Computer oriented image acquisition of the liver: Toward a better numerical model for radiofrequency ablation.

Simulating physiological interventions for planning purposes requires an accurate virtual liver model as computation input. To best meet the demands the data acquisition has to be oriented on image processing purposes. We provide a CT imaging protocol which makes it possible to extract much more vessels with the same segmentation algorithms than when using them on data sets from the clinical routine. Medical evaluation of physiological models demand a statistical evaluation in a pre-clinical study, that means in a first step reproducible results for a large number of subjects. So data acquisition should be as automatic as possible without neglecting modeling demands. Image quality should be reproducible to guarantee an ongoing high quality of image processing results. We evaluate the protocol by comparison of segmentation results with results achieved on standard data sets from clinical routine using the same segmentation methods. Results show that typically up to ten times more vessels can be extracted and the surface accuracy is improved.