Publikace

We address the problem of estimating the uncertainty of optical flow algorithm results. Our method estimates the error magnitude at all points in the image. It can be used as a confidence measure. It is based on bootstrap resampling, which is a computational statistical inference technique based on repeating the optical flow calculation several times for different randomly chosen subsets of pixel contributions. As few as 10 repetitions are enough to obtain useful estimates of geometrical and angular errors. We use the combined local global optical flow method (CLG) which generalizes both Lucas-Kanade and Horn-Schunck type methods. However, the bootstrap method is very general and can be applied to almost any optical flow algorithm that can be formulated as a minimization problem. We show experimentally on synthetic as well as real video sequences with known ground truth that the bootstrap method performs better than all other confidence measures tested.

We explore the benefit of a spatio-temporal approach for deformable registration of respiratory-correlated imaging of the thorax. Our goal is obtain a more compact, restrictive parametrization of the deformation model in an attempt to reduce sen sitivity to artifacts and noise.

Ultrasound guidance is used for many surgical interventions such as biopsy and electrode insertion. We present a~method to localize a thin surgical tool such as a biopsy needle or a micro-electrode in a 3D ultrasound image. The proposed method starts with thresholding and model fitting using RANSAC for robust localization of the axis. Subsequent local optimization refines its position. Two different tool image models are presented, one simple and fast, the second using learned a priori information on the tool's voxel intensities and the background. Finally, the tip of the tool is localized by finding an intensity drop along the axis. The simulation study shows that our algorithm can localize the tool at nearly real-time speed, even using a MATLAB implementation, with accuracy better than 1mm. In an experimental comparison to several alternative localization methods, our method appears to be the fastest and the most robust one.

We present a novel algorithm for the registration of 2D image sequences that combines the principles of multiresolution B-spline based elastic registration and those of bidirectional consistent registration. In our method, consecutive triples of images are iteratively registered to gradually extend through the set of images the information of the whole sequence. The intermediate results are reused as initialization of the corresponding next triple registration. We interpolate the images and model the deformation fields using B-spline multiresolution pyramids.

We address the problem of estimating the uncertainty of pixel based image registration algorithms, given just the two images to be registered, for cases when no ground truth data is available. Our novel method uses bootstrap resampling. It is very general, applicable to almost any registration method based on minimizing a pixel-based similarity criterion; we demonstrate it using the SSD, SAD, correlation, and mutual information criteria. We show experimentally that the bootstrap method provides better estimates of the registration accuracy than the state-of-the-art Cramer-Rao bound method. Additionally, we evaluate also a fast registration accuracy estimation (FRAE) method which is based on quadratic sensitivity analysis ideas and has a negligible computational overhead. FRAE mostly works better than the Cramer-Rao bound method but is outperformed by the bootstrap method.

In surgical practice small metallic instruments are frequently used to perform various tasks inside human body. We address the problem of their accurate localization in the tissue. Recent experiments using medical ultrasound have shown that this modality is suitable for real-time visualization of anatomical structures as well as the position of surgical instruments. We propose an image processing algorithm that permits to automatically estimate the position of a line-segment shaped object. This method was applied to the localization of a thin metallic electrode in biological tissue. We show that the electrode axis can be found through maximizing the Parallel Integral Projection transform which is a form of the Radon transform. To accelerate this step hierarchical mesh-grid algorithm is implemented. Once the axis position is known, localization of the electrode tip is performed. The method was tested on simulated images, on ultrasound images of a tissue mimicking phantom containing a me

2-D Locally Regularized Tissue Strain Estimation From Radio-Frequency Ultrasound Images: Theoretical Developments and Results on Experimental Data

The Laplace-Cauchy problem of propagating Dirichlet and Neumann data from a portion to the rest of the boundary is an ill-posed inverse problem. Many regularizing algorithms have been recently proposed, in order to stabilize the solution with respect to noisy or incomplete data. Our main application is in electro-encephalography (EEG) where potential measurements available at part of the scalp are used to reconstruct the potential and the current on the inner skull surface. This problem, known as cortical mapping, and other applications - in fields such as nondestructive testing, or biomedical engineering - require to solve the problem in realistic, three-dimensional geometry. The goal of this article is to present a new boundary element based method for solving the Laplace-Cauchy problem in three dimensions, in a multilayer geometry. We validate the method experimentally on simulated data.

Parallel MRI (pMRI) is a way to increase the speed of the MRI acquisition by com bining data obtained simultaneously from several receiver coils with distinct sp atial sensitivities. We propose an algorithm th at uses B-spline functions to approximate the reconstruction map which reduces t he number of parameters to estimate and makes the reconstruction faster and less sensitive to noise. The proposed method is tested on both phantom and in vivo images. The results ar e compared with commercial implementation of GRAPPA and SENSE algorithms in term s of time complexity and quality of the reconstruction.

Generalized topology extension for the symmetric MEG/EEG boundary element method.