Ground Truth Model Research Articles

Optical Flow Approach Followed by SVM Classification Model to Recognize Abnormal Behavior of a Crowd

A many of researches have been carried out in the field of the crowd behavior recognition system. Recognizing crowd behavior in videos is most challenging and occlusions because of irregular human movement. This paper gives an overview of optical flow model along with the SVM (Support Vector Machine) classification model. This proposed approach evaluates sudden changes in motion of an event and classifies that event to a category: Normal and Abnormal. Geometric means of location, direction, and displacement of the feature points of each frame are estimated. Harris corner Detector is used in each frame for tracking a set of feature points. Proposed approach is very effective in real time scenario like public places where security is most important. After analyzing result ROC curve (receiver operating characteristics) is plotted which gives classification accuracy. We also presented frame level comparison with Ground truth and social force model (SFM) techniques. Our proposed approach is giving a promising result compare to all state of art methods.

THERMAL TEXTURE GENERATION AND 3D MODEL RECONSTRUCTION USING SFM AND GAN

Abstract. Realistic 3D models with textures representing thermal emission of the object are widely used in such fields as dynamic scene analysis, autonomous driving, and video surveillance. Structure from Motion (SfM) methods provide a robust approach for the generation of textured 3D models in the visible range. Still, automatic generation of 3D models from the infrared imagery is challenging due to an absence of the feature points and low sensor resolution. Recent advances in Generative Adversarial Networks (GAN) have proved that they can perform complex image-to-image transformations such as a transformation of day to night and generation of imagery in a different spectral range. In this paper, we propose a novel method for generation of realistic 3D models with thermal textures using the SfM pipeline and GAN. The proposed method uses visible range images as an input. The images are processed in two ways. Firstly, they are used for point matching and dense point cloud generation. Secondly, the images are fed into a GAN that performs the transformation from the visible range to the thermal range. We evaluate the proposed method using real infrared imagery captured with a FLIR ONE PRO camera. We generated a dataset with 2000 pairs of real images captured in thermal and visible range. The dataset is used to train the GAN network and to generate 3D models using SfM. The evaluation of the generated 3D models and infrared textures proved that they are similar to the ground truth model in both thermal emissivity and geometrical shape.

A study of MRI gradient echo signals from discrete magnetic particles with considerations of several parameters in simulations

Modeling MRI signal behaviors in the presence of discrete magnetic particles is important, as magnetic particles appear in nanoparticle labeled cells, contrast agents, and other biological forms of iron. Currently, many models that take into account the discrete particle nature in a system have been used to predict magnitude signal decays in the form of R2* or R2′ from one single voxel. Little work has been done for predicting phase signals. In addition, most calculations of phase signals rely on the assumption that a system containing discrete particles behaves as a continuous medium. In this work, numerical simulations are used to investigate MRI magnitude and phase signals from discrete particles, without diffusion effects. Factors such as particle size, number density, susceptibility, volume fraction, particle arrangements for their randomness, and field of view have been considered in simulations. The results are compared to either a ground truth model, theoretical work based on continuous mediums, or previous literature. Suitable parameters used to model particles in several voxels that lead to acceptable magnetic field distributions around particle surfaces and accurate MR signals are identified. The phase values as a function of echo time from a central voxel filled by particles can be significantly different from those of a continuous cubic medium. However, a completely random distribution of particles can lead to an R2′ value which agrees with the prediction from the static dephasing theory. A sphere with a radius of at least 4 grid points used in simulations is found to be acceptable to generate MR signals equivalent from a larger sphere. Increasing number of particles with a fixed volume fraction in simulations reduces the resulting variance in the phase behavior, and converges to almost the same phase value for different particle numbers at each echo time. The variance of phase values is also reduced when increasing the number of particles in a fixed voxel. These results indicate that MRI signals from voxels containing discrete particles, even with a sufficient number of particles per voxel, cannot be properly modeled by a continuous medium with an equivalent susceptibility value in the voxel.

Attractor learning in synchronized chaotic systems in the presence of unresolved scales

Recently, supermodels consisting of an ensemble of interacting models, synchronizing on a common solution, have been proposed as an alternative to the common non-interactive multi-model ensembles in order to improve climate predictions. The connection terms in the interacting ensemble are to be optimized based on the data. The supermodel approach has been successfully demonstrated in a number of simulation experiments with an assumed ground truth and a set of good, but imperfect models. The supermodels were optimized with respect to their short-term prediction error. Nevertheless, they produced long-term climatological behavior that was close to the long-term behavior of the assumed ground truth, even in cases where the long-term behavior of the imperfect models was very different. In these supermodel experiments, however, a perfect model class scenario was assumed, in which the ground truth and imperfect models belong to the same model class and only differ in parameter setting. In this paper, we consider the imperfect model class scenario, in which the ground truth model class is more complex than the model class of imperfect models due to unresolved scales. We perform two supermodel experiments in two toy problems. The first one consists of a chaotically driven Lorenz 63 oscillator ground truth and two Lorenz 63 oscillators with constant forcings as imperfect models. The second one is more realistic and consists of a global atmosphere model as ground truth and imperfect models that have perturbed parameters and reduced spatial resolution. In both problems, we find that supermodel optimization with respect to short-term prediction error can lead to a long-term climatological behavior that is worse than that of the imperfect models. However, we also show that attractor learning can remedy this problem, leading to supermodels with long-term behavior superior to the imperfect models.

Interactive patient-specific 3D approximation of scapula bone shape from 2D X-ray images using landmark-constrained statistical shape model fitting.

We report on an interactive tool for patientspecific 3D approximation of scapula bone shape from 2D X-ray images using landmark-constrained statistical shape model (SSM) fitting. The 3D localization of points on the 2D X-ray images was done through X-ray stereophotogrammetry. The inferior angle, acromion and the coracoid process were identified as reliable landmarks from the anteroposterior (AP) and oblique lateral views in a landmark selection study. The 3D scapula surface was approximated through fitting the scapula SSM to the 3D reconstructed coordinates of the selected landmarks. 3D point localization yielded average (X, Y, Z) coordinate reconstruction errors of (X=0.14, Y=0.07, Z=0.04) mm. The landmark-constrained fitting algorithm yielded an average error between the mean posterior model landmarks and the corresponding target landmarks of 0.49 mm using the three landmarks, and later 0.19 mm with sixteen landmarks. Average surface to surface error between the CT ground truth model and approximated model from within the dataset improved from 3.20 mm to 2.46 mm using three landmarks and using sixteen landmarks, respectively. Average surface to surface error between the CT ground truth model and the approximated model from outside the dataset improved from 4.28 mm to 3.20 mm using three landmarks and using sixteen landmarks, respectively.

Assessment of the Accuracy of Optical Shape Sensing for Needle Tracking Interventions

Accurate needle guidance is essential for a number of magnetic resonance imaging (MRI)-guided percutaneous procedures, such as radiofrequency ablation (RFA) of metastatic liver tumors. A promising technology to obtain real-time tracking of the shape and tip of a needle is by using high-frequency (up to 20 kHz) fiber Bragg grating (FBG) sensors embedded in optical fibers, which are insensitive to external magnetic fields. We fabricated an MRI-compatible needle designed for percutaneous procedures with a series of FBG sensors which would be tracked in an image-guidance system, allowing to display the needle shape within a navigation image. A series of phantom experiments demonstrated needle tip tracking errors of 1.05 ± 0.08 mm for a needle deflection up to 16.82 mm on a ground-truth model and showed nearly similar accuracy to electromagnetic (EM) tracking (i.e., 0.89 ± 0.09 mm). We demonstrated feasibility of the FBG-based tracking system for MRI-guided interventions with differences under 1 mm between tracking systems. This study establishes the needle tracking accuracy of FBG needle tracking for image-guided procedures.

Two-view underwater 3D reconstruction for cameras with unknown poses under flat refractive interfaces

In an underwater imaging system, a refractive interface is introduced when a camera looks into the water-based environment, resulting in distorted images due to the refraction of light. Simply ignoring the refraction effect or using the lens radial distortion model causes erroneous 3D reconstruction. This paper deals with a general underwater imaging setup using two cameras, of which each camera is placed in a separate waterproof housing with a flat glass window. In order to handle refraction properly, a simplified refractive camera model is used in this paper. Based on two new concepts, namely the Ellipse of Refrax (EoR) and the Refractive Depth (RD) of a scene point, we derive two new formulations of the underwater known rotation structure and motion (SaM) problem. One gives a globally optimal solution and the other is robust to outliers. The constraint of known rotation is further relaxed by incorporating the robust known rotation SaM into a new hybrid optimization framework. Our method is able to simultaneously perform underwater camera calibration and 3D reconstruction automatically without using any calibration object or additional calibration device. In order to evaluate the performance and practical applicability of our method, extensive experiments using synthetic data, synthetically rendered images and real underwater images were carried out. The experimental results demonstrate that the proposed method can significantly improve the accuracy of the reconstructed 3D structure (within 0.78 mm for an object of dimension over 200 mm compared with the ground truth model captured by a land-based system) and of the system parameters for underwater applications. Compared with bundle adjustment using the refractive camera model initialized with traditional 3D reconstruction methods, our proposed optimization method has significantly better completeness and accuracy and lower 3D errors in the reconstructed models.

Canny edge-based deformable image registration

This work focuses on developing a 2D Canny edge-based deformable image registration (Canny DIR) algorithm to register in vivo white light images taken at various time points. This method uses a sparse interpolation deformation algorithm to sparsely register regions of the image with strong edge information. A stability criterion is enforced which removes regions of edges that do not deform in a smooth uniform manner. Using a synthetic mouse surface ground truth model, the accuracy of the Canny DIR algorithm was evaluated under axial rotation in the presence of deformation. The accuracy was also tested using fluorescent dye injections, which were then used for gamma analysis to establish a second ground truth. The results indicate that the Canny DIR algorithm performs better than rigid registration, intensity corrected Demons, and distinctive features for all evaluation matrices and ground truth scenarios. In conclusion Canny DIR performs well in the presence of the unique lighting and shading variations associated with white-light-based image registration.

Distinguishing Normative Processes From Noise

This research evaluates the relative merits of two established and two newly proposed methods for modeling impressions of social events: stepwise regression, ANOVA, Bayesian model averaging, and Bayesian model sampling. Models generated with each method are compared against a ground truth model to assess performance at variable selection and coefficient estimation. We also assess the theoretical impacts of different modeling choices. Results show that the ANOVA procedure has a significantly lower false discovery rate than stepwise regression, whereas Bayesian methods exhibit higher true positive rates and comparable false discovery rates to ANOVA. Bayesian methods also generate coefficient estimates with less bias and variance than either stepwise regression or ANOVA. We recommend the use of Bayesian methods for model specification in affect control theory.

5-ALA Photodynamic Therapy in Neurosurgery, Towards the Design of a Treatment Planning System: A Proof of Concept

Purpose. Glioblastoma (GBM) treatment still remains a complex challenge. Among alternatives or adjuvant therapies, photodynamic therapy (5-ALA PDT) appears to be a promising approach. 5-ALA PDT can be delivered intraoperatively, early after tumor resection, or interstitially according to brain tumor location. A treatment planning system was designed to manage dosimetry issues before PDT delivery.Methods. The TPS was developed according to a specific workflow from stereotactic image registration to light fluence rate modeling. Here, we describe a proof of concept of a treatment planning system (TPS) dedicated to interstitial 5-ALA PDT. This tool enables the planning of a whole treatment in surgical stereotactic conditions. Stereotactic registration and dosimetry components are detailed and evaluated. The registration process is compared to a commercial solution (Leksell Gamma Plan®, Elekta®, Sweden) defined as the ground truth and dosimetry model implemented in our TPS and is compared to numerical simulations.Results. Registration achieved a sub-millimetric mean relative error that matched the standard MRI resolution. Dosimetry comparison showed a negligible error between analytical and numerical models and enabled a validation of the dosimetry algorithm implemented.Conclusions. A treatment planning system was designed to achieve 5-ALA PDT simulations before the patients underwent surgery. Similarly, for radiation therapy, we proposed a system to plan and evaluate the 5-ALA PDT dosimetry for optimizing treatment delivery. Although this system remains to be perfected, this preliminary work aimed to demonstrate the feasibility of planning 5-ALA PDT treatments in stereotactic conditions. Future improvements will mainly focus on the optimization of the treatment delivery, automatic segmentation and GPU-accelerated Monte-Carlo management to take into account GBM tissue heterogeneity.

Low-Cost Methodology for Skin Strain Measurement of a Flexed Biological Limb.

The purpose of this manuscript is to compute skin strain data from a flexed biological limb, using portable, inexpensive, and easily available resources. We apply and evaluate this approach on a person with bilateral transtibial amputations, imaging left and right residual limbs in extended and flexed knee postures. We map 3-D deformations to a flexed biological limb using freeware and a simple point-and-shoot camera. Mean principal strain, maximum shear strain, as well as lines of maximum, minimum, and nonextension are computed from 3-D digital models to inform directional mappings of the strain field for an unloaded residual limb. Peak tensile strains are ∼0.3 on the anterior surface of the knee in the proximal region of the patella, whereas peak compressive strains are ∼ -0.5 on the posterior surface of the knee. Peak maximum shear strains are ∼0.3 on the posterior surface of the knee. The accuracy and precision of this methodology are assessed for a ground-truth model. The mean point location distance is found to be 0.08 cm, and the overall standard deviation for point location difference vectors is 0.05 cm. This low-cost and mobile methodology may prove critical for applications such as the prosthetic socket interface where whole-limb skin strain data are required from patients in the field outside of traditional, large-scale clinical centers. Such data may inform the design of wearable technologies that directly interface with human skin.

Simulating cryo electron tomograms of crowded cell cytoplasm for assessment of automated particle picking.

BackgroundCryo-electron tomography is an important tool to study structures of macromolecular complexes in close to native states. A whole cell cryo electron tomogram contains structural information of all its macromolecular complexes. However, extracting this information remains challenging, and relies on sophisticated image processing, in particular for template-free particle extraction, classification and averaging. To develop these methods it is crucial to realistically simulate tomograms of crowded cellular environments, which can then serve as ground truth models for assessing and optimizing methods for detection of complexes in cell tomograms.ResultsWe present a framework to generate crowded mixtures of macromolecular complexes for realistically simulating cryo electron tomograms including noise and image distortions due to the missing-wedge effects. Simulated tomograms are then used for assessing the template-free Difference-of-Gaussian (DoG) particle-picking method to detect complexes of different shapes and sizes under various crowding and noise levels. We identified DoG parameter settings that maximize precision and recall for detecting particles over a wide range of sizes and shapes. We observed that medium sized DoG scaling factors showed the overall best performance. To further improve performance, we propose a combination strategy for integrating results from multiple parameter settings. With increasing macromolecular crowding levels, the precision of particle picking remained relatively high, while the recall was dramatically reduced, which limits the detection of sufficient copy numbers of complexes in a crowded environment. Over a wide range of increasing noise levels, the DoG particle picking performance remained stable, but dramatically reduced beyond a specific noise threshold.ConclusionsAutomatic and reference-free particle picking is an important first step in a visual proteomics analysis of cell tomograms. However, cell cytoplasm is highly crowded, which makes particle detection challenging. It is therefore important to test particle-picking methods in a realistic crowded setting. Here, we present a framework for simulating tomograms of cellular environments at high crowding levels and assess the DoG particle picking method. We determined optimal parameter settings to maximize the performance of the DoG particle-picking method.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-016-1283-3) contains supplementary material, which is available to authorized users.

Inferring connectivity in networked dynamical systems: Challenges using Granger causality.

Determining the interactions and causal relationships between nodes in an unknown networked dynamical system from measurement data alone is a challenging, contemporary task across the physical, biological, and engineering sciences. Statistical methods, such as the increasingly popular Granger causality, are being broadly applied for data-driven discovery of connectivity in fields from economics to neuroscience. A common version of the algorithm is called pairwise-conditional Granger causality, which we systematically test on data generated from a nonlinear model with known causal network structure. Specifically, we simulate networked systems of Kuramoto oscillators and use the Multivariate Granger Causality Toolbox to discover the underlying coupling structure of the system. We compare the inferred results to the original connectivity for a wide range of parameters such as initial conditions, connection strengths, community structures, and natural frequencies. Our results show a significant systematic disparity between the original and inferred network, unless the true structure is extremely sparse or dense. Specifically, the inferred networks have significant discrepancies in the number of edges and the eigenvalues of the connectivity matrix, demonstrating that they typically generate dynamics which are inconsistent with the ground truth. We provide a detailed account of the dynamics for the Erdős-Rényi network model due to its importance in random graph theory and network science. We conclude that Granger causal methods for inferring network structure are highly suspect and should always be checked against a ground truth model. The results also advocate the need to perform such comparisons with any network inference method since the inferred connectivity results appear to have very little to do with the ground truth system.

Erratum: Evaluation of the tool "Reg Refine" for user-guided deformable image registration.

"Reg Refine" is a tool available in the MIM Maestro v6.4.5 platform (www.mim-software.com) that allows the user to actively participate in the deformable image registration process. The purpose of this work was to evaluate the efficacy of this tool and investigate strategies for how to apply it effectively. This was done by performing DIR on two publicly available ground-truth models, the Pixel-based Breathing Thorax Model (POPI) for lung, and the Deformable Image Registration Evaluation Project (DIREP) for head and neck. Image noise matched in both magnitude and texture to clinical CBCT scans was also added to each model to simulate the use case of CBCT-CT alignment. For lung, the results showed Reg Refine effective at improving registration accuracy when controlled by an expert user within the context of large lung deformation. CBCT noise was also shown to have no effect on DIR performance while using the MIM algorithm for this site. For head and neck, the results showed CBCT noise to have a large effect on the accuracy of registration, specifically for low-contrast structures such as the brain-stem and parotid glands. In these cases, the Reg Refine tool was able to improve the registration accuracy when controlled by an expert user. Several strategies for how to achieve these results have been outlined to assist other users and provide feedback for developers of similar tools.

A multi-criteria evaluation platform for segmentation algorithms.

The purpose of this paper is to present a platform for evaluating segmentation algorithms that detect anatomical structures in medical images. Structure detection being subject to human interpretation, we first describe a method to define a ground truth model, i.e. a generated bronze standard, that will be the reference for subsequent analysis. This bronze standard will be characterized in order to retrieve its confidence level that will later be used to normalize the algorithm evaluation. We then describe how the developed platform helps in evaluating algorithm performances described using five evaluation criteria: accuracy, reliability, robustness, under/over segmentation sensitivity and outlier sensitivity. First, we explain how to extract those evaluation criteria using specific normalized metrics commonly found in the literature, then we present how to combine all the information in order to get a global evaluation of segmentation algorithms. Lastly, a radar-style graph analysis is presented for easy multi-criteria interpretation.

SU‐F‐J‐140: Using Handheld Stereo Depth Cameras to Extend Medical Imaging for Radiation Therapy Planning

Purpose:A correct body contour is essential for the accuracy of dose calculation in radiation therapy. While modern medical imaging technologies provide highly accurate representations of body contours, there are times when a patient's anatomy cannot be fully captured or there is a lack of easy access to CT/MRI scanning. Recently, handheld cameras have emerged that are capable of performing three dimensional (3D) scans of patient surface anatomy. By combining 3D camera and medical imaging data, the patient's surface contour can be fully captured.Methods:A proof‐of‐concept system matches a patient surface model, created using a handheld stereo depth camera (DC), to the available areas of a body contour segmented from a CT scan. The matched surface contour is then converted to a DICOM structure and added to the CT dataset to provide additional contour information. In order to evaluate the system, a 3D model of a patient was created by segmenting the body contour with a treatment planning system (TPS) and fabricated with a 3D printer. A DC and associated software were used to create a 3D scan of the printed phantom. The surface created by the camera was then registered to a CT model that had been cropped to simulate missing scan data. The aligned surface was then imported into the TPS and compared with the originally segmented contour.Results:The RMS error for the alignment between the camera and cropped CT models was 2.26 mm. Mean distance between the aligned camera surface and ground truth model was −1.23 +/−2.47 mm. Maximum deviations were < 1 cm and occurred in areas of high concavity or where anatomy was close to the couch.Conclusion:The proof‐of‐concept study shows an accurate, easy and affordable method to extend medical imaging for radiation therapy planning using 3D cameras without additional radiation.Intel provided the camera hardware used in this study.

SuReSim: simulating localization microscopy experiments from ground truth models.

Super-resolution fluorescence microscopy has become a widely used tool in many areas of research. However, designing and validating super-resolution experiments to address a research question in a technically feasible and scientifically rigorous manner remains a fundamental challenge. We developed SuReSim, a software tool that simulates localization data of arbitrary three-dimensional structures represented by ground truth models, allowing users to systematically explore how changing experimental parameters can affect potential imaging outcomes.

Postoperative 3D spine reconstruction by navigating partitioning manifolds.

The postoperative evaluation of scoliosis patients undergoing corrective treatment is an important task to assess the strategy of the spinal surgery. Using accurate 3D geometric models of the patient's spine is essential to measure longitudinal changes in the patient's anatomy. On the other hand, reconstructing the spine in 3D from postoperative radiographs is a challenging problem due to the presence of instrumentation (metallic rods and screws) occluding vertebrae on the spine. This paper describes the reconstruction problem by searching for the optimal model within a manifold space of articulated spines learned from a training dataset of pathological cases who underwent surgery. The manifold structure is implemented based on a multilevel manifold ensemble to structure the data, incorporating connections between nodes within a single manifold, in addition to connections between different multilevel manifolds, representing subregions with similar characteristics. The reconstruction pipeline was evaluated on x-ray datasets from both preoperative patients and patients with spinal surgery. By comparing the method to ground-truth models, a 3D reconstruction accuracy of 2.24 ± 0.90 mm was obtained from 30 postoperative scoliotic patients, while handling patients with highly deformed spines. This paper illustrates how this manifold model can accurately identify similar spine models by navigating in the low-dimensional space, as well as computing nonlinear charts within local neighborhoods of the embedded space during the testing phase. This technique allows postoperative follow-ups of spinal surgery using personalized 3D spine models and assess surgical strategies for spinal deformities.

DeepBag: Recognizing Handbag Models

In this paper, we address the problem of branded handbag recognition. It is a challenging problem due to the non-rigid deformation, illumination changes, and inter-class similarity. We propose a novel framework based on deep convolutional neural network (CNN). Concretely, we propose a new CNN model, called feature selective joint classification - regression CNN (FSCR-CNN). Its advantages lie in two folds: 1) it alleviates the illumination changes by a feature selection strategy to focus on the color- nondiscriminative features in the network learning, and 2) rather than only targeting on the hard label (i.e., the handbag model), it also incorporates a soft label (i.e., a distribution measuring the similarity between the ground truth model and all the models to be trained) to construct the loss function for training CNN, which leads to a better classifier for handbags with large inter-class similarity. We evaluate the performance of our framework on a newly built branded handbag dataset. The results show that it performs favorably for recognizing handbags with 94.48% in accuracy. We also apply the proposed FSCR-CNN model in recognizing other fine-grained objects with state-of-the-art CNN architectures, which is able to achieve over 5% improvement in accuracy.

A robust likelihood function for 3D human pose tracking.

Recent works on 3D human pose tracking using unsupervised methods typically focus on improving the optimization framework to find a better maximum in the likelihood function (i.e., the tracker). In contrast, in this paper, we focus on improving the likelihood function, by making it more robust and less ambiguous, thus making the optimization task easier. In particular, we propose an exponential chamfer distance for model matching that is robust to small pose changes, and a part-based model that is better able to localize partially occluded and overlapping parts. Using a standard annealing particle filter and simple diffusion motion model, the proposed likelihood function obtains significantly lower error than other unsupervised tracking methods on the HumanEva dataset. Noting that the joint system of the tracker’s body model is different than the joint system of the motion capture ground-truth model, we propose a novel method for transforming between the two joint systems. Applying this bias correction, our part-based likelihood obtains results equivalent to state-of-the-art supervised tracking methods.