gms | German Medical Science

GMDS 2013: 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

01. - 05.09.2013, Lübeck

Integrated Processing and Visualization Framework for Signal and Signal-to-Image Data

Meeting Abstract

  • Stephan Jonas - Uniklinik RWTH Aachen, Aachen, DE
  • Ali Hadizadeh - Uniklinik RWTH Aachen, Aachen, DE
  • Julian Gehrenkemper - Uniklinik RWTH Aachen, Aachen, DE
  • Nikolaus Marx - Uniklinik RWTH Aachen, Aachen, DE
  • Thomas Deserno - Uniklinik RWTH Aachen, Aachen, DE

GMDS 2013. 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Lübeck, 01.-05.09.2013. Düsseldorf: German Medical Science GMS Publishing House; 2013. DocAbstr.336

doi: 10.3205/13gmds267, urn:nbn:de:0183-13gmds2678

Veröffentlicht: 27. August 2013

© 2013 Jonas et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.de). Er darf vervielfältigt, verbreitet und öffentlich zugänglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.


Gliederung

Text

Introduction: Long-term electrocardiography (ECG) is an important tool for the diagnosis of cardiac-related diseases. Interpretation of long-term ECG data is usually reduced to the detection of short periods of abnormal beat morphology. However, the general “fitness” of the heart might be of higher interest. Furthermore, scientific long-term ECG studies produce a lot of data (12 channels, 1000 Hz, sometimes multiple days), which cannot be evaluated manually and requires special software for processing and visualization. While signal processing has been used to quantify and classify ECG signals exclusively so far, our goal is to apply recent developments in image processing to ECG data. Therefore, a new framework is proposed combining signal and image processing techniques on ECG data and additional signals from other sources.

Methods: The proposed Matlab-based (http://www.mathworks.de/) software consists of an intuitive graphical user interface (GUI) for visualization that allows users to choose what ECG channels with which combination of filters or other processing steps to be displayed. In addition, ECG and other signal data (e.g., blood glucose, respiratory movement) is visualized as image data in multiple ways. For example based on the signals self-similarity, cycles and/or channels are considered as lines of columns of an image, and the signal’s amplitude is coded as gray scale. Due to the large amount of data, all processing is done only on demand for the displayed portion of the signal by a real-time processing engine. Additionally, the software supports partial loading. This means that only the part of the data is read from the recording file into memory that is necessary for the current calculations and visualizations. This allows for maximum flexibility at a minimum of computational cost. The framework was built in a modular manner for easy enhancement and further development of functionality. For example, new filters such as smoothing; segmentation, partitioning, and clustering; as well as special event detection can be integrated by simply registering the corresponding class in a directory.

Results: The software is currently used for visualization and analysis of long-term ECG data (7days, 12channels) in a multicenter clinical trial on dialysis patients. Easily added to the display by checkboxes, a 12 lead ECG may be represented as an image with 12 pixels height, or pivoting single beats in a single channel. Pivoting normalizes the beat-length and represents it column wise in the image. Using this technique, timing-changes of different parts of the beat are revealed. These visualizations are superimposed to selected channel plots, and stretched according to the selected window. Visualization artifacts due to insufficient resolution are clearly marked to the user. The framework allows easy integration of 1D and 2D filters for signal and image processing, respectively. The data will be analyzed for differences in the cardiac cycle depending on the time since, or during dialysis.

Discussion and Conclusion: Combining medical signal and image processing techniques provides impact to understanding long-term ECG. Our Framework is fast, reliable and easy to use. Applying image-based texture analysis to medical signals will provide further insights in physiology and medicine.