Voxel Scale Research Articles

Quantitative MRI analysis of structural changes in tomato tissues resulting from dehydration.

A quantitative magnetic resonance imaging (MRI) analysis at 1.5T of the effects of different dehydration regimes on transverse relaxation parameters measured in tomato tissue is presented. Multi-exponential T2 maps have been estimated for the first time, providing access to spatialized microstructural information at voxel scale. The objective was to provide a better understanding of the changes in the multi-exponential transverse relaxation parameters induced by dehydration in tomato tissues and to unravel the effects of microstructure and composition on relaxation parameters. The results led to the hypothesis that the multi-exponential relaxation signal reflects cell compartmentation and tissue heterogeneity, even at the voxel scale. Multi-exponential relaxation times provided information about water loss from specific cell compartments and seem to indicate that the dehydration process mainly affects large cells. By contrast, total signal intensity showed no sensitivity to variations in water content in the range investigated in the present study (between 95% [fresh tissue] and 90% [after dehydration]). The variation in relaxation times resulting from water loss was due to both changes in solute concentration and compartment size. The comparative analysis of the two contrasted tissues in terms of microporosity demonstrated that magnetic susceptibility effects, caused by the presence of air in the placenta tissue, significantly impact the effective relaxation and might be the dominant effect in the variations observed in relaxation times in this tissue.

Potential of combined neutron and X‐ray imaging to quantify local carbon contents in soil

AbstractIn this study, we investigated the potential and limitations of using joint X‐ray and time‐of‐flight (TOF) neutron imaging for mapping the 3‐dimensional organic carbon distribution in soil. This approach is viable because neutron and X‐ray beams have complementary attenuation properties. Soil minerals consist to a large part of silicon and aluminium, and elements that are relatively translucent to neutrons but attenuate X‐rays. In contrast, attenuation of neutrons is strong for hydrogen, which is abundant in soil organic matter (SOM), while hydrogen barely attenuates X‐rays. In theory, TOF neutron imaging does further more allow the imaging of Bragg edges, which correspond to d‐spacings in minerals. This could help to distinguish between SOM and clay minerals, the mineral group in soil that is most strongly associated with hydrogen atoms. We collected TOF neutron image data at the IMAT beamline at the ISIS facility and synchrotron X‐ray image data at the I12 beamline at the Diamond Light source, both located within the Rutherford Appleton Laboratory, Harwell, UK. The white beam (the full energy spectrum) neutron image clearly showed variations in neutron attenuation within soil aggregates at approximately constant X‐ray attenuations. This indicates a constant bulk density with varying organic matter and/or clay content. Unfortunately, the combination of TOF neutron and X‐ray imaging was not suited to allow for a distinction between SOM and clay minerals at the voxel scale. While such a distinction is possible in theory, it is prevented by technical limitations. One of the main reasons is that the neutron frequencies available at modern neutron sources are too large to capture the main d‐spacings of clay minerals. As a result, inference to voxel scale SOM concentrations is presently not feasible. Future improved neutron sources and advanced detector designs will eventually overcome the technical problems encountered here. On the positive side, combined X‐ray and TOF neutron imaging demonstrated abilities to identify quartz grains and to distinguish between plastics and plant seeds.Highlights Full understanding of biogeochemical processes requires three‐dimensional (3‐D) maps of organic matter in soil (SOM). This study investigates a novel method to map voxel‐scale SOM contents with 3‐D resolution. The method is based a combination of X‐ray and time‐of‐flight neutron tomography. At present, technical limitations prevent distinguishing between SOM and clay mineral contents. More advanced neutron sources are required to overcome the encountered technical obstacles.

Multi-scale Monte Carlo simulations of gold nanoparticle-induced DNA damages for kilovoltage X-ray irradiation in a xenograft mouse model using TOPAS-nBio

BackgroundGold nanoparticles (AuNPs) are considered as promising agents to increase the radiosensitivity of tumor cells. However, the biological mechanisms of radiation enhancement effects of AuNPs are still not well understood. We present a multi-scale Monte Carlo simulation framework within TOPAS-nBio to investigate the increase of DNA damage due to the presence of AuNPs in mouse tumor models.MethodsA tumor was placed inside a voxel mouse model and irradiated with either 100-kVp or 200-kVp X-ray beams. Phase spaces were employed to transfer particles from the macroscopic (voxel) scale to the microscopic scale, which consists of a cell geometry including a detailed mouse DNA model. Radiosensitizing effects were calculated in the presence and absence of hybrid nanoparticles with a text{Fe}_2text{O}_3 core surrounded by a gold layer (AuFeNPs). To simulate DNA damage even for very small energy tracks, Geant4-DNA physics and chemistry models were used on microscopic scale.ResultsAn AuFeNP-induced enhancement of both dose and DNA strand breaks has been established for different scenarios. Produced chemical radicals including hydroxyl molecules, which were assumed to be responsible for DNA damage through chemical reactions, were found to be significantly increased. We further observed a dependency of the results on the location of the cells within the tumor for 200-kVp X-ray beams.ConclusionOur multi-scale approach allows to study irradiation-induced physical and chemical effects on cells. We showed a potential increase in cell radiosensitization caused by relatively small concentrations of AuFeNPs. Our new methodology allows the individual adjustment of parameters in each simulation step and therefore can be used for other studies investigating the radiosensitizing effects of AuFeNPs or AuNPs in living cells.

Matrix moments of the diffusion tensor distribution and matrix-variate Gamma approximation

Diffusion MRI allows for the non-invasive investigation of the microscopic architecture of biological tissues in vivo. While being sensitive to microstructural tissue changes, diffusion MRI yields poor specificity regarding the cause of these changes. This specificity can be enhanced by using (i) signal representations that retrieve global diffusion properties of the voxel content from the features of the measured diffusion signal or (ii) models that relate these features to predefined sub-voxel tissue-specific compartments. Either on the voxel scale or within diffusion compartments, tissue microstructure can be described using a diffusion tensor distribution (DTD) P(D), which captures a collection of microscopic diffusion environments. A convenient way to estimate DTDs relies on choosing a plausible parametric functional form approximating them. In particular, the mean diffusion tensor 〈D〉 and covariance tensor C associated with this choice enable derivation of measures quantifying specific microstructural diffusion properties. However, high-dimensional DTDs can be quite intractable in practice. In this tutorial paper, we present the matrix moments of P(D), which allow computation of the mean diffusion tensor, covariance tensor and higher-order statistical tensors for any arbitrary functional choice approximating P(D). In turn, these statistical tensors are related to common diffusion measures, such as the mean diffusivity, microscopic diffusion anisotropy and variance of isotropic diffusivities. Using the well-known case of the matrix-variate Gaussian distribution to validate the matrix moments, we then use these tools on the more complex non-central matrix-variate Gamma distribution, computing its covariance tensor for the first time. Finally, we utilize this covariance tensor to design a new voxel-scale signal representation, the matrix-variate Gamma approximation, which we evaluate in vivo and compare to q-space trajectory imaging (QTI) in silico. While the matrix-variate Gamma approximation possesses important limitations explained by the structure of its newly computed covariance tensor, it presents a greater accuracy than QTI in systems exhibiting high orientational order, which suggests that non-central matrix-variate Gamma distributions appropriately describe anisotropic diffusion compartments, as in the DIAMOND model. The framework of matrix moments promotes a more widespread use of matrix-variate distributions as plausible approximations of the DTD by alleviating their intractability, thereby facilitating their design and validation.

Sensitivity of voxel-based estimations of leaf area density with terrestrial LiDAR to vegetation structure and sampling limitations: A simulation experiment

The need for fine scale description of vegetation structure is increasing as Leaf Area Density (LAD, m2/m3) becomes a critical parameter to understand ecosystem functioning and energy and mass fluxes in heterogeneous ecosystems. Terrestrial Laser Scanning (TLS) has shown great potential for retrieving the foliage area at stand, plant or voxel scales. Several sources of measurement errors have been identified and corrected over the past years. However, measurements remain sensitive to several factors, including, 1) voxel size and vegetation structure within voxels, 2) heterogeneity in sampling from TLS instrument (occlusion and shooting pattern), the consequences of which have been seldom analyzed at the scale of forest plots. In the present paper, we aimed at disentangling biases and errors in plot-scale measurements of LAD with TLS in a simulated vegetation scene. Two negative biases were formerly attributed to (i) the unsampled voxels and to (ii) the subgrid vegetation heterogeneity (i.e. clumping effect), and then quantified, thanks to a the simulation experiment providing known LAD references at voxel scale, vegetation manipulations and unbiased point estimators. We used confidence intervals to evaluate voxel-scale measurement accuracy.We found that the unsampled voxel effect (i) led to underestimations with the “mean layer” method –commonly used to fill unsampled voxels- for small voxels (0.1–0.2 m) and/or low number of scans (<4). It was explained by the spatial correlations in vegetation, which induced that dense voxels were more often occluded by dense neighbors than light voxels. The distribution of the bias was heterogeneous in canopy, the bias being stronger at mid canopy where occlusion started, but smaller in highly-occluded upper layers. This somehow counterintuitive result was explained by a more random sampling of upper layers, but could highly depend on vegetation structure.The subgrid vegetation heterogeneity effect (ii) was confirmed to increase with voxel size, yet, the magnitude of this bias -quantified with vegetation manipulation- was found to be more homogeneously-distributed than the unsampled voxel effect.Overall, we found that no scenario was unbiased. However, an intermediate voxel size (0.5 m) was the best option, because the relatively homogeneous subgrid effect could be handled with a single correction factor and voxel-scale measurements errors were reasonable. On the contrary, smaller voxels led to poor voxel-scale measurements and variable biases in magnitude and spatial distribution with sampling design. However, more similar research in other context is required to adapt these conclusions to other forest plots.

Individual Tree Crown Segmentation Directly from UAV-Borne LiDAR Data Using the PointNet of Deep Learning

Accurate individual tree crown (ITC) segmentation from scanned point clouds is a fundamental task in forest biomass monitoring and forest ecology management. Light detection and ranging (LiDAR) as a mainstream tool for forest survey is advancing the pattern of forest data acquisition. In this study, we performed a novel deep learning framework directly processing the forest point clouds belonging to the four forest types (i.e., the nursery base, the monastery garden, the mixed forest, and the defoliated forest) to realize the ITC segmentation. The specific steps of our approach were as follows: first, a voxelization strategy was conducted to subdivide the collected point clouds with various tree species from various forest types into many voxels. These voxels containing point clouds were taken as training samples for the PointNet deep learning framework to identify the tree crowns at the voxel scale. Second, based on the initial segmentation results, we used the height-related gradient information to accurately depict the boundaries of each tree crown. Meanwhile, the retrieved tree crown breadths of individual trees were compared with field measurements to verify the effectiveness of our approach. Among the four forest types, our results revealed the best performance for the nursery base (tree crown detection rate r = 0.90; crown breadth estimation R2 &gt; 0.94 and root mean squared error (RMSE) &lt; 0.2m). A sound performance was also achieved for the monastery garden and mixed forest, which had complex forest structures, complicated intersections of branches and different building types, with r = 0.85, R2 &gt; 0.88 and RMSE &lt; 0.6 m for the monastery garden and r = 0.80, R2 &gt; 0.85 and RMSE &lt; 0.8 m for the mixed forest. For the fourth forest plot type with the distribution of crown defoliation across the woodland, we achieved the performance with r = 0.82, R2 &gt; 0.79 and RMSE &lt; 0.7 m. Our method presents a robust framework inspired by the deep learning technology and computer graphics theory that solves the ITC segmentation problem and retrieves forest parameters under various forest conditions.

Voxel-Scale Conversion Mapping Informs Intrinsic Resolution in Stereolithographic Additive Manufacturing

In high-resolution stereolithography, printed parts deviate significantly from the projected photomask because of the reaction complexity on the voxel scale. To better understand the reaction proce...

Forces and flow induced by a moving intruder in a granular packing: coarse-graining and DEM simulations versus experiments

The study of stress transmission and particles’ mobility in powder beds constitutes a challenging issue to improve processes such as milling, agglomeration and kneading. In this paper the ability to promote particle mobilities by means of an intruder is investigated. An experimental setup was developed to characterize the particles’ mobility at the voxel scale. A particle image velocimetry (PIV) algorithm was employed to determine the velocity fields in the vicinity of an intruder in ascendant vertical motion. A discrete element method model was developed to simulate the same system at the particle scale. Velocity maps were determined relying on a coarse-graining procedure and compared to PIV results. During the rise of the intruder the experimental and simulated drag forces are in good agreement. Spatio-temporal correlation between the granular mobility and the state of the force network are analyzed as a function of the height of the intruder. The drag force profile exhibits a relaxation trend, noised by fluctuations whose origin are in successive loading/rupture events. An interpretation of these fluctuations is proposed with regard to the development of preferential paths in the normal force network, and to the localization of sliding at the contacts scale. Finally, we show that global strains are a consequence of the ejection of particles directly in contact with the ends of the intruder.

Impact de la méthode de segmentation dans la caractérisation des adénocarcinomes pulmonaires en TEP-TDM au 18FDG

ObjectiveThe treatment of lung adenocarcinomas is conditioned by the presence of certain genetic abnormalities. Certain quantitative parameters obtained from FDG PET-CT, at the voxel scale, provide tumour shape and texture characteristics and might predict their mutational status. Our objective was to determine the impact of the segmentation method in the characterization of lung adenocarcinomas in FDG PET-CT. MethodsForty-nine patients with pulmonary adenocarcinomas were retrospectively included, with their initial FDG PET-CT image. The studied tumours were big, heterogeneous and difficult to segment automatically. The automatic FLAB algorithm was used with and without manual adjustment. The parameters were extracted and compared to the ALK, PDL1, and KRAS status, in order to compare the performances of the two segmentation methods. Their performance was determined by the ROC curve method. ResultsSeveral parameters were significant to predict genetic status (AUC>0.65). The best performing parameters were different according to the genes studied and according to the resampling methods used. The results were less dependent on resampling in automatic segmentation without manual adjustment. The best performing parameters were volume dependent parameters for segmentation with manual adjustment, and texture parameters for automatic segmentation without adjustment. ConclusionThe study of texture parameters is more efficient in automatic segmentation that is not manually adjusted, and it is advantageous to use a manual adjustment when studying volume-dependent parameters in the case of very heterogeneous tumors.

PET imaging of neurotransmission using direct parametric reconstruction

Receptor ligand-based dynamic Positron Emission Tomography (PET) permits the measurement of neurotransmitter release in the human brain. For single-scan paradigms, the conventional method of estimating changes in neurotransmitter levels relies on fitting a pharmacokinetic model to activity concentration histories extracted after PET image reconstruction. However, due to the statistical fluctuations of activity concentration data at the voxel scale, parametric images computed using this approach often exhibit low signal-to-noise ratio, impeding characterization of neurotransmitter release. Numerous studies have shown that direct parametric reconstruction (DPR) approaches, which combine image reconstruction and kinetic analysis in a unified framework, can improve the signal-to-noise ratio of parametric mapping. However, there is little experience with DPR in imaging of neurotransmission and the performance of the approach in this application has not been evaluated before in humans. In this report, we present and evaluate a DPR methodology that computes 3-D distributions of ligand transport, binding potential (BPND) and neurotransmitter release magnitude (γ) from a dynamic sequence of PET sinograms. The technique employs the linear simplified reference region model (LSRRM) of Alpert et al. (2003), which represents an extension of the simplified reference region model that incorporates time-varying binding parameters due to radioligand displacement by release of neurotransmitter. Estimation of parametric images is performed by gradient-based optimization of a Poisson log-likelihood function incorporating LSRRM kinetics and accounting for the effects of head movement, attenuation, detector sensitivity, random and scattered coincidences. A 11C-raclopride simulation study showed that the proposed approach substantially reduces the bias and variance of voxel-wise γ estimates as compared to standard methods. Moreover, simulations showed that detection of release could be made more reliable and/or conducted using a smaller sample size using the proposed DPR estimator. Likewise, images of BPND computed using DPR had substantially improved bias and variance properties. Application of the method in human subjects was demonstrated using 11C-raclopride dynamic scans and a reward task, confirming the improved quality of the estimated parametric images using the proposed approach.

Mitigating occlusion effects in Leaf Area Density estimates from Terrestrial LiDAR through a specific kriging method

Terrestrial Laser Scanning (TLS) has been used during the past decade to capture the complexity of 3D forest canopy structures, especially Leaf or Plant Area Density (LAD/PAD). TLS data, i.e. point cloud, can be divided into voxels to estimate the three-dimensional distribution of LAD/PAD. However, the combination effects of vegetation occlusion and shooting pattern of TLS scanners lead to a highly heterogeneous sampling, which limits the reliability of some local estimates, since several voxels are either not explored or poorly-explored by laser beams. In practice, recommendations vary regarding the minimum number of beams crossing voxels or the minimum path lengths required to provide reliable predictions. In addition, assigning a value to non-explored and poorly-explored voxels is still an open question.The present work proposes a new method, called LAD-kriging, to mitigate the impact of non-uniform sampling and to increase the accuracy of LAD estimates in non-explored and poorly-explored voxels. The method takes advantage of i) an unbiased LAD estimator of known variance, which was recently developed; ii) the spatial correlation of the LAD field, which derives from vegetation clumping. LAD-kriging computes kriging weights from mathematical derivations, which takes into account both spatial dependencies in the LAD field and the reliability of the estimate available in each voxel. It was evaluated through numerical experiments, which enabled to validate the algorithm and to evaluate its performance through comparison with true references. An example application to field data shows that such spatial correlations truly exist in the field and that LAD-kriging entails to reduce sampling errors (with respect to full resolution scanning). Although a real validation is impossible due to the lack of precise references at the voxel scale, this example allows to gain confidence for its application to field data.In our realistic numerical experiment, up to 25–30% of voxels can be explored by <100 beams, thereby leading to unreliable estimates of the LAD within these voxels and frequent large local errors. In cases where voxels were not explored at all, simple methods previously reported in earlier studies, such as ignoring occlusion or assigning “mean layer” values to unexplored voxels were inefficient (RMSE = 0.72 m−1 and poor local accuracy). By contrast, LAD-kriging exhibited the smallest errors (RMSE = 0.48 m−1) and was locally more accurate. LAD-Kriging was also efficient for correcting voxels explored with few beams, i.e. unreliable, dropping RMSE from 0.92 to 0.42 m−1 in these volumes. In practice, LAD-Kriging does not require defining a reliability threshold or any arbitrary parameterization, which is convenient for field applications and can be extended to other applications than LAD with TLS, such as UAS/VAS scanning, provided that unbiased estimators of known variance are available.

Voxelated soft matter via multimaterial multinozzle 3D printing.

There is growing interest in voxelated matter that is designed and fabricated voxel by voxel1-4. Currently, inkjet-based three-dimensional (3D) printing is the only widely adopted method that is capable of creating 3D voxelated materials with high precision1-4, but the physics of droplet formation requires the use of low-viscosity inks to ensure successful printing5. By contrast, direct ink writing, an extrusion-based 3D printing method, is capable of patterning a much broader range of materials6-13. However, it is difficult to generate multimaterial voxelated matter by extruding monolithic cylindrical filaments in a layer-by-layer manner. Here we report the design and fabrication of voxelated soft matter using multimaterial multinozzle 3D (MM3D) printing, in which the composition, function and structure of the materials are programmed at the voxel scale. Our MM3D printheads exploit the diode-like behaviour that arises when multiple viscoelastic materials converge at a junction to enable seamless, high-frequency switching between up to eight different materials to create voxels with a volume approaching that of the nozzle diameter cubed. As exemplars, we fabricate a Miura origami pattern14 and a millipede-like soft robot that locomotes by co-printing multiple epoxy and silicone elastomer inks of stiffness varying by several orders of magnitude. Our method substantially broadens the palette of voxelated materials that can be designed and manufactured in complex motifs.

Ground Surface Recognition at Voxel Scale From Mobile Laser Scanning Data in Urban Environment

Ground filtering is an essential process for further classification of other features from a point cloud. This letter aims to develop an efficient method for the automatic recognition of a ground surface in urban environments from a point cloud acquired by mobile laser scanning (MLS). The MLS point cloud has large amounts of data. To decrease the calculation effort, the point clouds are first segmented into regular voxels using an octree structure. Then, voxel cloud connectivity segmentation (VCCS) is applied to generate supervoxels, which alleviate the issue of boundary cross between voxels and ground truth. Finally, several geometric features on the voxel scale are selected for a support vector machine (SVM) to label segments as a ground surface and nonground objects. For an objective comparison with other methods, the accuracy of the proposed method is evaluated by public MLS data sets. Experimental results show that all ground type objects can be well recognized with least susceptibility to scenic change.

Microstructure-based multi-scale evaluation of fluid flow in an anthracite coal sample with partially-percolating voxels

Understanding fluid flow behavior in coal is of great significance for coal-bed methane exploration. X-ray CT and image segmentation have been widely used to extract pore network and generate flow field grids for flow simulation in coal samples. However, these techniques have fundamental limitations for the multi-scale characterization of coal samples, where the sub-voxel scale details could not be resolved for millimeter scale macroscopic samples. This makes it difficult to simulate the multi-scale flow behavior of fluid transport in coal sample with varying pore scales. The primary challenge is to make connection between simulation results of different scales. In the present work, multi-scale fluid flow in an anthracite coal sample was simulated by incorporating the data-constrained modeling (DCM), molecular dynamics (MD) method and partially-percolating lattice Boltzmann method (PP-LBM). In this multi-scale simulation method, three-dimensional (3D) flow field containing multi-scale structural information of the coal sample was generated by combining DCM with multi-energy synchrotron radiation CT. Multi-scale fluid flow was simulated by PP-LBM. In PP-LBM, an effective percolation fraction parameter which represents the effective volume fraction of the fluid that contributed to the flow for the voxel was used as a bridge to connect the fluid flow pattern of sub-voxel scales and voxel scales. The effective percolation fraction of a voxel versus its porosity was derived by MD simulations at the sub-voxel size level. The 3D distribution of fluid speed in the coal sample and its permeability were obtained by this multi-scale method. The numerical results are consistent with published laboratory measurements. Our proposed approach incorporated multi-scale effects and offered a more realistic fluid transport simulation method for a coal sample with varying pore size scales from the microscopic to macroscopic level. The method would be applicable for fluid transport simulations for other multi-scale porous materials.

IMAge/enGINE: a freely available software for rapid computation of high-dimensional quantification.

High-dimensional image data including diffusion weighted imaging, diffusion tensor imaging and dynamic imaging are important in exploring the connectivity, cellularity, pharmacokinetic and blood supply. IMAge/enGINE is software especially designed for high-dimensional medical image computing. IMAge/enGINE is implemented based on open-source and cross-platform tools such as Qt, ITK and VTK. It processes the high-dimensional image data in a slice-by-slice computation mechanism. For computational efficiency, C++ is used for implementing IMAge/enGINE and multi-thread computing is handled in the scale of voxels. The architecture of IMAge/enGINE is modularized for easier extension. IMAge/enGINE has following features: (I) IMAge/enGINE is free for research use; (II) it has an easy-to-use graphic user interface designed for clinical users without programming or engineering background; (III) its frame work is open-source and extensible. Developers can implement algorithms as modules and integrate them into IMAge/enGINE or generate their own application. The source of IMAge/enGINE is hosted at https://github.com/VusionMed/IMAge-enGINE. Multiple diffusion and perfusion models are implemented and integrated into IMAge/enGINE and its binaries can be downloaded freely at http://www.vusion.com.cn/?page_id=14971.

Comparative assessment of linear least-squares, nonlinear least-squares, and Patlak graphical method for regional and local quantitative tracer kinetic modeling in cerebral dynamic 18 F-FDG PET.

Dynamic 18 F-FDG PET allows quantitative estimation of cerebral glucose metabolism both at the regional and local (voxel) level. Although sensitive to noise and highly computationally expensive, nonlinear least-squares (NLS) optimization stands as the reference approach for the estimation of the kinetic model parameters. Nevertheless, faster techniques, including linear least-squares (LLS) and Patlak graphical method, have been proposed to deal with high resolution noisy data, representing a more adaptable solution for routine clinical implementation. Former research investigating the relative performance of the available algorithms lack precise evaluation of kinetic parameter estimates under realistic acquisition conditions. The present study aims at the systematic comparison of the feasibility and pertinence of kinetic modeling of dynamic cerebral 18 F-FDG PET using NLS, LLS, and Patlak method, based on numerical simulations and patient data. Numerical simulations were used to study the bias and variance of K1 and Ki parameters estimation under representative noise levels. Patient data allowed to assess the concordance between the three methods at the regional and voxel scale, and to evaluate the robustness of the estimations with respect to patient head motion. Our findings indicate that at the regional level NLS and LLS provide kinetic parameter estimates (K1 and Ki ) with similar bias and variance characteristics (K1 bias±relative standard deviation [RSD] 0.0±5.1% and 0.1%±4.9% for NLS and LLS respectively, Ki bias±RSD 0.1%±4.5% and -0.7%±4.4% for NLS and LLS respectively). NLS estimates appear, however, to be slightly less sensitive to patient motion. At the voxel level, provided that patient motion is negligible or corrected, LLS offers an appealing alternative solution for local K1 mapping. It yields K1 estimates that are highly correlated, with high correlation with NLS values (Pearson's r=0.95 on actual data) within computations times less than two orders of magnitude lower. Last, Patlak method appears as the most robust and accurate technique for the estimation of Ki values at the regional and voxel scale, with or without head motion. It provides low bias/low variance Ki quantification (bias±RSD -1.5±9.5% and -4.1±19.7% for Patlak and NLS respectively) as well as smooth parametric images suitable for visual assessment.

MVF-CNN: Fusion of Multilevel Features for Large-Scale Point Cloud Classification

Due to the disorder, sparsity, and irregularity of point clouds, the accurate classification of large-scale point clouds is a challenging problem. Voxelization-based deep learning methods have been applied to point cloud classification and have achieved good performances. However, there are some problems in the methods, such as voxels lacking color information, single receptive fields being considered only at the same voxel scale, and only the global features of voxels being considered in the classification. This paper proposes a deep learning-based algorithm for large-scale point cloud classification through the fusion of multiscale voxels and features (MVF-CNN). First, the point cloud is transformed into voxels of two different sizes. Then, the voxels are input into the 3D convolutional neural network (3D CNN) with three different scale receptive fields for the feature extraction. Next, the output features of the 3D CNN are input into the proposed global and local feature fusion classification network (GLNet), which fuses the global features of the voxels at the main branch and the local features of each voxel at the auxiliary branch. Finally, the multiscale features of the main branch are fused, and its classification results are obtained. We have conducted the experiments on six-point cloud scenes. The experimental results show that the proposed algorithm accurately classifies large-scale point clouds. Compared with several semisupervised/supervised learning methods, the proposed algorithm obtains better classification results. In addition, the experimental results also demonstrate that the proposed algorithm has a strong generalization ability and obviously has a better classification performance than the compared algorithms.

Anatomical accuracy of standard-practice tractography algorithms in the motor system - A histological validation in the squirrel monkey brain

For two decades diffusion fiber tractography has been used to probe both the spatial extent of white matter pathways and the region to region connectivity of the brain. In both cases, anatomical accuracy of tractography is critical for sound scientific conclusions. Here we assess and validate the algorithms and tractography implementations that have been most widely used - often because of ease of use, algorithm simplicity, or availability offered in open source software. Comparing forty tractography results to a ground truth defined by histological tracers in the primary motor cortex on the same squirrel monkey brains, we assess tract fidelity on the scale of voxels as well as over larger spatial domains or regional connectivity. No algorithms are successful in all metrics, and, in fact, some implementations fail to reconstruct large portions of pathways or identify major points of connectivity. The accuracy is most dependent on reconstruction method and tracking algorithm, as well as the seed region and how this region is utilized. We also note a tremendous variability in the results, even though the same MR images act as inputs to all algorithms. In addition, anatomical accuracy is significantly decreased at increased distances from the seed. An analysis of the spatial errors in tractography reveals that many techniques have trouble properly leaving the gray matter, and many only reveal connectivity to adjacent regions of interest. These results show that the most commonly implemented algorithms have several shortcomings and limitations, and choices in implementations lead to very different results. This study should provide guidance for algorithm choices based on study requirements for sensitivity, specificity, or the need to identify particular connections, and should serve as a heuristic for future developments in tractography.

Estimators and confidence intervals for plant area density at voxel scale with T-LiDAR

Estimating leaf and plant area density with Terrestrial Laser Scanners (TLS) continues to be more and more popular, as tridimensional point clouds appear as an appealing measurement technique for heterogeneous environments. Some approaches implement a discretization of the point cloud in a grid (referred to as “voxel-based”) to account for this vegetation heterogeneity and significant work has been done in this recent research field, but no general theoretical analysis is available. Although estimators have been proposed and several causes of biases have been identified, their unbiasedness (zero bias) and efficiency (smallest error) have not been evaluated. Also, confidence intervals are almost never provided.In the present paper, we assumed that the vegetation elements were randomly distributed within voxels and that TLS beams were infinitely thin, in order to focus on the remaining sources of biases and errors. In this simplified context, we both solve the transmittance equation and use the Maximum Likelihood Estimation (MLE), to derive some new estimators for the attenuation coefficient, which is proportional to leaf area density at voxel scale in this idealized context. These estimators include bias corrections and confidence intervals, and account for the number of beams crossing the voxel (beam number), the inequality of path lengths in voxel, the size of vegetation elements, as well as for the variability of element positions between vegetation samples. These theoretical derivations are complemented by numerous numerical simulations for the evaluation of estimator bias and efficiency, as well as the assessment of the coverage probabilities of confidence intervals. Our simulations reveal that the usual estimators are biased and exhibit 95% confidence intervals on the order of ±100% of the estimate, when the beam number is smaller than 30. Second, our bias-corrected estimators -especially the bias-corrected MLE- are truly unbiased and efficient in a wider range of validity than the usual ones, even for beam number as low as 5. Third, we found that the confidence intervals can be as high as ≈ ± 50% when the projected area of a single element was on the order of 10% of voxel cross-sectional area and vegetation was dense (optical depth of the voxel equal to 2), even for a beam number larger than 1000. This is explained by the variability of element positions between vegetation samples, which implies that a significant part of residual error is caused by random effects. When LAD estimates are averaged over several small voxels -typically to determine a vertical profile at plot scale or to compute the total leaf area of a single plant-, confidence intervals are typically on the order of ±5 to 10% with bias-corrected estimators, which is twice as small as with usual estimators.Our study provides some new ready-to-use estimators and confidence intervals for attenuation coefficients, which are unbiased and efficient within a fairly large range of parameter values. The unbiasedness is achieved for a fairly low beam number, which is promising for application to airborne LiDAR data. They permit to raise the level of understanding and confidence on LAD estimation. Among other applications, their usage should help determine the most suitable voxel size, for given vegetation types and scanning density, whereas existing guidelines are highly variable among studies, probably because of differences in vegetation, scanning design and estimators. The impact of other sources of biases and errors, such as vegetation heterogeneity inside voxels or TLS specifications are not addressed in the present manuscript and would require further investigations.

Measuring and modelling energy partitioning in canopies of varying complexity using MAESPA model

Evapotranspiration and energy partitioning are complex to estimate because they result from the interaction of many different processes, especially in multi-species and multi-strata ecosystems. We used MAESPA model, a mechanistic, 3D model of coupled radiative transfer, photosynthesis, and balances of energy and water, to simulate the partitioning of energy and evapotranspiration in homogeneous tree plantations, as well as in heterogeneous multi-species, multi-strata agroforests with diverse spatial scales and management schemes. The MAESPA model was modified to add (1) calculation of foliage surface water evaporation at the voxel scale; (2) computation of an average within-canopy air temperature and vapour pressure; and (3) use of (1) and (2) in iterative calculations of soil and leaf temperatures to close ecosystem-level energy balances. We tested MAESPA model simulations on a simple monospecific Eucalyptus stand in Brazil, and also in two complex, heterogeneous Coffea agroforests in Costa Rica. MAESPA satisfactorily simulated the daily and seasonal dynamics of net radiation (RMSE = 29.6 and 28.4 W m−2; R2 = 0.99 and 0.99 for Eucalyptus and Coffea sites respectively) and its partitioning between latent-(RMSE = 68.1 and 37.2 W m−2; R2 = 0.87 and 0.85) and sensible-energy (RMSE = 54.6 and 45.8 W m−2; R2 = 0.57 and 0.88) over a one-year simulation at half-hourly time-step. After validation, we use the modified MAESPA to calculate partitioning of evapotranspiration and energy between plants and soil in the above-mentioned agro-ecosystems. In the Eucalyptus plantation, 95% of the outgoing energy was emitted as latent-heat, while the Coffea agroforestry system’s partitioning between sensible and latent-heat fluxes was roughly equal. We conclude that MAESPA process-based model has an appropriate balance of detail, accuracy, and computational speed to be applicable to simple or complex forest ecosystems and at different scales for energy and evapotranspiration partitioning.