Controlling the segmentation parameters by case-based reasoning

We propose a case-based image segmentation method, which takes the non-image information and the image characteristics and selects among a set of cases the case which fits best to the current case. The segmentation parameter associated to the close case are applied to the segmentation unit and taken for segmentation of the current case. By taking into account the non-image and image information we break down our complex solution space to a subspace of relevant cases where the variation among the cases is limited. Besides that with case based reasoning we can incrementally learn new image segmentation parameters. We use our approach for determination of brain/liquor ratio in CT images. This parameter is used for diagnosis of Alzheimer disease