The novel high-density mapping system RhythmiaTM Medical (Boston Scientific, Marlborough, USA) allows a fast and automatic acquisition of intracardiac electrograms (EGMs). For recording the ORION mini-basket catheter is used. Due to the small electrode surface, the spatial averaging is smaller than with other commonly used mapping catheters. This results in a higher quality of unipolar signals. However, these are still corrupted by noise such as high frequency interference. Within this project, methods were developed and benchmarked that can be applied to detect and remove these undesired components. An algorithm was implemented to detect and eliminate artificial peaks in the spectrum of the EGM. The filtered signals showed improved quality in time domain. The performance of the spectral peak detection resulted in a median sensitivity of 92.1% and in a median positive predictive value of 91.9%.
Student Theses (2)
S. Huck. Implementation and Evaluation of Signal Preprocessing Techniques for Novel Intracardiac High-Density Mapping. Institute of Biomedical Engineering, Karlsruhe Institute of Technology (KIT). Bachelorarbeit. 2016
The novel high-density mapping system RhythmiaTM Medical from Boston Scientific allows a fast and automatic acquisition of endocardial signals. For recording the ORION mini-basket catheter is used which consists of 64 electrodes. Due to the small electrode surface of 0.4mm2, the spatial averaging is smaller than in other commonly used mapping catheters. This results in a higher quality of unipolar signals. Furthermore, the big number of electrodes and the small inter-electrode spacing lead to a high resolution of local activation time maps.An export of the raw data, which were recorded during mapping, was possible. In this thesis, preprocessing techniques for the unipolar signals were implemented and evaluated. Furthermore, local activation time maps were generated. Two preprocessing techniques were implemented with the requirement to identify and eliminate artificial peaks in a spectrum. The peak detection algorithm automatically identified artificial peaks by analysis of their heights in a spectrum. The fundamental frequency calculation was based on the automatic peak detection, but additionally considered if peaks were harmonics of a fundamental frequency. For suppression a Gaussian Notch filter was used. The performance of each identification technique was evaluated using respectively a data set containing manual annotations of 28 spectra. The peak detection resulted in a median sensitivity of 92.1% and in a median positive predictive value of 91.9%. The performance of the fundamental frequency calculation method yielded a median sensitivity and positive predictive value of 100%. Furthermore, the quality of the unipolar electrograms were compared with electrograms derived from the EnSite Velocity mapping system, recorded using the Constellation big-basket catheter. Therefore, the signal-to-noise ratio was calculated. Electrograms which were recorded using the ORION catheter demonstrated a higher signal quality. Three local activation time maps were generated. These maps contained 1403 - 1756 collected measuring points. The visual inspection confirmed a qualitative agreement. The comparison in phase space showed good quantitative agreement with resulting root-mean-square errors of 12.0 - 19.6 ms.
S. Huck. 3D Contactless Respiration Measurement. Institute of Biomedical Engineering, Karlsruhe Institute of Technology (KIT); Philips Medizin Systeme Böblingen GmbH. Masterarbeit. 2018
An increased respiration rate is an indicator of several diseases, such as sepsis. Even nowadays, the common way to measure the respiration rate on the general ward, is the manual counting of breaths within a minute. These measurements are inaccurate and thus unreliable. A contactless respiration measurement technique at the general ward could solve this problem and could contribute to the early detection of diseases. In this thesis, three feature extraction methods were implemented and evaluated which measured the respiration rate from 3D data of a Time of Flight camera. The spectral approach determined the respiration rate by analyzing the pixel signals in the frequency domain. In the second approach, the principal component analysis was applied to the pixel signals with the aim of extracting the respiration signal from the noisy data. The third approach computed volume changes over time. Additionally, a movement detection was implemented to avoid failures. The performance of each approach was evaluated by comparing the results to a reference capnography signal. 85 recordings from six subjects were analyzed. The spectral and the PCA approach achieved positive predictive values of more than 99 %. The volume approach showed, in contrast to the others, dependencies on the subject’s pose. The work in this thesis showed that a respiration signal can be obtained from 3D data. Two out of three of the proposed methods can be used for further development to enable the usage in a medical device.