CAREER: Novel Geometric Techniques for Optimal Surface Detection in Medical Images

点击翻译按钮获取中文摘要

CAREER: Novel Geometric Techniques for Optimal Surface Detection in Medical ImagesXiaodong WuDepartment of Electrical and Computer EngineeringDepartment of Radiation OncologyThe University of IowaAutomated detection of globally optimal surfaces representing object boundaries in volumetric images is fundamental and remains challenging in modern computer-assisted medical diagnosis and treatment as well as other important medical applications. The analyses of 3-D data are typically performed visually in routine clinical practice, and are done in 2-D on a slice-by-slice basis. Maximum intensity projections providing 2-D images of the 3-D data or qualitative volume renderings are also common. However, reliable quantitative volumetric 3-D and 4-D segmentation methods that are robust, user-friendly, and applicable to clinical practice are not routinely available. As such, a significant and possibly critical portion of the image information available in the acquired data may unfortunately be never used for clinical diagnostic purposes. This research involves directly addressing the needs of automated segmentation and will result in novel methodologies and practical tools that greatly accelerate the pace of medical image analysis. The investigator studies the specific problems of optimally detecting single and multiple interacting surfaces in volumetric medical images. Surfaces with high-order smoothness are of particular interest. The newly introduced edge-weighted geometric graph representation for the surface detection problems allows incorporating a wide spectrum of image features into the formulation, admitting highly accurate segmentation results. However, it also brings tremendous computational challenges. The target geometric problems include highly smooth surface identification, geometric layer partitioning, surface reconstruction from noisy data, digital rays, constrained k-means clustering, and geometric multiple labeling. The application of computational geometry techniques to medical imaging is intellectually deep and will result in advances in both computational geometry and medicine. An important goal of this research is dissemination of implemented algorithms in software to medical domains. In doing this, it can help to further integrate computer science with medicine. Beyond its research impact, the educational goal of this project is to facilitate the institution-wide effort to foster and promote the extension of the strength of biomedical research at the University of Iowa to successful interdisciplinary undergraduate and graduate programs.