Visual analysis of cerebral perfusion data – four interactive approaches and a comparison
Abstract
Cerebral perfusion data are acquired to characterize the regional blood supply of brain tissue. One of their major diagnostic applications is ischemicstroke assessment. We present a comparison of four interactive approaches to analyzingcerebral perfusion data from ischemic stroke patients which are based on (1) concentration-time curves (CTC) derived from the original data, (2) parameters describing the CTC shape, (3) enhancement trends computed in a statistical analysis, and (4) semi-quantitative perfusion parameters derived via parametric modelling and deconvolution. The comparison is carried out with regard to the involved data pre-processing, the complexity of the interactive analysis and the resulting tissue selections. It is supported by a visual analysis framework that integrates the different approaches. The rich information content in time-dependent 3D perfusion data is both an opportunity for improved diagnosis and a challenge how to optimize the assessment of such rich data. With our comparison we contribute to a discussion between data-near and model-near assessment strategies and their respective opportunities.
S. Oeltze, B. Preim, H. Hauser, J. RØ. rvik, and A. Lundervold, "Visual analysis of cerebral perfusion data – four interactive approaches and a comparison," in Proceedings of the 6th Intern. Symp. on Image and Signal Processing and Analysis (ISPA 2009), 2009, p. 582–589.
[BibTeX]
Cerebral perfusion data are acquired to characterize the regional blood supply of brain tissue. One of their major diagnostic applications is ischemicstroke assessment. We present a comparison of four interactive approaches to analyzingcerebral perfusion data from ischemic stroke patients which are based on (1) concentration-time curves (CTC) derived from the original data, (2) parameters describing the CTC shape, (3) enhancement trends computed in a statistical analysis, and (4) semi-quantitative perfusion parameters derived via parametric modelling and deconvolution. The comparison is carried out with regard to the involved data pre-processing, the complexity of the interactive analysis and the resulting tissue selections. It is supported by a visual analysis framework that integrates the different approaches. The rich information content in time-dependent 3D perfusion data is both an opportunity for improved diagnosis and a challenge how to optimize the assessment of such rich data. With our comparison we contribute to a discussion between data-near and model-near assessment strategies and their respective opportunities.
@INPROCEEDINGS {oeltze09perfusion,
author = "Steffen Oeltze and Bernhard Preim and Helwig Hauser and Jarle R{\O }rvik and Arvid Lundervold",
title = "Visual analysis of cerebral perfusion data -- four interactive approaches and a comparison",
booktitle = "Proceedings of the 6th Intern. Symp. on Image and Signal Processing and Analysis (ISPA 2009)",
year = "2009",
pages = "582--589",
month = "Sept.",
abstract = "Cerebral perfusion data are acquired to characterize the regional blood supply of brain tissue. One of their major diagnostic applications is ischemicstroke assessment. We present a comparison of four interactive approaches to analyzingcerebral perfusion data from ischemic stroke patients which are based on (1) concentration-time curves (CTC) derived from the original data, (2) parameters describing the CTC shape, (3) enhancement trends computed in a statistical analysis, and (4) semi-quantitative perfusion parameters derived via parametric modelling and deconvolution. The comparison is carried out with regard to the involved data pre-processing, the complexity of the interactive analysis and the resulting tissue selections. It is supported by a visual analysis framework that integrates the different approaches. The rich information content in time-dependent 3D perfusion data is both an opportunity for improved diagnosis and a challenge how to optimize the assessment of such rich data. With our comparison we contribute to a discussion between data-near and model-near assessment strategies and their respective opportunities.",
images = "images/oeltze09perfusion1.jpg, images/oeltze09perfusion2.jpg",
thumbnails = "images/oeltze09perfusion1_thumb.jpg, images/oeltze09perfusion2_thumb.jpg"
}