Automated organ localization in medical images can be very effective in enhancing a radiologist's productivity through smarter workflow design and reporting tools. In this paper, we present an algorithm for organ localization in an image through registration with an atlas data. The challenge comes from supporting non-rigid registration of multi-modality inter-patient data. The proposed solution is an extension of blob-based nonrigid registration framework for multi-modality registration. Blob-based framework is a fast non-rigid registration approach providing flexibility of constraints across different image regions or blobs, and which facilitates controlling topological changes inside blobs. An important step in this framework is defining right blobs for a good inter-patient registration. We describe solution to choosing the right blobs through body region identification for acceptable organ localization.
展开▼
