Jyrki Lötjönen

Model extraction from magnetic resonance volume data using the deformable pyramid

A general framework for automatic model extraction from magnetic resonance (MR) images is described. The framework is based on a two-stage algorithm. In the first stage, a geometrical and topological multiresolution prior model is constructed. It is based on a pyramid of graphs. In the second stage, a matching algorithm is described. This algorithm is used to deform the prior pyramid in a constrained manner. The topological and the main geometrical properties of the model are preserved, and at the same time, the model adapts itself to the input data. We show that it performs a fast and robust model extraction from image data containing unstructured information and noise. The efficiency of the deformable pyramid is illustrated on a synthetic image. Several examples of the method applied to MR volumes are also represented.

A 3-D model-based registration approach for the PET, MR and MCG cardiac data fusion

In this paper, a new approach is presented for the assessment of a 3-D anatomical and functional model of the heart including structural information from magnetic resonance imaging (MRI) and functional information from positron emission tomography (PET) and magnetocardiography (MCG). The method uses model-based co-registration of MR and PET images and marker-based registration for MRI and MCG. Model-based segmentation of MR anatomical images results in an individualized 3-D biventricular model of the heart including functional parameters from PET and MCG in an easily interpretable 3-D form.

Methods of Artificial Enlargement of the Training Set for Statistical Shape Models

Due to the small size of training sets, statistical shape models often over-constrain the deformation in medical image segmentation. Hence, artificial enlargement of the training set has been proposed as a solution for the problem to increase the flexibility of the models. In this paper, different methods were evaluated to artificially enlarge a training set. Furthermore, the objectives were to study the effects of the size of the training set, to estimate the optimal number of deformation modes, to study the effects of different error sources, and to compare different deformation methods. The study was performed for a cardiac shape model consisting of ventricles, atria, and epicardium, and built from magnetic resonance (MR) volume images of 25 subjects. Both shape modeling and image segmentation accuracies were studied. The objectives were reached by utilizing different training sets and datasets, and two deformation methods. The evaluation proved that artificial enlargement of the training set improves both the modeling and segmentation accuracy. All but one enlargement techniques gave statistically significantly (p < 0.05) better segmentation results than the standard method without enlargement. The two best enlargement techniques were the nonrigid movement technique and the technique that combines principal component analysis (PCA) and finite element model (FEM). The optimal number of deformation modes was found to be near 100 modes in our application. The active shape model segmentation gave better segmentation accuracy than the one based on the simulated annealing optimization of the model weights.

Reconstruction of 3-D geometry using 2-D profiles and a geometric prior model

A method has been developed to reconstruct three-dimensional (3-D) surfaces from two-dimensional (2-D) projection data. It is used to produce individualized boundary element models, consisting of thorax and lung surfaces, for electro- and magnetocardiographic inverse problems. Two orthogonal projections are utilized. A geometrical prior model, built using segmented magnetic resonance images, is deformed according to profiles segmented from projection images. In the authors method, virtual X-ray images of the prior model are first constructed by simulating real X-ray imaging. The 2-D profiles of the model are segmented from the projections and elastically matched with the profiles segmented from patient data. The displacement vectors produced by the elastic 2-D matching are back projected onto the 3-D surface of the prior model. Finally, the model is deformed, using the back-projected vectors. Two different deformation methods are proposed. The accuracy of the method is validated by a simulation. The average reconstruction error of a thorax and lungs was 1.22 voxels, corresponding to about 5 mm.

A triangulation method of an arbitrary point set for biomagnetic problems

A new triangulation method has been developed for extracting isosurface from volume data. The nodes for triangulation can be selected arbitrarily from the surface of the object of interest. The Voronoi polygons for nodes are searched on the surface and triangulation is accomplished by connecting the neighboring Voronoi areas. The method is basically Delaunay triangulation using geodesic distances instead of Euclidean ones. In areas where the curvature of the surface is low, the Delaunay criteria are fulfilled. When the curvature is high, the geometry of the object is described more accurately than in Euclidean Delaunay methods. Since geodesic distances are utilized, i.e., the surface information is used in triangulation, the topology of the object can be preserved more easily than in the Euclidean cases. Our fully automatic method has been developed for boundary element modeling and it has been successfully applied in magnetocardiographic and electrocardiographic forward and inverse studies. However, the method can be utilized in any triangulation problem if the surface description is provided.

Correction of Movement Artifacts from 4-D Cardiac Short- and Long-Axis MR Data

Typically a cardiac MR cine series consists of images over several time points but only from one spatial location. The volumetric information is obtained by combining 2-D slices from different image series. If a patient moves during an MR imaging session, the slices from different image series shift relative to each other, and the 3-D volume reconstructed does not represent the real geometry. In this study, an algorithm was developed to correct movement artifacts simultaneously from short- and long-axis MR cine series. The performance of the algorithm was evaluated by calculating the accuracy of the method against simulated movements imposed on real data, and by visually inspecting the results with real patient images. In both cases, the algorithm reduced significantly movement artifacts.

Elastic Matching Using a Deformation Sphere

A novel method is proposed for elastic matching of two data volumes. A combination of mutual information, gradient information and smoothness of transformation is used to guide the deformation of another of the volumes. The deformation is accomplished in a multiresolution process by spheres containing a vector field. Position and radius of the spheres are varied. The feasibility of the method is demonstrated in two cases: matching inter-patient MR images of the head and intra-patient cardiac MR and PET images.

Bioelectromagnetic localization of a pacing catheter in the heart

The accuracy of localizing source currents within the human heart by non-invasive magneto- and electrocardiographic methods was investigated in 10 patients. A non-magnetic stimulation catheter inside the heart served as a reference current source. Biplane fluoroscopic imaging with lead ball markers was used to record the catheter position. Simultaneous multichannel magnetocardiographic (MCG) and body surface potential mapping (BSPM) recordings were performed during catheter pacing. Equivalent current dipole localizations were computed from MCG and BSPM data, employing standard and patient-specific boundary element torso models. Using individual models with the lungs included, the average MCG localization error was 7+/-3 mm, whereas the average BSPM localization error was 25+/-4 mm. In the simplified case of a single homogeneous standard torso model, an average error of 9+/-3 mm was obtained from MCG recordings. The MCG localization accuracies obtained in this study imply that the capability of multichannel MCG to locate dipolar sources is sufficient for clinical purposes, even without constructing individual torso models from x-ray or from magnetic resonance images.

Early familial dilated cardiomyopathy: identification with determination of disease state parameter from cine MR image data.

PURPOSEnTo characterize early changes in cardiac anatomy and function for lamin A/C gene (LMNA) mutation carriers by using magnetic resonance (MR) imaging and to develop tools to analyze and visualize the findings.nnnMATERIALS AND METHODSnThe ethical review board of the institution approved the study, and informed written consent was obtained. The patient group consisted of 12 subjects, seven women (mean age, 36 years; age range, 18-54 years) and five men (mean age, 28 years; age range, 18-39 years) of Finnish origin, who were each heterozygotes with one LMNA mutation that may cause familial dilated cardiomyopathy (DCM). All the subjects were judged to be healthy with transthoracic echocardiography. The control group consisted of 14 healthy subjects, 11 women (mean age, 41 years; range, 23-54 years) and three men (mean age, 45 years; range, 34-57 years), of Finnish origin. Cine steady state free precession MR imaging was performed with a 1.5-T system. The volumes, wall thickness, and wall motion of both left ventricle (LV) and right ventricle were assessed. A method combining multiple MR image parameters was used to generate a global cardiac function index, the disease state parameter (DSP). A visual fingerprint was generated to assess the severity of familial DCM.nnnRESULTSnThe mean DSP of the patient group (0.69 +/- 0.15 [standard deviation]) was significantly higher than that of the control group (0.32 +/- 0.13) (P = .00002). One subject had an enlarged LV.nnnCONCLUSIONnSubclinical familial DCM was identified by determination of the DSP with MR imaging, and this method might be used to recognize familial DCM at an early stage.

Nonfluoroscopic localization of an amagnetic stimulation catheter by multichannel magnetocardiography.

This study was performed to: (1) evaluate the accuracy of noninvasive magnetocardiographic (MCG) localization of an amagnetic stimulation catheter; (2) validate the feasibility of this multipurpose catheter; and (3) study the characteristics of cardiac evoked fields. A stimulation catheter specially designed to produce no magnetic disturbances was inserted into the heart of five patients after routine electrophysiological studies. The catheter position was documented on biplane cine x‐ray images. MCG signals were then recorded in a magnetically shielded room during cardiac pacing. Noninvasive localization of the catheters tip and stimulated depolarization was computed from measured MCG data using a moving equivalent current‐dipole source in patient‐specific boundary element torso models. In all five patients, the MCG localizations were anatomically in good agreement with the catheter positions defined from the x‐ray images. The mean distance between the position of the tip of the catheter defined from x‐ray fluoroscopy and the MCG localization was 11 ± 4 mm. The mean three‐dimensional difference between the MCG localization at the peak stimulus and the MCG localization, during the ventricular evoked response about 3 ms later, was 4 ± 1 mm calculated from signal‐averaged data. The 95% confidence interval of beat‐to‐beat localization of the tip of the stimulation catheter from ten consecutive beats in the patients was 4 ± 2 mm. The propagation velocity of the equivalent current dipole between 5 and 10 ms after the peak stimulus was 0.9 ± 0.2 m/s. The results show that the use of the amagnetic catheter is technically feasible and reliable in clinical studies. The accurate three‐dimensional localization of this multipurpose catheter by multichannel MCG suggests that the method could be developed toward a useful clinical tool during electrophysiological studies.