Wolfgang Härer

Event-synchronous cancellation of the heart interference in biomedical signals

A two-pass adaptive filtering algorithm is proposed for cancellation of recurrent interferences such as the heart interference in biomedical signals. In the first pass, an average waveform in one period of the interference is estimated by event-synchronous (QRS-synchronous) averaging of the corrupted signal. In a second pass, an adaptive Schur recursive least squares (RLS) lattice filter is used to cancel the interference by using the event synchronously repeated estimated average waveform of the interference as an artificial reference signal. One key feature of this approach is that the ECG is only used for QRS synchronization and not directly as a reference signal for adaptive filtering. Thus the proposed algorithm can be applied to interference problems where ECG and true interference are almost synchronous but show considerably different waveforms. This is usually the case with the heart interference in biomedical signals. Both off-line and real-time implementations of the event synchronous interference canceller are described. The method is applied to the cancellation of the heart interference in magnetoencephalogram (MEG) signals and to the effective isolation of ventricular extrasystoles (VES) in magnetocardiogram (MCG) signals. Experimental results are shown. The new method typically attenuates the amplitudes of R-wave and T-wave interference components by an amplitude factor of 30 without influencing the MEG events of interest.<<ETX>>

Pattern recognition in biomagnetic signals by spatio-temporal correlation and application to the localisation of propagating neuronal activity

A combined correlation analysis of temporal and spatial patterns in a biomagnetic multichannel recording is proposed to extract patterns of transient, randomly repetitive physiological events from biomagnetic data sets with a low signal-to-noise ratio. This method is especially sensitive in the differentiation of signals from different source propagation pathways. The recognition and subsequent averaging of spike events in the MEG of an epileptic patient by this method is shown to provide an essential improvement in the signal-to-noise ratio. The source localisation from the averaged data showed electrical activity which propagated from the primary epileptogenic focus to distant parts of the brain.

A Novel Approach for Filtered Backprojection in Tomosynthesis Based on Filter Kernels Determined by Iterative Reconstruction Techniques

Breast tomosynthesis is a new 3D imaging technique providing 3D slices of the breast. The computation of 3D slices from a set of 2D projections is performed with a reconstruction algorithm --- either a filtered backprojection (FBP) or an algebraic iterative reconstruction method. Both approaches yield different image characteristics of the reconstructed object. The algebraic method has the main disadvantage of a very long processing time. We experimentally developed a set of filter kernels for FBP, determined by iterative reconstruction providing similar image characteristics and quality as an algebraic reconstruction. Additionally we showed that it is possible to approximate these kernels by a polynomial function of order 4. Using clinical data sets we demonstrate how the image quality and the image impression can be varied by using different reconstruction methods.

Spatial Differences of the Duration of Ventricular Late Fields in the Signal‐Averaged Magnetocardiogram in Patients with Ventricular Late Potentials

Magnetocardiography (MCG) allows one to noninvasively localize cardiac electrical activity in three dimensions. It was the purpose of this study to obtain information about the spatial variations of signal‐averoged ventricular Jate magnetic fields recorded by a biomagnetic multichannel system. Biomagnetic signals of 170–600 heart cycles obtained hy the 37‐channel system KRENIKONR (Siemens Medical Engineering Group) were simultaneously averaged in all channels. The absolute values of the filtered signals (digital, bidirectional, four‐pole butterworth, bandpass filter [3‐dB range, 40–250 Hz]) were calculated in each channel. The noise level was determined within the TP segment. The onset of the terminal low amplitude signals (TLAS) was defined when the signals became lower than 1/23 of Rmaxof the QRS complex for the channel with the largest filtered QRS complex after filtering. The TLAS ended when the signal was lower than twice the standard deviation (2 sigma) above the mean noise level. Ventricular late fields were defined as present when the TLAS had a duration of more than 39 msec. In this study, five patients with ventricular late potentials (four with sustained ventricular tachycardia) and three healthy individuals were examined. Ventricular late fields were detected in the patient group in 2–15 MCG channels with a mean length of 49.6 msec (43–60 msec). The spatial distribution of the ventricular late fields was consistently found to exhibit maximum duration in a certain area. In the normal subjects no ventricular late fields were detected. Thus, MCG is able to detect ventricular late fields and their spatial variations. In addition to the information obtained hy signal averaging from the surface ECG, averaging of biomagnetic signals with a multichannel device can reveal spatial inhomogeneity of delayed myocardial excitation.

Clinical magnetocardiography: Experience with a biomagnetic multichannel system

The magnetic fields caused by the human hearts electrical activity were coherently recorded with a biomagnetic multichannel system (KRENIKON®) during 1 to 10 minutes in 49 patients. 31 to 37 magnetic channels were recorded simultaneously with the ECG and respiration.Comparison of a magnetic index and the Sokolow-Lyon index to echocardiographic findings in the quantification of left ventricular hypertrophy demonstrated the superiority of the magnetocardiogram (MCG) as compared to the ECG. The magnetocardiographic investigation of patients with WPW-Syndrome, ventricular extrasystoles, ventricular tachycardia, and paced ventricular beats demonstrated that multichannel magnetocardiography permits the non-invasive three dimensional localization of arrhythmogenic tissue with high spatial accuracy.

Iterative extended field of view reconstruction

Incomplete data due to the object extent beyond the scanning field of view (SFOV) is a common problem in computed tomography. In these cases, there are parts of the object to be reconstructed for which only incomplete projections of less than 180o are available. Applying iterative algorithms like algebraic reconstruction technique (ART) or simultaneous algebraic reconstruction Technique (SART) onto the problem of truncated projections can not produce a satisfying solution unless special constraints are used. To regularize the reconstruction algorithm, we extend iterative reconstruction algorithms by introducing information regarding the statistics of the attenuation values of the reconstructed object in terms of the log likelihood function of attenuation values. This information can be taken from the regions of the image still inside the SFOV but close to the region where the object exceeds the SFOV. The information can be utilized in an algebraic reconstruction method by adding a constraint term to the cost function that shall be minimized. Experiments show that for not severely truncated projections, as they are common for CT applications, including this information yields good estimates about the object.

Constriction of cone-beam artifacts by the Z-smart reconstruction method

We introduce a novel approach for 3D image reconstruction from CB data acquired along a full circular trajectory. Our approach, which we refer to as the Z-smart reconstruction method, follows the scheme of ID filtering and 3D backprojection, while providing flexibility in the choice of the filtering directions. This flexibility allows us to modify the appearance of CB artifacts in the reconstruction results, so that, for certain imaging tasks, the Z-smart method can yield image quality that is significantly superior to that achievable with most other reconstruction algorithms for the full-scan circular trajectory. In particular, when imaging objects that have strong but localized heterogeneities in the axial direction, the Z-smart method can outperform both, the popular filtered-backprojection approach of Feldkamp et ah, and also ART, i.e. a fully 3D iterative reconstruction method. We provide a numerical evaluation of our algorithm using simulated CB data of an analytically defined tube phantom, with and without added noise.

Signal processing for the biomagnetic reconstruction of time-variant sources

Abstract not available.

The Neural Impulse Response Filter

The neural network filter architecture presented in this paper is well suited for a restricted class of nonlinear adaptive filter applications. Our filter can model systems with long time responses and it is able to learn to take into account additional parameters that influence the filter response nonlinearly. Our filter was successfully used in a biomedical application for the removal of the cardiac interference from magnetoencephalographic (MEG) data.

Isolation of ventricular extrasystrolic signals from the preceding heart activity by adaptive filtering

Noninvasive 3-dimensional localization of the origin of ventricular extrasystoles (VES) is possible from multichannel MEG data after separation of VES onset from the preceding repolarization activity. An adaptive filter algorithm for signal separation is suggested. Compared to conventional digital subtraction, the adaptive filter saves interactive computer time, and first experiences with multichannel MCG data indicate that the result is more accurate.