Efficient Visual Analysis of Dynamic Medical Image Data

Spatial and temporal resolution of tomographic medical image data (CT, MRI, etc.) being acquired in medical diagnostics and clinical studies has increased substantially and will increase further. Particularly for dynamic image data, the evaluation software does not sufficiently exploit the rich information. A framework shall be developed that combines image interpretation techniques with visual analysis of 4D dynamic medical image data. Perfusion data is an important and representative example for dynamic medical image data. These data are acquired, e.g., in ischemic stroke, cardiac, and tumor diagnosis. A multi-dimensional space of perfusion parameters needs to be explored to perform a reliable diagnosis. For the first time, adaptive model-based segmentation techniques will be developed to delineate regions of interest in these 4D data sets. Such a visually supported analysis has several advantages:

• Implicit training lets the user adapt the tool for specializing it to selected problems in perfusion analysis.
• An efficient general solution is provided which might be adapted according to the specific imaging device, the imaging sequence, or the type of contrast agent administration.
• Interpretation tools can be extended to similar analysis problems, e.g. fMRI data evaluation.

Techniques from cluster analysis, dimension reduction and image segmentation will be used to extract features for visualization. 3D visualization techniques will be refined and adapted to the peculiarities of high resolution perfusion data. Data exploration will support researching physicians and medical physicists to assess the influence on image acquisition parameters on the expressiveness of perfusion parameters and combinations thereof.

Das Projekt ist Teil des DFG-SPP (Scalable Visual Analytics: Interaktive visuelle Analysesysteme für komplexe Informationswelten).

http://infovis.uni-konstanz.de/spp/index.php?lang=de

Spatial and temporal resolution of tomographic medical image data (CT, MRI, etc.) being acquired in medical diagnostics and clinical studies has increased substantially and will increase further. Particularly for dynamic image data, the evaluation software does not sufficiently exploit the rich information. A framework shall be developed that combines image interpretation techniques with visual analysis of 4D dynamic medical image data. Perfusion data is an important and representative example for dynamic medical image data. These data are acquired, e.g., in ischemic stroke, cardiac, and tumor diagnosis. A multi-dimensional space of perfusion parameters needs to be explored to perform a reliable diagnosis. For the first time, adaptive model-based segmentation techniques will be developed to delineate regions of interest in these 4D data sets. Such a visually supported analysis has several advantages:

• Implicit training lets the user adapt the tool for specializing it to selected problems in perfusion analysis.
• An efficient general solution is provided which might be adapted according to the specific imaging device, the imaging sequence, or the type of contrast agent administration.
• Interpretation tools can be extended to similar analysis problems, e.g. fMRI data evaluation.

Techniques from cluster analysis, dimension reduction and image segmentation will be used to extract features for visualization. 3D visualization techniques will be refined and adapted to the peculiarities of high resolution perfusion data. Data exploration will support researching physicians and medical physicists to assess the influence on image acquisition parameters on the expressiveness of perfusion parameters and combinations thereof.

This Project is part of the DFG-SPP Scalable Visual Analytics.

http://infovis.uni-konstanz.de/spp/index.php?lang=de