Voxel-based Methods Research Articles

Enhancing Point Features with Spatial Information for Point-Based 3D Object Detection

Currently, there are many kinds of voxel-based multisensor 3D object detectors, while point-based multisensor 3D object detectors have not been fully studied. In this paper, we propose a new 3D two-stage object detection method based on point cloud and image fusion to improve the detection accuracy. To address the problem of insufficient semantic information of point cloud, we perform multiscale deep fusion of LiDAR point and camera image in a point-wise manner to enhance point features. Due to the imbalance of LiDAR points, the object point cloud in the long-distance area is sparse. We design a point cloud completion module to predict the spatial shape of objects in the candidate boxes and extract the structural information to improve the feature representation ability to further refine the boxes. The framework is evaluated on widely used KITTI and SUN-RGBD dataset. Experimental results show that our method outperforms all state-of-the-art point-based 3D object detection methods and has comparable performance to voxel-based methods as well.

Optimal Surgical Extent for Memory and Seizure Outcome in Temporal Lobe Epilepsy.

ObjectivePostoperative memory decline is an important consequence of anterior temporal lobe resection (ATLR) for temporal lobe epilepsy (TLE), and the extent of resection may be a modifiable factor. This study aimed to define optimal resection margins for cognitive outcome while maintaining a high rate of postoperative seizure freedom.MethodsThis cohort study evaluated the resection extent on postoperative structural MRI using automated voxel‐based methods and manual measurements in 142 consecutive patients with unilateral drug refractory TLE (74 left, 68 right TLE) who underwent standard ATLR.ResultsVoxel‐wise analyses revealed that postsurgical verbal memory decline correlated with resections of the posterior hippocampus and inferior temporal gyrus, whereas larger resections of the fusiform gyrus were associated with worsening of visual memory in left TLE. Limiting the posterior extent of left hippocampal resection to 55% reduced the odds of significant postoperative verbal memory decline by a factor of 8.1 (95% CI 1.5–44.4, p = 0.02). Seizure freedom was not related to posterior resection extent, but to the piriform cortex removal after left ATLR. In right TLE, variability of the posterior extent of resection was not associated with verbal and visual memory decline or seizures after surgery.InterpretationThe extent of surgical resection is an independent and modifiable risk factor for cognitive decline and seizures after left ATLR. Adapting the posterior extent of left ATLR might optimize postoperative outcome, with reduced risk of memory impairment while maintaining comparable seizure‐freedom rates. The current, more lenient, approach might be appropriate for right ATLR. ANN NEUROL 2022;91:131–144

Multiple-region grey matter atrophy as a predictor for the development of dementia in a community: the Hisayama Study

ObjectiveTo assess the association of regional grey matter atrophy with dementia risk in a general older Japanese population.MethodsWe followed 1158 dementia-free Japanese residents aged ≥65 years for 5.0 years. Regional...

Activation and deactivation patterns in schizophrenia during performance of an fMRI adapted version of the stroop task

The Stroop task, which examines an aspect of executive function/cognitive control, the ability to inhibit prepotent responses, has been relatively little examined in schizophrenia, and the findings have been inconsistent. Whether performance of this task is associated with failure of de-activation in the disorder is also uncertain. We examined 42 schizophrenic patients and 61 healthy controls during performance of an fMRI-adapted version of the Stroop task, the counting Stroop task. Task-related activations (incongruent > congruent condition) and de-activations (baseline > incongruent) were examined using whole-brain, voxel-based methods. In the healthy controls, task performance was found to be associated with activations in the left dorsolateral prefrontal cortex and the dorsal anterior cingulate cortex, among other regions. De-activations were seen in the medial frontal cortex, the middle and posterior cingulate gyrus and cuneus, the parahippocampal gyrus and the hippocampus. The schizophrenic patients did not show reduced activation compared to the healthy controls. They did, however, show failure of de-activation in the medial frontal cortex. Our negative finding with respect to hypoactivation during performance of a task requiring inhibition of prepotent responses suggests that brain functional abnormality in schizophrenia may not affect all aspects of executive function/cognitive control. The finding of medial frontal cortex failure of de-activation adds to existing findings of default mode network dysfunction in the disorder.

Altered gray matter volume and functional connectivity in patients with herpes zoster and postherpetic neuralgia

Numerous neuroimaging studies on postherpetic neuralgia (PHN) and herpes zoster (HZ) have revealed abnormalities in brain structure/microstructure and function. However, few studies have focused on changes in gray matter (GM) volume and intrinsic functional connectivity (FC) in the transition from HZ to PHN. This study combined voxel-based morphometry and FC analysis methods to investigate GM volume and FC differences in 28 PHN patients, 25 HZ patients, and 21 well-matched healthy controls (HCs). Compared to HCs, PHN patients exhibited a reduction in GM volume in the bilateral putamen. Compared with HZ patients, PHN patients showed decreased GM volume in the left parahippocampal gyrus, putamen, anterior cingulate cortex, and right caudate and increased GM volume in the right thalamus. However, no regions with significant GM volume changes were found between the HZ and HC groups. Correlation analysis revealed that GM volume in the right putamen was positively associated with illness duration in PHN patients. Furthermore, lower FCs between the right putamen and right middle frontal gyrus/brainstem were observed in PHN patients than in HCs. These results indicate that aberrant GM volumes and FC in several brain regions, especially in the right putamen, are closely associated with chronification from HZ to PHN; moreover, these changes profoundly affect multiple dimensions of pain processing. These findings may provide new insights into the pathophysiological mechanisms of PHN.

A 3D template-based point generation network for 3D reconstruction from single images

Learning-based approaches in 3D reconstruction problem have attracted researchers, due to the excellent performance of this approach in image segmentation and image classification. The increasing attention to the learning- based approach for the 3D reconstruction application is also due to the availability of 3D datasets shared publicly, such as ShapeNet and ModelNet datasets. Several deep learning approaches use voxel representation-based approaches. However, voxel-based methods suffer from inefficiency and inability to create higher dimensional 3D results. Another representation is by using point cloud representation, an unstructured 3D points in the object’s surface. However, learning such irregular structures is a challenging task due to the unordered properties of such representations. This paper proposes a new framework for 3D reconstruction of 2D images that introduces a 3D template-based point generation network. The 3D template-based point generation network infers a 3D template and generates 3D point clouds representing the reconstructed 3D object, based on an input image. The proposed network introduces two inputs, the encoded 2D image and the encoded 3D point template produced by an image classification module and a 3D template generation module. Experiments on the ShapeNet dataset show better performance than existing methods in terms of the Chamfer distance between the 3D ground-truth data and the 3D reconstructed data.

The effect of function-based and voxel-based tropospheric tomography techniques on the GNSS positioning accuracy

Tropospheric wet delay, the main source of which is water vapor, is one of the major factors affecting the accuracy of positioning techniques using microwave. Tropospheric tomography is a powerful method to reconstruct the water vapor content in four-dimensional (4D) space. This paper studies the effect of using function-based and voxel-based tropospheric tomography methods on the positioning accuracy. This examination is performed on the static and kinematic positioning modes using Global Navigation Satellite Systems (GNSS) stations under different weather conditions. After validating the results of tomography methods using radiosonde observations, the tomography-based positioning solutions, including function-based and voxel-based approaches, are compared with the positions obtained using tropospheric models. The results of two GPS stations show that the accuracy increases when applying tomography approaches. The function-based tomography is able to increase the accuracy of the up component of the static and kinematic modes by about 0.42 and 0.79 cm, respectively, compared to the voxel-based method. In addition, the use of the function-based tropospheric tomography can decrease the convergence time of the kinematic Precise Point Positioning (PPP) solution.

Cortical Thickness Estimation: A Comparison of FreeSurfer and Three Voxel-Based Methods in a Test-Retest Analysis and a Clinical Application.

The cortical thickness has been used as a biomarker to assess different cerebral conditions and to detect alterations in the cortical mantle. In this work, we compare methods from the FreeSurfer software, the Computational Anatomy Toolbox (CAT12), a Laplacian approach and a new method here proposed, based on the Euclidean Distance Transform (EDT), and its corresponding computational phantom designed to validate the calculation algorithm. At region of interest (ROI) level, within- and inter-method comparisons were carried out with a test-retest analysis, in a subset comprising 21 healthy subjects taken from the Multi-Modal MRI Reproducibility Resource (MMRR) dataset. From the Minimal Interval Resonance Imaging in Alzheimer's Disease (MIRIAD) data, classification methods were compared in their performance to detect cortical thickness differences between 23 healthy controls (HC) and 45 subjects with Alzheimer's disease (AD). The validation of the proposed EDT-based method showed a more accurate and precise distance measurement as voxel resolution increased. For the within-method comparisons, mean test-retest measures (percentages differences/intraclass correlation/Pearson correlation) were similar for FreeSurfer (1.80%/0.90/0.95), CAT12 (1.91%/0.83/0.91), Laplacian (1.27%/0.89/0.95) and EDT (2.20%/0.88/0.94). Inter-method correlations showed moderate to strong values (R > 0.77) and, in the AD comparison study, all methods were able to detect cortical alterations between groups. Surface- and voxel-based methods have advantages and drawbacks regarding computational demands and measurement precision, while thickness definition was mainly associated to the cortical thickness absolute differences among methods. However, for each method, measurements were reliable, followed similar trends along the cortex and allowed detection of cortical atrophies between HC and patients with AD.

Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection

Recent advances on 3D object detection heavily rely on how the 3D data are represented, i.e., voxel-based or point-based representation. Many existing high performance 3D detectors are point-based because this structure can better retain precise point positions. Nevertheless, point-level features lead to high computation overheads due to unordered storage. In contrast, the voxel-based structure is better suited for feature extraction but often yields lower accuracy because the input data are divided into grids. In this paper, we take a slightly different viewpoint --- we find that precise positioning of raw points is not essential for high performance 3D object detection and that the coarse voxel granularity can also offer sufficient detection accuracy. Bearing this view in mind, we devise a simple but effective voxel-based framework, named Voxel R-CNN. By taking full advantage of voxel features in a two-stage approach, our method achieves comparable detection accuracy with state-of-the-art point-based models, but at a fraction of the computation cost. Voxel R-CNN consists of a 3D backbone network, a 2D bird-eye-view (BEV) Region Proposal Network, and a detect head. A voxel RoI pooling is devised to extract RoI features directly from voxel features for further refinement. Extensive experiments are conducted on the widely used KITTI Dataset and the more recent Waymo Open Dataset. Our results show that compared to existing voxel-based methods, Voxel R-CNN delivers a higher detection accuracy while maintaining a real-time frame processing rate, i.e., at a speed of 25 FPS on an NVIDIA RTX 2080 Ti GPU. The code is available at https://github.com/djiajunustc/Voxel-R-CNN.

Shape from projections via differentiable forward projector for computed tomography

In computed tomography, the reconstruction is typically obtained on a voxel grid. In this work, however, we propose a mesh-based reconstruction method. For tomographic problems, 3D meshes have mostly been studied to simulate data acquisition, but not for reconstruction, for which a 3D mesh means the inverse process of estimating shapes from projections. In this paper, we propose a differentiable forward model for 3D meshes that bridge the gap between the forward model for 3D surfaces and optimization. We view the forward projection as a rendering process, and make it differentiable by extending recent work in differentiable rendering. We use the proposed forward model to reconstruct 3D shapes directly from projections. Experimental results for single-object problems show that the proposed method outperforms traditional voxel-based methods on noisy simulated data. We also apply the proposed method on electron tomography images of nanoparticles to demonstrate the applicability of the method on real data.

A Comparison of Two Statistical Mapping Tools for Automated Brain FDG-PET Analysis in Predicting Conversion to Alzheimer’s Disease in Subjects with Mild Cognitive Impairment

Automated voxel-based analysis methods are used to detect cortical hypometabolism typical of Alzheimer's Disease (AD) on FDG-PET brain scans. We compared the accuracy of two clinically validated tools for their ability to identify those MCI subjects progressing to AD at followup, to evaluate the impact of the analysis method on FDG-PET diagnostic performance. SPMGrid and BRASS (Hermes Medical Solutions, Stockholm, Sweden) were tested on 131 MCI and elderly healthy controls from the EADC PET dataset. The concordance between the tools was tested by correlating the quantitative parameters (z- and t-values), calculated by the two software tools, and by measuring the topographical overlap of the abnormal regions (Dice score). Three independent expert readers blindly assigned a diagnosis based on the two map sets. We used conversion to AD dementia as the gold standard. The t-map and z-map calculated with SPMGrid and BRASS, respectively, showed a good correlation (R > .50) for the majority of individual cases (128/131) and for the majority of selected regions of interest (ROIs) (98/116). The overlap of the hypometabolic patterns from the two tools was, however, poor (Dice score .36). The diagnostic performance was comparable, with BRASS showing significantly higher sensitivity (.82 versus .59) and SPMGrid showing higher specificity (.87 versus .52). Despite similar diagnostic performance in predicting conversion to AD in MCI subjects, the two tools showed significant differences, and the maps provided by the tools showed limited overlap. These results underline the urgency for standardization across FDG-PET analysis methods for their use in clinical practice.

Estimating cover fraction from TLS return intensity in coniferous and broadleaved tree shoots

Terrestrial laser scanning (TLS) provides a unique opportunity to study forest canopy structure and its spatial patterns such as foliage quantity and dispersal. Using TLS point clouds for estimating leaf area density with voxel-based methods is biased by the physical dimensions of laser beams, which violates the common assumption of beams being infinitely thin. Real laser beams have a footprint size larger than several millimeters. This leads to difficulties in estimating leaf area density from light detection and ranging (LiDAR) in vegetation, where the target objects can be of similar or even smaller size than the beam footprint. To compensate for this bias, we propose a method to estimate the per-pulse cover fraction, defined as the fraction of laser beams’ footprint area that is covered by vegetation targets, using the LiDAR return intensity and an experimental calibration measurement. We applied this method to a Leica P40 single-return instrument, and report our experimental results. We found that conifer foliage had a lower average per-pulse cover fraction than broadleaved foliage, indicating an increased number of partial hits in conifer foliage. We further discuss limitations of our method that stem from unknown target properties that influence the LiDAR return intensity and highlight potential ways to overcome the limitations and manage the remaining uncertainty. Our method’s output, the per-beam cover fraction, may be useful in a weight function for methods that estimate leaf area density from LiDAR point clouds.

Resolving bundle-specific intra-axonal T2 values within a voxel using diffusion-relaxation tract-based estimation

At the typical spatial resolution of MRI in the human brain, approximately 60–90% of voxels contain multiple fiber populations. Quantifying microstructural properties of distinct fiber populations within a voxel is therefore challenging but necessary. While progress has been made for diffusion and T1-relaxation properties, how to resolve intra-voxel T2 heterogeneity remains an open question. Here a novel framework, named COMMIT-T2, is proposed that uses tractography-based spatial regularization with diffusion-relaxometry data to estimate multiple intra-axonal T2 values within a voxel. Unlike previously-proposed voxel-based T2 estimation methods, which (when applied in white matter) implicitly assume just one fiber bundle in the voxel or the same T2 for all bundles in the voxel, COMMIT-T2 can recover specific T2 values for each unique fiber population passing through the voxel. In this approach, the number of recovered unique T2 values is not determined by a number of model parameters set a priori, but rather by the number of tractography-reconstructed streamlines passing through the voxel. Proof-of-concept is provided in silico and in vivo, including a demonstration that distinct tract-specific T2 profiles can be recovered even in the three-way crossing of the corpus callosum, arcuate fasciculus, and corticospinal tract. We demonstrate the favourable performance of COMMIT-T2 compared to that of voxelwise approaches for mapping intra-axonal T2 exploiting diffusion, including a direction-averaged method and AMICO-T2, a new extension to the previously-proposed Accelerated Microstructure Imaging via Convex Optimization (AMICO) framework.

Brainstem Volumetric Integrity in Preclinical and Prodromal Alzheimer's Disease.

Neuropathological studies have suggested the tau pathology observed in Alzheimer's disease (AD) originates in brainstem nuclei, but no studies to date have quantified brainstem volumes in clinical populations with biomarker-confirmed mild cognitive impairment (MCI) or dementia due to AD or determined the value of brainstem volumetrics in predicting dementia. The present study examined whether MRI-based brainstem volumes differ among cognitively normal older adults and those with MCI or dementia due to AD and whether preclinical brainstem volumes predict future progression to dementia. Alzheimer's Disease Neuroimaging Initiative participants (N = 1,629) underwent baseline MRI scanning with variable clinical follow-up (6-120 months). Region of interest and voxel-based morphometric methods assessed brainstem volume differences among cognitively normal (n = 814), MCI (n = 542), and AD (n = 273) participants, as well as subsets of cerebrospinal fluid biomarker-confirmed MCI (n = 203) and AD (n = 160) participants. MCI and AD cases showed smaller midbrain volumes relative to cognitively normal participants when normalizing to whole brainstem volume, and showed smaller midbrain, locus coeruleus, pons, and whole brainstem volumes when normalizing to total intracranial volume. Cognitively normal individuals who later progressed to AD dementia diagnosis exhibited smaller baseline midbrain volumes than individuals who did not develop dementia, and voxel-wise analyses revealed specific volumetric reduction of the locus coeruleus. Findings are consistent with neuropathological observations of early AD-related pathology in brainstem nuclei and further suggest the clinical relevance of brainstem substructural volumes in preclinical and prodromal AD.

A voxel-based lesion symptom mapping analysis of chronic pain in multiple sclerosis.

Pain is one of the most disabling symptoms in multiple sclerosis. Chronic pain in multiple sclerosis is often neuropathic in nature, although a clear-cut distinction with nociceptive pain is not easy. The aim of our study was to analyze the MRIs of multiple sclerosis patients with chronic pain in order to explore possible associations with lesion sites, on a voxel-by-voxel basis. We enrolled patients aged > 18 years with multiple sclerosis in accordance with the 2010 McDonald criteria. Patients meeting criteria for persistent pain (frequent or constant pain lasting > 3 months) were included in the "pain group". The other patients were included in the "no pain group". We outlined lesions on FLAIR MRI scans using a semi-automated edge finding tool. To detect the association between lesion localization and persistent pain, images were analysed with the voxel-based lesion symptom mapping methods implemented in the (nonparametric mapping software included into the MRIcron. We enrolled 208 MS patients (140 F, mean age 55.2 ± 9.4 years; 176 RR, 28 progressive MS; mean EDSS 2.0 + 2.0). Pain group included 96 patients and no pain group 112 patients. Lesions of the right dorsolateral prefrontal area were significantly more prevalent in patients without pain, whereas periventricular posterior lesions were significantly more prevalent in patients with persistent pain. Our data suggest a role of the right dorsolateral prefrontal cortex in the modulation of pain perception and in the occurrence of chronic pain in MS patients. Our data also support a hemispheric asymmetry in pain perception and modulation.

Multiphase Conductivity Imaging With Electrical Impedance Tomography and B-Spline Level Set Method

This article presents a B-spline level set (BLS)-based reconstruction scheme for multiphase conductivity imaging with electrical impedance tomography (EIT). The EIT inverse problem is recast as a shape reconstruction problem where both the inclusion interface and conductivity values are the unknown. In using this scheme, the unknown conductivity to be reconstructed is assumed to be piecewise constant while the inclusion interface is represented by the zero level set of a BLS surface, which, in turn, is represented as a linear combination of B-spline basis functions. Through such a low-order representation technique, the number of unknowns in the reconstruction problem is significantly reduced compared with the traditional pixel-/voxel-based reconstruction methods. The performance of the BLS scheme is tested with heart-and-lungs phantom studies and the cross section of the industrial pipeline phantom studies. The results demonstrate that the BLS scheme is capable of preserving sharp features of inclusions and providing estimates to both the interface locations and unknown conductivity values with reliable accuracy. In addition, we also found that the BLS scheme is quite robust to the selection of the regularization weighting parameter in the least-square-based optimization regime and the additional parameter in the Heaviside function used for modeling the conductivity profile.

Application Research on Computer Aided Simulation of Coastal Wetland Topography and Seawater Inundation Range

The computer-aided simulation of terrain optimization scheme can effectively solve the analysis problem of coastal wetland terrain and seawater inundation. The construction of three-dimensional model uses the traditional polygon-based method, which has great expression in terrain complexity and flexibility The limitations of voxel-based terrain production methods have brought various innovative ways to improve the quality of the 3D world scene while saving costs. A large part of the success of simulation technology now comes from the highly realistic 3D virtual world, which makes people immersed. And the application based on computer modeling can well meet the requirements of high-quality scenes. With the development and maturity of computer modeling technology, we are left with both opportunities and challenges. This article analyzes them.

Preclinical Voxel-Based Dosimetry in Theranostics: a Review.

Due to the increasing use of preclinical targeted radionuclide therapy (TRT) studies for the development of novel theranostic agents, several studies have been performed to accurately estimate absorbed doses to mice at the voxel level using reference mouse phantoms and Monte Carlo (MC) simulations. Accurate dosimetry is important in preclinical theranostics to interpret radiobiological dose-response relationships and to translate results for clinical use. Direct MC (DMC) simulation is believed to produce more realistic voxel-level dose distribution with high precision because tissue heterogeneities and nonuniform source distributions in patients or animals are considered. Although MC simulation is considered to be an accurate method for voxel-based absorbed dose calculations, it is time-consuming, computationally demanding, and often impractical in daily practice. In this review, we focus on the current status of voxel-based dosimetry methods applied in preclinical theranostics and discuss the need for accurate and fast voxel-based dosimetry methods for pretherapy absorbed dose calculations to optimize the dose computation time in preclinical TRT.

BionoiNet: ligand-binding site classification with off-the-shelf deep neural network.

Fast and accurate classification of ligand-binding sites in proteins with respect to the class of binding molecules is invaluable not only to the automatic functional annotation of large datasets of protein structures but also to projects in protein evolution, protein engineering and drug development. Deep learning techniques, which have already been successfully applied to address challenging problems across various fields, are inherently suitable to classify ligand-binding pockets. Our goal is to demonstrate that off-the-shelf deep learning models can be employed with minimum development effort to recognize nucleotide- and heme-binding sites with a comparable accuracy to highly specialized, voxel-based methods. We developed BionoiNet, a new deep learning-based framework implementing a popular ResNet model for image classification. BionoiNet first transforms the molecular structures of ligand-binding sites to 2D Voronoi diagrams, which are then used as the input to a pretrained convolutional neural network classifier. The ResNet model generalizes well to unseen data achieving the accuracy of 85.6% for nucleotide- and 91.3% for heme-binding pockets. BionoiNet also computes significance scores of pocket atoms, called BionoiScores, to provide meaningful insights into their interactions with ligand molecules. BionoiNet is a lightweight alternative to computationally expensive 3D architectures. BionoiNet is implemented in Python with the source code freely available at: https://github.com/CSBG-LSU/BionoiNet. Supplementary data are available at Bioinformatics online.

Measurement error and reliability of three available 3D superimposition methods in growing patients

IntroductionCone-Beam Computed Tomography (CBCT) images can be superimposed, allowing three-dimensional (3D) evaluation of craniofacial growth/treatment effects. Limitations of 3D superimposition techniques are related to imaging quality, software/hardware performance, reference areas chosen, and landmark points/volumes identification errors. The aims of this research are to determine/compare the intra-rater reliability generated by three 3D superimposition methods using CBCT images, and compare the changes observed in treated cases by these methods.MethodsThirty-six growing individuals (11–14 years old) were selected from patients that received orthodontic treatment. Before and after treatment (average 24 months apart) CBCTs were analyzed using three superimposition methods. The superimposed scans with the two voxel-based methods were used to construct surface models and quantify differences using SlicerCMF software, while distances in the landmark-derived method were calculated using Excel. 3D linear measurements of the models superimposed with each method were then compared.ResultsRepeated measurements with each method separately presented good to excellent intraclass correlation coefficient (ICC ≥ 0.825). ICC values were the lowest when comparing the landmark-based method and both voxel-based methods. Moderate to excellent agreement was observed when comparing the voxel-based methods against each other. The landmark-based method generated the highest measurement error.ConclusionsFindings indicate good to excellent intra-examiner reliability of the three 3D superimposition methods when assessed individually. However, when assessing reliability among the three methods, ICC demonstrated less powerful agreement. The measurements with two of the three methods (CMFreg/Slicer and Dolphin) showed similar mean differences; however, the accuracy of the results could not be determined.