Peter Karasev

Anatomic and Morphological Evaluation of the Quadriceps Tendon Using 3-Dimensional Magnetic Resonance Imaging Reconstruction Applications for Anterior Cruciate Ligament Autograft Choice and Procurement

Background: The autograft of choice for anterior cruciate ligament (ACL) reconstruction remains controversial. The quadriceps tendon is the least utilized and least studied of the potential autograft options. Purpose: To determine if the quadriceps tendon has the anatomic characteristics to produce a graft whose length and volume are adequate, reproducible, and predictable when compared with the other commonly used autografts. Study Design: Cross-sectional study; Level of evidence, 3. Methods: Axial proton density magnetic resonance imaging (MRI) scans of 60 skeletally mature patients (30 male and 30 female) were evaluated. Volumetric analysis of 3-dimensional models of the patellar and quadriceps tendons was performed before and after the removal of a 10 mm–wide graft from both tendons. Length, thickness, and width measurements of the quadriceps tendon were made at predetermined locations. Anthropometric data were collected, and subgroup analysis, sex analysis, and linear regression were performed. Results: The mean percentage of volume remaining after removal of a 10 mm–wide graft from the patellar tendon was 56.6%, compared with 61.3% when harvesting an 80 mm–long graft of the same width from the quadriceps tendon. The intra-articular volume of the proposed quadriceps tendon graft was 87.5% greater than that of the patellar tendon graft. The mean length of the quadriceps tendon was 73.5 ± 12.3 mm in female patients and 81.1 ± 10.6 mm in male patients. These measurements were most highly correlated with patient height. The width of the quadriceps tendon decreased as one proceeded proximally from its insertion, and the thickness of the quadriceps tendon remained relatively constant. Conclusion: The quadriceps tendon has the anatomic characteristics to produce a graft whose length and volume are both reproducible and predictable, while yielding a graft with a significantly greater intra-articular volume than a patellar tendon graft with a similar width.

Fire and smoke detection in video with optimal mass transport based optical flow and neural networks

Detection of fire and smoke in video is of practical and theoretical interest. In this paper, we propose the use of optimal mass transport (OMT) optical flow as a low-dimensional descriptor of these complex processes. The detection process is posed as a supervised Bayesian classification problem with spatio-temporal neighborhoods of pixels;feature vectors are composed of OMT velocities and R,G,B color channels. The classifier is implemented as a single-hidden-layer neural network. Sample results show probability of pixels belonging to fire or smoke. In particular, the classifier successfully distinguishes between smoke and similarly colored white wall, as well as fire from a similarly colored background.

Optical Flow Estimation for Flame Detection in Videos

Computational vision-based flame detection has drawn significant attention in the past decade with camera surveillance systems becoming ubiquitous. Whereas many discriminating features, such as color, shape, texture, etc., have been employed in the literature, this paper proposes a set of motion features based on motion estimators. The key idea consists of exploiting the difference between the turbulent, fast, fire motion, and the structured, rigid motion of other objects. Since classical optical flow methods do not model the characteristics of fire motion (e.g., non-smoothness of motion, non-constancy of intensity), two optical flow methods are specifically designed for the fire detection task: optimal mass transport models fire with dynamic texture, while a data-driven optical flow scheme models saturated flames. Then, characteristic features related to the flow magnitudes and directions are computed from the flow fields to discriminate between fire and non-fire motion. The proposed features are tested on a large video database to demonstrate their practical usefulness. Moreover, a novel evaluation method is proposed by fire simulations that allow for a controlled environment to analyze parameter influences, such as flame saturation, spatial resolution, frame rate, and random noise.

Interactive Medical Image Segmentation Using PDE Control of Active Contours

Segmentation of injured or unusual anatomic structures in medical imagery is a problem that has continued to elude fully automated solutions. In this paper, the goal of easy-to-use and consistent interactive segmentation is transformed into a control synthesis problem. A nominal level set partial differential equation (PDE) is assumed to be given; this open-loop system achieves correct segmentation under ideal conditions, but does not agree with a human experts ideal boundary for real image data. Perturbing the state and dynamics of a level set PDE via the accumulated user input and an observer-like system leads to desirable closed-loop behavior. The input structure is designed such that a user can stabilize the boundary in some desired state without needing to understand any mathematical parameters. Effectiveness of the technique is illustrated with applications to the challenging segmentations of a patellar tendon in magnetic resonance and a shattered femur in computed tomography.

A sparsity-inducing optimization based algorithm for planar patches extraction from noisy point-cloud data

Currently, much of the manual labor needed to generate as-built building information models (BIMs) of existing facilities is spent converting raw point cloud data sets (PCDs) to BIM descriptions. Automating the PCD conversion process can drastically reduce the cost of generating as-built BIMs. Due to the widespread existence of planar structures in civil infrastructures, detecting and extracting planar patches from raw PCDs is a fundamental step in the conversion pipeline from PCDs to BIMs. However, existing methods cannot effectively address both automatically detecting and extracting planar patches from infrastructure PCDs. The existing methods cannot resolve the problem due to the large scale and model complexity of civil infrastructure, or due to the requirements of extra constraints or known information. To address the problem, this article presents a novel framework for automatically detecting and extracting planar patches from large-scale and noisy raw PCDs. The proposed method automatically detects planar structures, estimates the parametric plane models, and determines the boundaries of the planar patches. The first step recovers existing linear dependence relationships amongst points in the PCD by solving a group-sparsity inducing optimization problem. Next, a spectral clustering procedure based on the recovered linear dependence relationships segments the PCD. Then, for each segmented group, model parameters of the extracted planes are estimated via singular value decomposition (SVD) and maximum likelihood estimation sample consensus (MLESAC). Finally, the α-shape algorithm detects the boundaries of planar structures based on a projection of the data to the planar model. The proposed approach is evaluated comprehensively by experiments on two types of PCDs from real-world infrastructures, one captured directly by laser scanners and the other reconstructed from video using structure-from-motion techniques. To evaluate the performance comprehensively, five evaluation metrics are proposed which measure different aspects of performance. Experimental results reveal that the proposed method outperforms the existing methods, in the sense that the method automatically and accurately extracts planar patches from large-scaled raw PCDs without any extra constraints nor user assistance.

Obtaining a 35x Speedup in 2D Phase Unwrapping Using Commodity Graphics Processors

Graphics processing units (GPUs) are a powerful tool for numerical computation. The GPU architecture and computational model are uniquely designed for high-resolution high-speed grid-based calculations. This capability can be utilized to accelerate certain classes of compute-intensive radar signal processing algorithms. Characteristics of a problem well-suited for computation on a GPU include high levels of data parallelism, low control logic, uniform boundary conditions, and well-defined input and output. We describe the implementation of two-dimensional multigrid least-squares weighted phase unwrapping on a GPU and demonstrate a large speedup over C and MATLAB implementations. Details of the GPU computation are provided. Background information on the GPU architecture and its applicability to general-purpose computation is discussed.

Kernel-based high-dimensional histogram estimation for visual tracking

We propose an approach for non-rigid tracking that represents objects by their set of distribution parameters. Compared to joint histogram representations, a set of parameters such as mixed moments provides a significantly reduced size representation. The discriminating power is comparable to that of the corresponding full high- dimensional histogram yet at far less spatial and computational complexity. The proposed method is robust in the presence of noise and illumination changes, and provides a natural extension to the use of mixture models. Experiments demonstrate that the proposed method outperforms both full color mean-shift and global covariance searches.

Interactive MRI segmentation with controlled active vision

Partitioning Magnetic-Resonance-Imaging (MRI) data into salient anatomic structures is a problem in medical imaging that has continued to elude fully automated solutions. Implicit functions are a common way to model the boundaries between structures and are amenable to control-theoretic methods. In this paper, the goal of enabling a human to obtain accurate segmentations in a short amount of time and with little effort is transformed into a control synthesis problem. Perturbing the state and dynamics of an implicit functions driving partial differential equation via the accumulated user inputs and an observer-like system leads to desirable closed-loop behavior. Using a Lyapunov control design, a balance is established between the influence of a data-driven gradient flow and the humans input over time. Automatic segmentation is thus smoothly coupled with interactivity. An application of the mathematical methods to orthopedic segmentation is shown, demonstrating the expected transient and steady state behavior of the implicit segmentation function and auxiliary observer.

Depth invariant visual servoing

This paper studies the problem of achieving consistent performance for visual servoing. Given the nonlinearities introduced by the camera projection equations in monocular visual servoing systems, many control algorithms experience non-uniform performance bounds. The variable performance bounds arise from depth dependence in the error rates. In order to guarantee depth invariant performance bounds, the depth nonlinearity must be cancelled, however estimating distance along the optical axis is problematic when faced with an object with unknown geometry. By tracking a planar visual feature on a given target, and measuring the area of the planar feature, a distance invariant input to state stable visual servoing controller is derived. Two approaches are given for achieving the visual tracking. Both of these approaches avoid the need to maintain long-term tracks of individual feature points. Realistic image uncertainty is captured in experimental tests that control the camera motion in a 3D renderer using the observed image data for feedback.

A video analytics framework for amorphous and unstructured anomaly detection

Video surveillance systems are often used to detect anomalies: rare events which demand a human response, such as a fire breaking out. Automated detection algorithms enable vastly more video data to be processed than would be possible otherwise. This note presents a video analytics framework for the detection of amorphous and unstructured anomalies such as fire, targets in deep turbulence, or objects behind a smoke-screen. Our approach uses an off-line supervised training phase together with an on-line Bayesian procedure: we form a prior, compute a likelihood function, and then update the posterior estimate. The prior consists of candidate image-regions generated by a weak classifier. Likelihood of a candidate region containing an object of interest at each time step is computed from the photometric observations coupled with an optimal-mass-transport optical-flow field. The posterior is sequentially updated by tracking image regions over time and space using active contours thus extracting samples from a properly aligned batch of images. The general theory is applied to the video-fire-detection problem with excellent detection performance across substantially varying scenarios which are not used for training.