of 1,879 images, randomly selected across trawl stations from both surveys, comprising 482 images of blue whiting, 456 images of Atlantic herring, 341 images of Atlantic mackerel, 335 images of mesopelagic fishes and 265 images containing a mixture of the four categories.

PURPOSE We propose a deep learning method that classifies focal liver lesions (FLLs) into cysts, hemangiomas, and metastases from portal phase abdominal CT images. We propose a synthetic data augmentation process to alleviate the class imbalance and the Lesion INformation Augmented (LINA) patch to improve the learning efficiency. METHODS A dataset of 502 portal phase CT scans of 1,290 FLLs was used. First, to alleviate the class imbalance and to diversify the training data patterns, we suggest synthetic training data augmentation using DCGAN-based lesion mask synthesis and pix2pix-based mask-to-image translation. Second, to improve the learning efficiency of convolutional neural networks (CNNs) for the small lesions, we propose a novel type of input patch termed the LINA patch to emphasize the lesion texture information while also maintaining the lesion boundary information in the patches. Third, we construct a multi-scale CNN through a model ensemble of ResNet-18 CNNs trained on LINA patches of various mini-patch sizes. RESULTS The experiments demonstrate that (a) synthetic data augmentation method shows characteristics different but complementary to those in conventional real data augmentation in augmenting data distributions, (b) the proposed LINA patches improve classification performance compared to those by existing types of CNN input patches due to the enhanced texture and boundary information in the small lesions, and (c) through an ensemble of LINA patch-trained CNNs with different mini-patch sizes, the multi-scale CNN further improves overall classification performance. As a result, the proposed method achieved an accuracy of 87.30%, showing improvements of 10.81%p and 15.0%p compared to the conventional image patch-trained CNN and texture feature-trained SVM, respectively. CONCLUSIONS The proposed synthetic data augmentation method shows promising results in improving the data diversity and class imbalance, and the proposed LINA patches enhance the learning efficiency compared to the existing input image patches.

Balancing allocation of assigning units to two treatment groups to minimize the allocation differences is important in biomedical research. The complete randomization, rerandomization, and pairwise sequential randomization (PSR) procedures can be employed to balance the allocation. However, the first two do not allow a large number of covariates. In this article, we generalize the PSR procedure and propose a k-resolution sequential randomization (k-RSR) procedure by minimizing the Mahalanobis distance between both groups with equal group size. The proposed method can be used to achieve adequate balance and obtain a reasonable estimate of treatment effect. Compared to PSR, k-RSR is more likely to achieve the optimal value theoretically. Extensive simulation studies are conducted to show the superiorities of k-RSR and applications to the clinical synthetic data and GAW16 data further illustrate the methods.

Purpose: To evaluate an algorithm for calibrationless parallel imaging to reconstruct undersampled parallel transmit field maps for the body and brain. Methods: Using synthetic data, body, and brain measurements of relative transmit maps, three different approaches to a joint transmit-receive low-rank tensor completion algorithm are evaluated. These methods included: (i) virtual coils using the product of receive and transmit sensitivities, (ii) joint-receiver coils that enforces a low rank structure across receive coils of all transmit modes, and (iii) transmit low rank (TxLR) that uses a low rank structure for both receive and transmit modes simultaneously. The performance of each are investigated for different noise levels and different acceleration rates on an 8-channel parallel transmit 7T system. Results: The virtual coils method broke down with increasing noise levels or acceleration rates greater than two producing RMS error greater than 0.1. The joint receiver coils method worked well up to acceleration factors of four, beyond which the RMS error exceeded 0.1. While TxLR enabled an eight-fold acceleration with most RMS errors remaining below 0.1. Conclusion: This work demonstrates that under-sampling factors of up to eight-fold are feasible for transmit array mapping and can be reconstructed using calibrationless parallel imaging methods., 30 pages, 14 figures

Different geophysical methods use different model parameterizations and inversion algorithms. Thus, combining these different inversion systems and yet adding the nonlinear cross‐gradient constraint in a joint inversion framework might be a big challenge, for instance, as explained further by Moorkamp et al. in 2011, there is a complex interaction between the data misfit terms, regularization and cross‐gradient terms and an imperfect fit to the data is expected. In this paper, we use a sequential algorithm for a two‐dimensional joint inversion of gravity and magnetic data, which tries to avoid these issues by decoupling the gravity inversion, the magnetic inversion and the cross‐gradient minimization processes. The efficiency of the algorithm and developed code is demonstrated by the joint inversion of noisy synthetic data. The results show a significant improvement in the respective models obtained by introducing the cross‐gradient joint inversion over the models obtained by separate inversions for synthetic data and then for field data targeting potash ore source in the AjiChai salt deposit in north‐western Iran. In this application case, the lower density of salt minerals such as potash, compared to its surrounding sedimentary sequences, motivates a gravity study. Furthermore, the relative lower susceptibility of these salt minerals, alongside their diamagnetic effect, makes them a suitable target for magnetic surveys. Separate gravity and magnetic studies had been performed over the deposit; however, a constitutive relationship between density and magnetization within the area of interest supporting a joint inversion had not been established. In this paper, we apply the sequential cross‐gradient approach to perform the first full joint inversion for the AjiChai salt deposit. The magnetic inversion here is performed to recover the magnetization amplitudes rather than the magnetization vector. In fact, we assume there is no remanent magnetization and, therefore, that the magnetization vector is constant and parallel to the geomagnetic field direction. The constructed density and magnetization models are of high concordance with available geological information and previous studies including drilling results. In addition, unlike previous separate inversion models, the models are structurally and geometrically similar.

This study introduces a new attribute to identify seismic erratic noise, i.e. outlier, in the context of unsupervised anomaly detection and is defined as local outlier probabilities. The local outlier probabilities calculate scores of degrees of isolation, i.e. outlier‐ness, for each object in a data set, which represents how far an object is deviated from its surrounding objects. Since the local outlier probabilities combines a density‐based outlier detection method with a statistically oriented scheme, its scoring system provides regularized outlier‐ness, which is an outlier probability, to be used for making a binary decision to do inclusion or exclusion of an object; such a decision only requires a simple and straightforward threshold on a probability. Based on the binary decision that flags outliers versus non‐outliers, local outlier probabilities‐denoising workflows are developed by combining multiple steps to complete an application of the local outlier probabilities to attenuate seismic erratic noise. Higher stability and improved robustness in the detection and rejection of seismic erratic noise have been achieved by implementing moving windows and decision tree‐based processes. To avoid loss of useful signal energy, signal enhancement applications are additionally suggested. Numerical experiments on synthetic data investigate the applicability of the proposed algorithms to seismic erratic noise attenuation. Field data examples demonstrate the feasibility of a local outlier probabilities‐denoising application as an effective tool in seismic denoising portfolio.

Obtaining accurate velocity models plays a crucial role in many routine seismic imaging algorithms. Seismic velocity models are normally made through seismic velocity analysis workflows. The routine workflows are not capable of dealing with polarity variations across moveout curves. We address this limitation by proposing a straight‐forward and robust semblance‐based workflow, which is a modified version of the conventional semblance function. The coherency function applies semblance analysis on separate clusters of receivers followed by averaging the corresponding coherency measures from all the clusters. The proposed approach is suitable for any case of amplitude variations including attenuation and any class of amplitude‐versus‐offset effects. The ability of the proposed workflow is demonstrated to two synthetic data as well as two field‐recorded common‐midpoint gathers. We perform accuracy analysis by comparing the results from the proposed approach with the results achieved from conventional velocity analysis, and another semblance‐based algorithm that is developed to address the polarity variation task. We also studied noise sensitivity analysis by computing and comparing mathematical expectations between theory and practice.

Accurate and precise estimation of microseismic event location in anisotropic reservoirs is challenging with borehole acquisition, due to limited azimuthal and inclination coverage. Inclusion of shear‐wave splitting information from microseismic data can improve estimates of event directivity, source location and the degree of reservoir anisotropy compared with using P‐waves alone. Additionally, S‐waves typically have a higher signal‐to‐noise ratio than P‐waves, due to higher amplitudes, decreasing the overall azimuthal and depth uncertainty in comparison with P‐wave only location methodologies. We present a joint P‐ and S‐wave hodogram workflow to reduce azimuthal uncertainty and estimate shear‐wave splitting. In this paper, finite‐difference waveform synthetics verify the workflow and illustrate interference effects associated with reflections and head‐waves. Multiple shear‐wave splitting techniques are applied to the synthetic data set in order to understand the limitations between splitting methodologies. Combining the P‐wave hodogram calculation of azimuth and inclination with the shear‐wave splitting analysis lowers the azimuthal uncertainty and improves waveform rotation results. From the comparison of three splitting methods, the minimum second eigenvalue is the most optimum for the data set. Post‐splitting analysis was key as the results are affected by the complexities in the waveforms and arrivals. Using receiver‐by‐receiver splitting quality analysis with shot and receiver clustering improves the identification of quality splitting results.

Population genetics relies heavily on simulated data for validation, inference, and intuition. In particular, since the evolutionary “ground truth” for real data is always limited, simulated data is crucial for training supervised machine learning methods. Simulation software can accurately model evolutionary processes, but requires many hand-selected input parameters. As a result, simulated data often fails to mirror the properties of real genetic data, which limits the scope of methods that rely on it. Here, we develop a novel approach to estimating parameters in population genetic models that automatically adapts to data from any population. Our method,pg-gan, is based on a generative adversarial network that gradually learns to generate realistic synthetic data. We demonstrate that our method is able to recover input parameters in a simulated isolation-with-migration model. We then apply our method to human data from the 1000 Genomes Project, and show that we can accurately recapitulate the features of real data.

The images produced by conventional vertical seismic profile imaging methods often suffer from the problems of illumination limitation and uncompleted wavefield separation. In this paper, we give a deep insight into the illumination behaviours of the upgoing primary and downgoing multiple signals and illustrate their merits and demerits in illuminating the subsurface structure around the well trajectory. Based on this, we develop an efficient joint imaging method of acoustic vertical seismic profile data based on the generalized Radon transform migration inversion theory. By analysing the interference characteristics of the primary and multiple reflections, we verify the capability of this method that can handle the joint imaging directly. The proposed method combines the advantages of primary and multiple contributions to the subsurface image without additional wavefield separation. The numerical results of synthetic data and field data demonstrate high‐quality construction of the structure in the vicinity of the borehole, also away from the well trajectory.

In this paper, we propose a method for revealing hidden reservoirs that are shielded by strong amplitudes. The bi‐dimensional empirical mode decomposition algorithm is used to decompose pre‐stack seismic data into several localized components, which generated from different discontinuities in the subsurface elastic properties. The two‐dimensional Teager–Kaiser energy operator process is applied to the first component, which includes a strong signal, to further locate the strong event. According to the located results, an energy weight matrix is established. By weighted summation of all the components, the strong event is suppressed, and the hidden reservoir becomes more prominent. Tests on a synthetic data and field data from Daniudi confirm that this method can separate strong signals from weaker responses and efficiently reveal shielded reservoirs.

PURPOSE A pharmacokinetic analysis of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data is subject to inaccuracy and instability partly owing to the partial volume effect (PVE). We proposed a new multicompartment model for a tissue-specific pharmacokinetic analysis in DCE-MRI data to solve the PVE problem and to provide better kinetic parameter maps. METHODS We introduced an independent parameter named fractional volumes of tissue compartments in each DCE-MRI pixel to construct a new linear separable multicompartment model, which simultaneously estimates the pixel-wise time-concentration curves and fractional volumes without the need of the pure-pixel assumption. This simplified convex optimization model was solved using a special type of non-negative matrix factorization (NMF) algorithm called the minimum-volume constraint NMF (MVC-NMF). RESULTS To test the model, synthetic datasets were established based on the general pharmacokinetic parameters. On well-designed synthetic data, the proposed model reached lower bias and lower root mean square fitting error compared to the state-of-the-art algorithm in different noise levels. In addition, the real dataset from QIN-BREAST-DCE-MRI was analyzed, and we observed an improved pharmacokinetic parameter estimation to distinguish the treatment response to chemotherapy applied to breast cancer. CONCLUSION Our model improved the accuracy and stability of the tissue-specific estimation of the fractional volumes and kinetic parameters in DCE-MRI data, and improved the robustness to noise, providing more accurate kinetics for more precise prognosis and therapeutic response evaluation using DCE-MRI.

Nowadays the least‐squares reverse time migration has become the most used migration method because of its accuracy in amplitude recovery and high‐resolution imaging, specifically its priority to image the beneath of structural domes such as salt domes. However, errors in the migration velocity model, inadequate physics of modelling/migration and too sparse data decrease the quality of the migrated image. Sparsity constraints help to mitigate the shortcomings of the least‐squares reverse time migration and stabilize the migration image, but the computational burden of sparse solvers is still a big challenge. In this paper, we propose a fast sparsity‐promoting least‐squares reverse time migration algorithm based on the Bregmanized operator splitting algorithm. In particular, we solve the least‐squares reverse time migration with l1‐norm regularization to increase the image resolution while removing the artefacts which cannot be suppressed by the traditional least‐squares reverse time migration. Also we develop a preconditioned Bregmanized operator splitting algorithm where iteratively using of a preconditioner decreases the computational burden. The proposed method is applied to a few sets of synthetic data, and results are compared with reverse time migration or least‐squares reverse time migration to verify its superiority. Numerical tests demonstrate that the proposed preconditioned Bregmanized operator splitting algorithm converges to the desired migration image in a small number of iterations.

The Farquhar-von Caemmerer-Berry (FvCB) model is extensively used to model photosynthesis from gas exchange measurements. Since its publication, many methods have been developed to measure, or more accurately estimate, parameters of this model. Here, we have created a tool that uses Bayesian statistics to fit photosynthetic parameters using concurrent gas exchange and chlorophyll fluorescence measurements whilst evaluating the reliability of the parameter estimation. We have tested this tool on synthetic data and experimental data from rice leaves. Our results indicate that reliable parameter estimation can be achieved whilst only keeping one parameter, Km , that is, Michaelis constant for CO2 by Rubisco, prefixed. Additionally, we show that including detailed low CO2 measurements at low light levels increases reliability and suggests this as a new standard measurement protocol. By providing an estimated distribution of parameter values, the tool can be used to evaluate the quality of data from gas exchange and chlorophyll fluorescence measurement protocols. Compared to earlier model fitting methods, the use of a Bayesian statistics-based tool minimizes human interaction during fitting, reducing the subjectivity which is essential to most existing tools. A user friendly, interactive Bayesian tool script is provided.

SummaryIn contrast to differential gene expression analysis at single gene level, gene regulatory networks (GRN) analysis depicts complex transcriptomic interactions among genes for better understandings of underlying genetic architectures of human diseases and traits. Recently, single-cell RNA sequencing (scRNA-seq) data has started to be used for constructing GRNs at a much finer resolution than bulk RNA-seq data and microarray data. However, scRNA-seq data are inherently sparse which hinders direct application of the popular Gaussian graphical models (GGMs). Furthermore, most existing approaches for constructing GRNs with scRNA-seq data only consider gene networks under one condition. To better understand GRNs under different but related conditions with single-cell resolution, we propose to construct Joint Gene Networks with scRNA-seq data (JGNsc) using the GGMs framework. To facilitate the use of GGMs, JGNsc first proposes a hybrid imputation procedure that combines a Bayesian zero-inflated Poisson (ZIP) model with an iterative low-rank matrix completion step to efficiently impute zero-inflated counts resulted from technical artifacts. JGNsc then transforms the imputed data via a nonparanormal transformation, based on which joint GGMs are constructed. We demonstrate JGNsc and assess its performance using synthetic data. The application of JGNsc on two cancer clinical studies of medulloblastoma and glioblastoma identifies novel findings in addition to confirming well-known biological results.

Seismic attenuation compensation is a spectrum‐broadening technique for enhancing the resolution of non‐stationary seismic data. The single‐trace attenuation compensation algorithms ignore the prior information that the seismic reflection events are generally continuous along seismic traces, thus, the compensated result may have poor spatial continuity and low signal‐to‐noise ratio. To address this problem, we extend the single‐trace approaches to the multi‐trace algorithms and furthermore propose a multi‐trace attenuation compensation with a spatial constraint. The frequency‐space prediction filters are the key to construct this spatial regularization. We test the effectiveness of the proposed spatially constrained attenuation compensation algorithm by applying both synthetic and field data examples. Synthetic data tests indicate that the proposed multi‐trace attenuation compensation approach can provide a better compensated result than single‐trace attenuation compensation algorithm in terms of suppressing noise amplification and guaranteeing structural continuities. Field data applications further confirm its stability and practicality to improve seismic resolution.

Seismically derived amplitude‐versus‐angle attributes along with well constraints are the base inputs into inverting seismic into subsurface properties. Conditioning the common image gathers is a common workflow in quantitative inversion and leads to a more accurate inversion product due to the removal of post‐migration artefacts. Here, we apply a neural network to condition the post‐migration gathers. The network is a cycle generative adversarial network, CycleGAN, which was designed for image‐to‐image translation. This can be considered the same problem as translating an artefact rich seismic gather to an artefact free seismic gather. To assess the feasibility of applying the network to pre‐stack conditioning, synthetic data sets were generated to train different networks for different tasks. The networks were trained to remove white noise, residual de‐multiples, gather flattening and a combination of the above for conditioning. The results show that a trained network was able remove white noise providing a more robust amplitude‐versus‐offset calculation. Another network trained using synthetic gathers with and without multiples assisted in multiple removal. However, instability around primary preservation has been observed so the network works better as a residual de‐multiple method. For gather conditioning, a network was trained with the unpaired artefact‐rich and artefact‐free training data where the artefacts included complex moveout, noise and multiples. When applied to the test data sets, the networks cleaned the artefact‐rich test data and translated complex moveout into flat gathers whilst preserving the amplitude response. Finally, two networks are applied to real data where a gather based on the well logs is used to quantify the match between the conditioned gathers and the raw gathers. The first network used synthetic data to train the network and, when applied to real data, provided a better tie with the well. The second network was trained with synthetic gathers whose properties were constrained by real seismic gathers from near the well. As anticipated, the network trained on the representative training data outperforms the network trained using the unconstrained data. However, the ability of the first network to condition the gather indicates that a sweep of networks can be trained without the need for real data and applied in a manner analogous to the way parameters are adjusted in traditional geophysical methods. The results show that the different neural networks can offer an alternative or augmentation to the existing geophysical workflow for conditioning pre‐stack seismic gathers.

Fil: Macchioli Grande, Franco. CONICET, Facultad de Ciencias Astronómicas y Geofísicas, Universidad Nacional de La Plata; Argentina. Fil: Zyserman, Fabio. CONICET, Facultad de Ciencias Astronómicas y Geofísicas, Universidad Nacional de La Plata; Argentina. Fil: Monachesi, Leonardo. Universidad Nacional de Río Negro. Instituto de Investigación en Paleobiología y Geología. Río Negro, Argentina. Fil: Jouniaux, Laurence. Institut de Physique du Globe de Strasbourg (UMR 7516), Université de Strasbourg et CNRS, Strasbourg; Francia. Fil: Rosas Carbajal, Marina. Institut de Physique du Globe de Paris; Francia. In glacial studies, properties such as glacier thickness and the basement permeability and porosity are key to understand the hydrological and mechanical behaviour of the system. The seismoelectric method could potentially be used to determine key properties of glacial environments. Here we analytically model the generation of seismic and seismoelectric signals by means of a shear horizontal seismic wave source on top of a glacier overlying a porous basement. Considering a one-dimensional setting, we compute the seismic waves and the electrokinetically induced electric field. We then analyse the sensitivity of the seismic and electromagnetic data to relevant model parameters, namely depth of the glacier bottom, porosity, permeability, shear modulus and saturating water salinity of the glacier basement. Moreover, we study the possibility of inferring these key parameters from a set of very low noise synthetic data, adopting a Bayesian framework to pay particular attention to the uncertainty of the model parameters mentioned above. We tackle the resolution of the probabilistic inverse problem with two strategies: (1) we compute the marginal posterior distributions of each model parameter solving multidimensional integrals numerically and (2) we use a Markov chain Monte Carlo algorithm to retrieve a collection of model parameters that follows the posterior probability density function of the model parameters, given the synthetic data set. Both methodologies are able to obtain the marginal distributions of the parameters and estimate their mean and standard deviation. The Markov chain Monte Carlo algorithm performs better in terms of numerical stability and number of iterations needed to characterize the distributions. The inversion of seismic data alone is not able to constrain the values of porosity and permeability further than the prior distribution. In turn, the inversion of the electric data alone, and the joint inversion of seismic and electric data are useful to constrain these parameters as well as other glacial system properties. Furthermore, the joint inversion reduces the uncertainty of the model parameters estimates and provides more accurate results.

Random noise attenuation utilizing predictive filtering achieves great performance in denoising seismic data. Conventional predictive filtering methods are based on fixed filter operators and neglect the complexity of structures. In this way, the denoised data cannot meet the requirement of balancing the signal preservation and noise removal. In this study, we proposed a structural complexity‐guided predictive filtering method that utilizes an adapted filter operator to adjust the changes of structural complexity. The proposed structural complexity‐guided predictive filtering mainly consists of two stages. A slope field information is acquired according to plane‐wave destruction to assess the structural complexity. In addition, an adaptive filter operator is obtained to denoise the seismic data according to the adaptive factor. Both synthetic data and real seismic profiles are employed to examine the denoising capacity and flexibility of the refined predictive filtering using adaptive lengths. The analysis of the predicted results shows that adaptive predictive filtering is powerful and has the ability to eliminate random noises with negligible distortions.

Neurite orientation dispersion and density imaging (NODDI) enables the assessment of intracellular, extracellular, and free water signals from multi-shell diffusion MRI data. It is an insightful approach to characterize brain tissue microstructure. Single-shell reconstruction for NODDI parameters has been discouraged in previous studies caused by failure when fitting, especially for the neurite density index (NDI). Here, we investigated the possibility of creating robust NODDI parameter maps with single-shell data, using the isotropic volume fraction (fISO ) as a prior. Prior estimation was made independent of the NODDI model constraint using a dictionary learning approach. First, we used a stochastic sparse dictionary-based network (DictNet), which is trained with data obtained from in vivo and simulated diffusion MRI data, to predict fISO . In single-shell cases, the mean diffusivity and raw T2 signal with no diffusion weighting (S0 ) was incorporated in the dictionary for the fISO estimation. Then, the NODDI framework was used with the known fISO to estimate the NDI and orientation dispersion index (ODI). The fISO estimated using our model was compared with other fISO estimators in the simulation. Further, using both synthetic data simulation and human data collected on a 3 T scanner (both high-quality HCP and clinical dataset), we compared the performance of our dictionary-based learning prior NODDI (DLpN) with the original NODDI for both single-shell and multi-shell data. Our results suggest that DLpN-derived NDI and ODI parameters for single-shell protocols are comparable with original multi-shell NODDI, and the protocol with b = 2000 s/mm2 performs the best (error ~ 5% in white and gray matter). This may allow NODDI evaluation of studies on single-shell data by multi-shell scanning of two subjects for DictNet fISO training.

Precise model personalization is a key step towards the application of cardiovascular physical models. In this manuscript, we propose to use deep learning (DL) to solve the parameter estimation problem in cardiovascular hemodynamics. Based on the convolutional neural network (CNN) and fully connected neural network (FCNN), a multi-input deep neural network (DNN) model is developed to map the nonlinear relationship between measurements and the parameters to be estimated. In this model, two separate network structures are designed to extract the features of two types of measurement data, including pressure waveforms and a vector composed of heart rate (HR) and pulse transit time (PTT), and a shared structure is used to extract their combined dependencies on the parameters. Besides, we try to use the transfer learning (TL) technology to further strengthen the personalized characteristics of a trained-well network. For assessing the proposed method, we conducted the parameter estimation using synthetic data and in vitro data respectively, and in the test with synthetic data, we evaluated the performance of the TL algorithm through two individuals with different characteristics. A series of estimation results show that the estimated parameters are in good agreement with the true values. Furthermore, it is also found that the estimation accuracy can be significantly improved by a multicycle combination strategy. Therefore, we think that the proposed method has the potential to be used for parameter estimation in cardiovascular hemodynamics, which can provide an immediate, accurate, and sustainable personalization process, and deserves more attention in the future.

The properties of the auroral electrojets are examined on the basis of a trained machine learning model. The relationships between solar-wind parameters and the AU and AL indices are modeled with an echo state network (ESN), a kind of recurrent neural network. We can consider this trained ESN model to represent nonlinear effects of the solar-wind inputs on the auroral electrojets. To identify the properties of auroral electrojets, we obtain various synthetic AU and AL data by using various artificial inputs with the trained ESN. The analyses of various synthetic data show that the AU and AL indices are mainly controlled by the solar-wind speed in addition to Bz of the interplanetary magnetic field (IMF) as suggested by the literature. The results also indicate that the solar-wind density effect is emphasized when solar-wind speed is high and when IMF Bz is near zero. This suggests some nonlinear effects of the solar-wind density.

We advance a principle directed to assigning numerical values to free parameters usually present in inversion methods. It may be formulated as: ‘Optimum estimates of free parameters in an inversion procedure must lead, in tests using synthetic data, to solutions whose geometrical expression reflects the main qualitative or semiquantitative geological characteristic of the study area.’ To this end, the interpreter should (i) specify a typical anomalous source geometry which incorporates the most relevant geological information for the study area, (ii) compute the corresponding gravity anomaly and (iii) invert the anomaly for the source geometry finding the numerical values of the free parameters that lead to a solution closest to the typical source. Application of the above methodology to synthetic and real data from the basement relief of a rift basin has asserted its efficacy.

The increased popularity of airborne measurements of the gravity gradient tensor for resource studies and geological mapping has resulted in a new awareness of the importance of terrain effects. In these measurements, the terrain effect often overwhelms that of the underlying crust and it becomes important to formulate a strategy for taking it into account when presenting the data and when inverting the data into density models. Using newly acquired data from Northern Sweden, we first attempted to estimate a variable terrain density model by inverting the data using a terrain model with a laterally varying density. Using data weights related to the topography variations, we find the best estimate of the lateral variation of the terrain density. We translate this model into a full three‐dimensional model such that all columns have the same vertical centre of mass as estimated from inspecting the radially averaged power spectrum of the area. This then defines a reference model for subsequent three‐dimensional inversion of the gravity gradient tensor dataset. We tested this approach first on synthetic data calculated from the measured topography including two density anomalies before we applied it to the measured data. The result is a model in which the surface density variations are propagated downwards in a systematic manner now in better agreement with measured densities of rock samples in the area.

In recent years, joint inversion has been widely used for integrated geological interpretation. We extended a data-space joint inversion algorithm of magnetotelluric, gravity and magnetic data to include first-arrival seismic travel-time and normalized cross-gradient constraints. We describe the main features of the algorithm and apply it to synthetic data generated for hypothetical models. For the synthetic data, we find that the joint inversion with multiple parameters is superior to the joint inversion with two or even three parameters, which can reduce the multisolution of inversion results more effectively. Furthermore, data-space joint inversion involves fewer memory requirements and better calculation speeds than traditional model-space joint inversion. The normalized cross-gradient constraints can better couple model parameters of different magnitudes compared with traditional unnormalized cross-gradient constraints, resulting in higher levels of structural similarity among resistivity, density, magnetic susceptibility and velocity models.

Cadzow filtering is currently considered as one of the most effective approaches for seismic data reconstruction. The basic version of Cadzow filtering first reorders each frequency slice of the seismic data (to be reconstructed) to a block Hankel/Toeplitz matrix, and then implements a rank‐reduction operator, that is truncated singular value decomposition, to the Hankel/Toeplitz matrix. However, basic Cadzow filtering cannot deal with the problem of recovering regularly missing data (up‐sampling) in the case of strongly aliased energy, because the regularly missing data will mix with signals in the singular spectrum. To solve this problem, it has been proposed to precondition the reconstruction of high‐frequency components using information from the low‐frequency components which are less aliased. In this paper, we further extend the de‐aliased Cadzow filtering approach to reconstruct regularly sampled seismic traces from the noisy observed data by modifying the reinserting operation, in which the high‐frequency components are projected onto the sub‐space spanned by several singular vectors of the low‐frequency components. At each iteration, the filtered data are weighted to the original data as a feedback. The weighting factor is related to the background noise level and changes with iteration. The feasibility of the proposed technique is validated via two‐dimensional, three‐dimensional and five‐dimensional synthetic data examples, as well as two‐dimensional post‐stack and three‐dimensional pre‐stack field data examples. The results demonstrate that the proposed technique can effectively interpolate regularly sampled data and is robust in noisy environments.

Purpose A new unsupervised learning method was developed to correct metal artifacts in MRI using 2 distorted images obtained with dual-polarity readout gradients. Methods An unsupervised learning method is proposed for a deep neural network architecture consisting of a deep neural network and an MR image generation module. The architecture is trained as an end-to-end process without the use of distortion-free images or off-resonance frequency maps. The deep neural network estimates frequency-shift maps between 2 distorted images that are obtained using dual-polarity readout gradients. From the estimated frequency-shift maps and 2 distorted input images, distortion-corrected images are obtained with the MR image generation module. Experiments using synthetic data and actual MR data were performed to compare images corrected by several metal-artifact-correction methods. Results The proposed method resolved the ripple and pile-up artifacts in the reconstructed images from synthetic data and actual MR data. The results from the proposed method were comparable to those from supervised-learning methods and superior to the compared model-based method. The proposed unsupervised learning method enabled the network to be trained without labels and to be more robust than supervised learning methods, for which overfitting problems can arise when using small training data sets. Conclusion Metal artifacts in the MR image were drastically corrected by the proposed unsupervised learning method. Two distorted images obtained with dual-polarity readout gradients are used as the input of the deep neural network. The proposed method can train networks without labels and does not overfit the network, even with small training data sets.

We develop a new methodology for determining the location and dynamics of brain activity from combined magnetoencephalography (MEG) and electroencephalography (EEG) data. The resulting inverse problem is ill‐posed and is one of the most difficult problems in neuroimaging data analysis. In our development we propose a solution that combines the data from three different modalities, magnetic resonance imaging (MRI), MEG and EEG, together. We propose a new Bayesian spatial finite mixture model that builds on the mesostate‐space model developed by Daunizeau & Friston [Daunizeau and Friston, NeuroImage 2007; 38, 67–81]. Our new model incorporates two major extensions: (i) We combine EEG and MEG data together and formulate a joint model for dealing with the two modalities simultaneously; (ii) we incorporate the Potts model to represent the spatial dependence in an allocation process that partitions the cortical surface into a small number of latent states termed mesostates. The cortical surface is obtained from MRI. We formulate the new spatiotemporal model and derive an efficient procedure for simultaneous point estimation and model selection based on the iterated conditional modes algorithm combined with local polynomial smoothing. The proposed method results in a novel estimator for the number of mixture components and is able to select active brain regions, which correspond to active variables in a high‐dimensional dynamic linear model. The methodology is investigated using synthetic data and simulation studies and then demonstrated on an application examining the neural response to the perception of scrambled faces. R software implementing the methodology along with several sample datasets are available at the following GitHub repository https://github.com/v2south/PottsMix. The Canadian Journal of Statistics 47: 688–711; 2019 © 2019 Statistical Society of Canada

Convolutional neural networks can provide a potential framework to characterize groundwater storage from seismic data. Estimation of key components such as the amount of groundwater stored in an aquifer and delineate water-table level, from active-source seismic data are investigated in this study. The data to train, validate, and test the neural networks are obtained by solving wave propagation in a coupled poroviscoelastic-elastic media. A discontinuous Galerkin method is applied to model wave propagation whereas a deep convolutional neural network is used for the parameter estimation problem. In the numerical experiment, the primary unknowns are the amount of stored groundwater and water-table level, and are estimated, while the remaining parameters, assumed to be of less of interest, are marginalized in the convolutional neural networks-based solution. Results, obtained through synthetic data, illustrate the potential of deep learning methods to extract additional aquifer information from seismic data, which otherwise would be impossible based on a set of reflection seismic sections or velocity tomograms.

In close-range applications, it is unclear whether optical photogrammetry is capable of accurately reconstructing submillimetre-scale roughness. This paper presents a study of fine-scale rock fracture roughness measurements with a careful assessment of the error. The workflow combines the techniques of structure from motion with focus stacking, using consumer-grade equipment and free or affordable software. The approach is tested firstly with synthetic data to check the influence of the number and position of cameras, object texture and image processing using focus stacking. Secondly, the optimised workflow is used to measure a natural shale rock fracture surface. To estimate the accuracy, the results were compared with a high-precision reference dataset provided by white-light interferometry. The standard deviation of error in the method is 6·5 μm, and is related with morphological structures with wavelengths below 150 μm and amplitudes smaller than 10 μm.

Purpose: To evaluate the accuracy and reproducibility of myocardial blood flow measurements obtained under different breathing strategies and motion correction techniques with arterial spin labeling. Methods: A prospective cardiac arterial spin labeling study was performed in 12 volunteers at 3 Tesla. Perfusion images were acquired twice under breath-hold, synchronized-breathing, and free-breathing. Motion detection based on the temporal intensity variation of a myocardial voxel, as well as image registration based on pairwise and groupwise approaches, were applied and evaluated in synthetic and in vivo data. A region of interest was drawn over the mean perfusion-weighted image for quantification. Original breath-hold datasets, analyzed with individual regions of interest for each perfusion-weighted image, were considered as reference values. Results: Perfusion measurements in the reference breath-hold datasets were in line with those reported in literature. In original datasets, prior to motion correction, myocardial blood flow quantification was significantly overestimated due to contamination of the myocardial perfusion with the high intensity signal of blood pool. These effects were minimized with motion detection or registration. Synthetic data showed that accuracy of the perfusion measurements was higher with the use of registration, in particular after the pairwise approach, which probed to be more robust to motion. Conclusion: Satisfactory results were obtained for the free-breathing strategy after pairwise registration, with higher accuracy and robustness (in synthetic datasets) and higher intrasession reproducibility together with lower myocardial blood flow variability across subjects (in in vivo datasets). Breath-hold and synchronized-breathing after motion correction provided similar results, but these breathing strategies can be difficult to perform by patients. Verónica Aramendía-Vidaurreta received PhD grant support from Asociación de Amigos de la Universidad de Navarra and Banco Santander

Magnetic resonance imaging of hyperpolarized pyruvate provides a new imaging biomarker for cancer metabolism, based on the dynamic in vivo conversion of hyperpolarized pyruvate to lactate. Methods for quantification of signal evolution need to be robust and reproducible across a range of experimental conditions. Pharmacokinetic analysis of dynamic spectroscopic imaging data from hyperpolarized pyruvate and its metabolites generally assumes that signal arises from ideal rectangular slice excitation profiles. In this study, we examined whether this assumption could lead to bias in kinetic analysis of hyperpolarized pyruvate and if so, whether such a bias can be corrected. A Bloch-McConnell simulator was used to generate synthetic data using a known set of “ground truth” pharmacokinetic parameter values. Signal evolution was then analyzed using analysis software that either assumed a uniform slice profile, or incorporated information about the slice profile into the analysis. To correct for slice profile effects, the expected slice profile was subdivided into multiple sub-slices in order to account for variable excitation angles along the slice dimension. An ensemble of sub-slices was then used to fit the measured signal evolution. A mismatch between slice profiles used for data acquisition and those assumed during kinetic analysis was identified as a source of quantification bias. Results indicate that imperfect slice profiles preferentially increase detected lactate signal, leading to an overestimation of the apparent metabolic exchange rate. The slice profile correction algorithm was tested in simulation, in phantom measurements, and applied to data acquired from a patient with prostate cancer. The results demonstrated that slice profile induced biases can be minimized by accounting for the slice profile during pharmacokinetic analysis. This algorithm can be used to correct data from either single or multi-slice acquisitions.

International audience; This paper introduces a method for dynamic 3D registration of female pelvic organs using 2D dynamic magnetic-resonance images (MRI). The aim is to provide a better knowledge and understanding of pathologies such as prolapsus or abnormal mobility of tissues. 2D dynamic MRI sequences are commonly used in nowadays clinical routines in order to evaluate the dynamic of organs, but due to the limited view, subjectivity related to human perception cannot be avoided in the diagnoses. A novel method for 2D/3D registration is proposed combining 3D Finite Element models with a priori knowledge of boundary conditions, in order to provide a 3D extrapolation of the dynamic of the organs observed in a single 2D MRI slice. The method is applied to the 4 main structures of the female pelvic floor (bladder, vagina, uterus and rectum), providing a full 3D visualization of the organs' displacements. The methodology is evaluated with two patient-specific data sets of volunteers presenting no pelvic pathology, and a sensitivity study is performed using synthetic data. The resulting simulations provide an estimation of the dynamic 3D shape of the organs facilitating diagnosis compared to 2D sequences. Moreover, the method follows a protocol compatible with current clinical constraints presenting this way potential short term medical applications.

Centrifugal sedimentation of an ideal suspension in a rotating tube or basket can be modelled by an initial-boundary-value problem for a scalar conservation law with a nonconvex flux function. The sought unknown is the volume fraction of solids as function of radial distance and time for constant initial data. The method of characteristics is used to construct entropy solutions. The qualitatively different solutions, which depend on the initial value and the vessel radial coordinates, are presented in detail along with numerical simulations. Based on the entropy solutions, a new method of flux identification, which does not require any prescribed functional expression, is presented and illustrated with synthetic data.

Nowadays, full-waveform inversion, based on fitting the measured surface data with modelled data, has become the preferred approach to recover detailed physical parameters from the subsurface. However, its application is computationally expensive for large inversion domains. Furthermore, when the subsurface has a complex geological setting, the inversion process requires an appropriate pre-conditioning scheme to retrieve the medium parameters for the desired target area in a reliable manner. One way of dealing with both aspects is by waveform inversion schemes in a target-oriented fashion. Therefore, we propose a prospective application of the convolution-type representation for the acoustic wavefield in the frequency–space domain formulated as a target-oriented waveform inversion method. Our approach aims at matching the observed and modelled upgoing wavefields at a target depth level in the subsurface, where the seismic wavefields, generated by sources distributed above this level, are available. The forward modelling is performed by combining the convolution-type representation for the acoustic wavefield with solving the two-way acoustic wave-equation in the frequency–space domain for the target area. We evaluate the effectiveness of our inversion method by comparing it with the full-domain full-waveform inversion process through some numerical examples using synthetic data from a horizontal well acquisition geometry, where the sources are located at the surface and the receivers are located along a horizontal well at the target level. Our proposed inversion method requires less computational effort and, for this particular acquisition, it has proven to provide more accurate estimates of the target zone below a complex overburden compared to both full-domain full-waveform inversion process and local full-waveform inversion after applying interferometry by multidimensional deconvolution to get local-impulse responses.

Accurate modelling of the conductivity structure of mineralisations can often be difficult. In order to remedy this, a parametric approach is often used. We have developed a parametric thin-sheet code, with a variable overburden. The code is capable of performing inversions of time-domain airborne electromagnetic data, and it has been tested successfully on both synthetic data and field data. The code implements an integral solution containing one or more conductive sheets, buried in a half-space with a laterally varying conductive overburden. This implementation increases the area of applicability compared to, for example, codes operating in free space, but it comes with a significant increase in computational cost. To minimise the cost, the code is parallelised using OpenMP and heavily optimised, which means that inversions of field data can be performed in hours on multiprocessor desktop computers. The code models the full system transfer function of the electromagnetic system, including variable flight height. The code is demonstrated with a synthetic example imitating a mineralisation buried underneath a conductive meadow. As a field example, the Valen mineral deposit, which is a graphite mineral deposit located in a variable overburden, is successfully inverted. Our results match well with previous models of the deposit; however, our predicted sheet remains inconclusive. These examples collectively demonstrate the effectiveness of our thin-sheet code.