Object Voxels Research Articles

AVFP-MVX: Multimodal VoxelNet With Attention Mechanism and Voxel Feature Pyramid

In the current 3-D object detection tasks, most algorithms are based on pure point cloud. Although LiDAR can provide target location information and contour information for detection, it is sparse, especially for long-distance objects. Besides, camera sensors can provide more detailed target color, texture information, and so on. However, if both point cloud and image data are used for object detection at the same time, the problem of large model capacity and overfitting will occur. Different modes will also produce different gradients for different subnetworks, and the entire network will be difficult to optimize. In order to solve these problems and continuously improve the performance of detection algorithms, this article designs a 3-D object detection method attention mechanism and voxel feature pyramid multimodal VoxelNet (AVFP-MVX), which uses both point cloud and image data to solve the above problems. By referring to MVX-Net, attention mechanism and voxel feature pyramid are used to improve the detection accuracy of 3-D objects. The visualization results show that the overall performance of AVFP-MVX does well, which can accurately select the target object and return to a good bounding box. Comparative tests show that the proposed method detected 91.24%, 80.45%, and 76.91% for the easy, mod, and hard targets of car, respectively, while the average accuracy of pedestrian and cyclist is 62.44% and 67.64%, respectively, which is better than the other methods. The results of ablation experiments show that when the attention mechanism and the voxel feature pyramid network (Voxel-FPN) were added, the detection accuracy of car, pedestrian, and cyclist was increased by 1.87%, 1.85%, and 1.88%, respectively.

An Efficient Connected-Component Labeling Algorithm for 3-D Binary Images

Conventional voxel-based algorithms for labeling connected components in 3D binary images use the same mask to process all object voxels. To reduce the number of times that neighboring voxels are checked when object voxels are processed, we propose an algorithm that uses two different masks for processing two different types of object voxels. One type of mask is used when the voxel preceding the object voxel being processed is an object voxel, and the other type is used otherwise. In either case, an optimal order is used for checking the voxels in the corresponding mask. Experimental results demonstrate that our proposed algorithm checked fewer voxels, and was more efficient than conventional algorithms.

An Artificial Visual System for Three Dimensional Motion Direction Detection

For mammals, enormous amounts of visual information are processed by neurons of the visual nervous system. The research of the direction selectivity is of great significance and local direction-selective ganglion neurons have been discovered. However, research is still at the one dimensional level and concentrated on a single cell. It remains challenging to explain the function and mechanism of the overall motion direction detection. In our previous papers, we have proposed a motion direction detection mechanism on the two dimensional level to solve these problems. The previous studies did not take into account that the information in the left and right retina is different and cannot be used to detect the three dimensional motion direction. Further effort is required to develop a more realistic system in three dimensions. In this paper, we propose a new three-dimensional artificial visual system to extend motion direction detection mechanism into three dimensions. We assumed that a neuron could detect the local motion of a single voxel object within three dimensional space. We also took into consideration that the information of the left and right retinas is different. Based on this binocular disparity, a realistic motion direction mechanism for three dimensions was established: the neurons received signals from the primary visual cortex of each eye and responded to motion in specific directions. There are a series of local direction-selective ganglion neurons arrayed on the retina by a logical AND operation. The response of each local direction detection neuron will be further integrated by the next neural layer to obtain the global motion direction. We carry out several computer simulations to demonstrate the validity of the mechanism. It shows that the proposed mechanism is capable of detecting the motion of complex three dimensional objects, which is consistent with most known physiological experimental results.

On the Relationship Between EB-3 Profiles and Microtubules Growth in Cultured Cells.

Microtubules are dynamic structures undergoing rapid growth and shrinkage in living cells and in vitro. The growth of microtubules in vitro was analyzed with subpixel precision (Maurer et al., Current Biology, 2014, 24 (4), 372–384); however, to what extent these results could be applied for microtubules growing in vivo remains largely unknown. Particularly, the question is whether microtubule growth velocity in cells could be sufficiently approximated by a Gaussian distribution or its variability requires a more sophisticated description? Addressing this question, we used time-lapse microscopy and mathematical modeling, and we analyzed EB-3 comets forming on microtubules of cultured cells with subpixel precision. Parameters of comets (shape, form, and velocity) were used as topological characteristics of 3D voxel objects. Using regression analysis, we determined the real positions of the microtubule tips in time-lapse sequences. By exponential decay fitting of the restored comet intensity profile, we found that in vivo EB-3 rapidly exchanges on growing microtubule ends with a decoration time ∼ 2 s. We next developed the model showing that the best correlation between comet length and microtubule end growth velocity is at time intervals close to the decoration time. In the cells, EB comet length positively correlates with microtubule growth velocity in preceding time intervals, while demonstrating no correlation in subsequent time intervals. Correlation between comet length and instantaneous growth velocity of microtubules remains under nocodazole treatment when mean values of both parameters decrease. Our data show that the growth of microtubules in living cells is well-approximated by a constant velocity with large stochastic fluctuations.

Prediction of multi-criteria optimization (MCO) parameter efficiency in volumetric modulated arc therapy (VMAT) treatment planning using machine learning (ML).

To predict the impact of optimization parameter changes on dosimetric plan quality criteria in multi-criteria optimized volumetric-modulated-arc therapy (VMAT) planning prior to optimization using machine learning (ML). A data base comprising a total of 21,266 VMAT treatment plans for 44 cranial and 18 spinal patient geometries was generated. The underlying optimization algorithm is governed by three highly composite parameters which model a combination of important aspects of the solution. Patient geometries were parametrized via volume- and shape properties of the voxel objects and overlap-volume histograms (OVH) of the planning-target-volume (PTV) and a relevant organ-at-risk (OAR). The impact of changes in one of the three optimization parameters on the maximally achievable value range of five dosimetric properties of the resulting dose distributions was studied. To predict the extent of this impact based on patient geometry, treatment site, and current parameter settings prior to optimization, three different ML-models were trained and tested. Precision-recall curves, as well as the area-under-curve (AUC) of the resulting receiver-operator-characteristic (ROC) curves were analyzed for model assessment. Successful identification of parameter regions resulting in a high variability of dosimetric plan properties depended on the choice of geometry features, the treatment indication and the plan property under investigation. AUC values between 0.82 and 0.99 could be achieved. The best average-precision (AP) values obtained from the corresponding precision/recall curves ranged from 0.71 to 0.99. Machine learning models trained on a database of pre-optimized treatment plans can help finding relevant optimization parameter ranges prior to optimization.

Holographic particle localization under multiple scattering

We introduce a novel framework that incorporates multiple scattering for large-scale 3D particle-localization using single-shot in-line holography. Traditional holographic techniques rely on single-scattering models which become inaccurate under high particle-density. We demonstrate that by exploiting multiple-scattering, localization is significantly improved. Both forward and back-scattering are computed by our method under a tractable recursive framework, in which each recursion estimates the next higher-order field within the volume. The inverse scattering is presented as a nonlinear optimization that promotes sparsity, and can be implemented efficiently. We experimentally reconstruct 100 million object voxels from a single 1-megapixel hologram. Our work promises utilization of multiple scattering for versatile large-scale applications.

МЕТОД АНАЛИЗА СНИМКОВ КОМПЬЮТЕРНОЙ ТОМОГРАФИИ НА ОСНОВЕ ПОЭТАПНОЙ БИНАРИЗАЦИИ ИЗОБРАЖЕНИЙ И ЕГО ПРОГРАММНАЯ РЕАЛИЗАЦИЯ

Introduction: Computed tomography shots of a body contain big amounts of information and require complex comprehensive analysis which takes a long time and often is difficult for a diagnostician. The diagnostics accuracy can be improved by the use of contrast dye, but this method is more expensive and has a number of contraindications. Purpose: Developing an automated object detection method for computed tomography shots which would provide physicians with the necessary information. Results: A parameterization method is proposed for objects on the layers of a CT shot. It performs sequential analysis of the layers and determines the area related to the examined object. In the course of the further analysis, the object is considered as an array of voxels, each of them characterizing an average density of the object in a particular zone. Furthermore, the areas of each layer are calculated, resulting in the total volume of the object. The determination of the selective dispersion for the density values of the object zones along with the density values histogram allow the diagnostician to make conclusions about the object properties and to evaluate the possibility of pathologies. Filtering the object voxels by their density values makes it possible to localize areas of the object with certain properties and to estimate the degree of symmetry for new growths in order to judge about their possible pathological nature. Practical relevance: The developed method along with its software implementation allows you to partially automate CT shot analysis by providing the medical specialists with the values of the examined object parameters: areas of each layer, total volume, sample variance of the density values, and symmetry estimation. This method makes it possible to localize the object zones with particular density values and to detect possible pathologic growths without the use of contrast dye.

A novel quantitative volumetric spreading index definition and assessment of astrocyte spreading in vitro.

A novel quantitative volumetric spreading index (VSI) is defined that depends on the total distance between object voxels and the contact surface plane in three-dimensional (3D) space. The VSI, which ranges from 0 to 1, is rotationally invariant around the z-axis. VSI can be used to quantify the degree of individual cell spreading, which is important for analysis of cell interactions with their environment. The VSIs of astrocytes cultured on a nanofibrillar surface and three different comparative planar surfaces have been calculated from confocal laser scanning microscope z-series images, and the effects of both culture surface and immunoreactivity on the degree of cell spreading were investigated. VSI calculations indicated a statistical correlation between increased reactivity, based on immunolabeling for glial fibrillary acidic protein, and decreased cell spreading. Further results provided a quantitative measure for the increased spreading of quiescent-like and reactive-like astrocytes on planar substrates functionalized with poly-l-lysine. © 2017 International Society for Advancement of Cytometry.

Calculation of computer-generated hologram (CGH) from 3D object of arbitrary size and viewing angle

We propose a method to calculate computer-generated hologram (CGH) from 3D object of arbitrary size and viewing angle. The spectrum relation between a 3D voxel object and the CGH wavefront is established. The CGH is generated via diffraction calculation from 3D voxel object based on the development of a scaled Fresnel diffraction algorithm. This CGH calculation method overcomes the limitations on sampling imposed by conventional Fourier transform based algorithm. The calculation and reconstruction of 3D object with arbitrary size and viewing angle is achieved. Both of simulation and optical experiments proves the validation of the proposed method.

Simple calculation of a computer-generated hologram for lensless holographic 3D projection using a nonuniform sampled wavefront recording plane.

In this paper, we present a method for calculation of a computer-generated hologram (CGH) from a 3D object. A virtual wavefront recording plane (WRP) which is close to the 3D object is established. This WRP is nonuniformly sampled according to the depth map of the 3D object. The generation of CGH only involves two nonuniform fast Fourier transform (NUFFT) and two fast Fourier transform (FFT) operations, the whole computational procedure is greatly simplified by diffraction calculation from a 2D planar image instead of 3D object voxels. Numerical simulations and optical experiments are carried out to confirm the feasibility of our proposed method. The CGH calculated with our method is capable to project zoomable 3D objects without lens.

Spatial attention enhances object coding in local and distributed representations of the lateral occipital complex

The modulation of neural activity in visual cortex is thought to be a key mechanism of visual attention. The investigation of attentional modulation in high-level visual areas, however, is hampered by the lack of clear tuning or contrast response functions. In the present functional magnetic resonance imaging study we therefore systematically assessed how small voxel-wise biases in object preference across hundreds of voxels in the lateral occipital complex were affected when attention was directed to objects. We found that the strength of attentional modulation depended on a voxel's object preference in the absence of attention, a pattern indicative of an amplificatory mechanism. Our results show that such attentional modulation effectively increased the mutual information between voxel responses and object identity. Further, these local modulatory effects led to improved information-based object readout at the level of multi-voxel activation patterns and to an increased reproducibility of these patterns across repeated presentations. We conclude that attentional modulation enhances object coding in local and distributed object representations of the lateral occipital complex.

Median prior constrained TV algorithm for sparse view low-dose CT reconstruction

It is known that lowering the X-ray tube current (mAs) or tube voltage (kVp) and simultaneously reducing the total number of X-ray views (sparse view) is an effective means to achieve low-dose in computed tomography (CT) scan. However, the associated image quality by the conventional filtered back-projection (FBP) usually degrades due to the excessive quantum noise. Although sparse-view CT reconstruction algorithm via total variation (TV), in the scanning protocol of reducing X-ray tube current, has been demonstrated to be able to result in significant radiation dose reduction while maintain image quality, noticeable patchy artifacts still exist in reconstructed images. In this study, to address the problem of patchy artifacts, we proposed a median prior constrained TV regularization to retain the image quality by introducing an auxiliary vector m in register with the object. Specifically, the approximate action of m is to draw, in each iteration, an object voxel toward its own local median, aiming to improve low-dose image quality with sparse-view projection measurements. Subsequently, an alternating optimization algorithm is adopted to optimize the associative objective function. We refer to the median prior constrained TV regularization as “TV_MP” for simplicity. Experimental results on digital phantoms and clinical phantom demonstrated that the proposed TV_MP with appropriate control parameters can not only ensure a higher signal to noise ratio (SNR) of the reconstructed image, but also its resolution compared with the original TV method.

Spherical blurred shape model for 3-D object and pose recognition: quantitative analysis and HCI applications in smart environments.

The use of depth maps is of increasing interest after the advent of cheap multisensor devices based on structured light, such as Kinect. In this context, there is a strong need of powerful 3-D shape descriptors able to generate rich object representations. Although several 3-D descriptors have been already proposed in the literature, the research of discriminative and computationally efficient descriptors is still an open issue. In this paper, we propose a novel point cloud descriptor called spherical blurred shape model (SBSM) that successfully encodes the structure density and local variabilities of an object based on shape voxel distances and a neighborhood propagation strategy. The proposed SBSM is proven to be rotation and scale invariant, robust to noise and occlusions, highly discriminative for multiple categories of complex objects like the human hand, and computationally efficient since the SBSM complexity is linear to the number of object voxels. Experimental evaluation in public depth multiclass object data, 3-D facial expressions data, and a novel hand poses data sets show significant performance improvements in relation to state-of-the-art approaches. Moreover, the effectiveness of the proposal is also proved for object spotting in 3-D scenes and for real-time automatic hand pose recognition in human computer interaction scenarios.

Complementary coded apertures for 4-dimensional x-ray coherent scatter imaging.

X-ray scattering has played a key role in non-destructive materials characterization due to the material-specific coherent scattering signatures. In the current energy dispersive coherent scatter imaging systems, including selected volume tomography and coherent scatter computed tomography, each object voxel is measured at a single scatter angle, which suffers from slow acquisition time. The employment of coded apertures in x-ray scatter imaging systems improves the photon collection efficiency, making it promising for real time volumetric imaging and material identification. In this paper, we propose a volumetric x-ray scatter imaging system using a pair of complementary coded apertures: a coded aperture on the detector side introduces multiplexed measurement on an energy-sensitive detector array; a complementary source-side coded aperture selectively illuminates the object to decouple the ambiguity due to the increased parallelization for 4D imaging. The system yields the 1D coherent scattering form factor at each voxel in 3D. We demonstrate tomographic imaging and material identification with the system and achieve a spatial resolution ~1 cm and a normalized momentum transfer resolution, Δq/q, of 0.2.

On the computation of the magnetic phase shift for magnetic nano-particles of arbitrary shape using a spherical projection model

The magnetic phase shift of an electron wave traveling through a magnetized object is computed by considering the object to be made up of a collection of uniformly magnetized spheres arranged on the nodes of a cubic grid. In the limit of vanishing grid size, this approach becomes equivalent to other numerical approaches. Update equations are derived for the change of the magnetic phase shift when the magnetization of a single object voxel is modified. Example phase shift calculations are presented for a uniformly magnetized sphere, circular disks with an infinitely sharp vortex core and a smooth core, and an oval disk with a pair of vortices and an antivortex.

A clinical gamma camera-based pinhole collimated system for high resolution small animal SPECT imaging

The main objective of the present study was to upgrade a clinical gamma camera to obtain high resolution tomographic images of small animal organs. The system is based on a clinical gamma camera to which we have adapted a special-purpose pinhole collimator and a device for positioning and rotating the target based on a computer-controlled step motor. We developed a software tool to reconstruct the target's three-dimensional distribution of emission from a set of planar projections, based on the maximum likelihood algorithm. We present details on the hardware and software implementation. We imaged phantoms and heart and kidneys of rats. When using pinhole collimators, the spatial resolution and sensitivity of the imaging system depend on parameters such as the detector-to-collimator and detector-to-target distances and pinhole diameter. In this study, we reached an object voxel size of 0.6 mm and spatial resolution better than 2.4 and 1.7 mm full width at half maximum when 1.5- and 1.0-mm diameter pinholes were used, respectively. Appropriate sensitivity to study the target of interest was attained in both cases. Additionally, we show that as few as 12 projections are sufficient to attain good quality reconstructions, a result that implies a significant reduction of acquisition time and opens the possibility for radiotracer dynamic studies. In conclusion, a high resolution single photon emission computed tomography (SPECT) system was developed using a commercial clinical gamma camera, allowing the acquisition of detailed volumetric images of small animal organs. This type of system has important implications for research areas such as Cardiology, Neurology or Oncology.

On the computation of the 〈3, 4, 5〉 curve skeleton of 3D objects

An algorithm to compute the curve skeleton of 3D objects in voxel images is presented. The skeleton is stable under isometric transformations of the object, since the algorithm is based on the use of the 〈3, 4, 5〉 weighted distance transform, which is a good approximation of the Euclidean distance transform. The 〈3, 4, 5〉 weighted distance transform is used both to identify suitable anchor points, and to efficiently check object voxels according to their distance to the background. The curve skeleton is symmetrically placed within the object, is topologically equivalent to the object, is unit-wide and provides a satisfactory representation of the object. Though the size of the object reconstructed from the curve skeleton via the reverse distance transformation is not as thick as that of the input, shape information is mostly retained by the skeleton, since all regions perceived as significant in the input can still be found in the reconstructed object.

Three-dimensional neuron tracing by voxel scooping

Tracing the centerline of the dendritic arbor of neurons is a powerful technique for analyzing neuronal morphology. In the various neuron tracing algorithms in use nowadays, the competing goals of computational efficiency and robustness are generally traded off against each other. We present a novel method for tracing the centerline of a neuron from confocal image stacks, which provides an optimal balance between these objectives. Using only local information, thin cross-sectional layers of voxels (‘scoops’) are iteratively carved out of the structure, and clustered based on connectivity. Each cluster contributes a node along the centerline, which is created by connecting successive nodes until all object voxels are exhausted. While data segmentation is independent of this algorithm, we illustrate the use of the ISODATA method to achieve dynamic (local) segmentation. Diameter estimation at each node is calculated using the Rayburst Sampling algorithm, and spurious end nodes caused by surface irregularities are then removed. On standard computing hardware the algorithm can process hundreds of thousands of voxels per second, easily handling the multi-gigabyte datasets resulting from high-resolution confocal microscopy imaging of neurons. This method provides an accurate and efficient means for centerline extraction that is suitable for interactive neuron tracing applications.

Polyhedral approximation and practical convex hull algorithm for certain classes of voxel sets

In this paper we introduce an algorithm for the creation of polyhedral approximations for certain kinds of digital objects in a three-dimensional space. The objects are sets of voxels represented as strongly connected subsets of an abstract cell complex. The proposed algorithm generates the convex hull of a given object and modifies the hull afterwards by recursive repetitions of generating convex hulls of subsets of the given voxel set or subsets of the background voxels. The result of this method is a polyhedron which separates object voxels from background voxels. The objects processed by this algorithm and also the background voxel components inside the convex hull of the objects are restricted to have genus 0. The second aim of this paper is to present some practical improvements to the discussed convex hull algorithm to reduce computation time.

A semi-supervised approach to space carving

In this paper, we present a semi-supervised approach to space carving by casting the recovery of volumetric data from multiple views into an evidence combining setting. The method presented here is statistical in nature and employs, as a starting point, a manually obtained contour. By making use of this user-provided information, we obtain probabilistic silhouettes of all successive images. These silhouettes provide a prior distribution that is then used to compute the probability of a voxel being carved. This evidence combining setting allows us to make use of background pixel information. As a result, our method combines the advantages of shape-from-silhouette techniques and statistical space carving approaches. For the carving process, we propose a new voxelated space. The proposed space is a projective one that provides a colour mapping for the object voxels which is consistent in terms of pixel coverage with their projection onto the image planes for the imagery under consideration. We provide quantitative results and illustrate the utility of the method on real-world imagery.