A new framework for multiscale and multiorientation image analysis and its application to medical volume data analysis

点击翻译按钮获取中文摘要

This project investigated a basic framework and clinical applications for the enhancement, segmentation, and quantification of line-like shapes such as blood vessels and sheet-like shapes such as articular cartilage in medical volume data. Based on multiscale Gaussian filters and the Hessian matrix of the volume function, line and sheet structures with various widths are enhanced. Local intensity structures are represented as continuous functions by further combining the gradient vectors. Medial axis/surface elements are locally determined based on the second-order approximations of local intensity structures. Diameter/thickness quantification is performed based on detected medial axis/surface elements, and various widths of structures are combined through multiscale integration. The utility of the framework was demonstrated by applications to the characterization of pulmonary nodules based on the spatial distribution of vessels surrounding pulmonary nodules, and articular cartilage thickness determination using real computed tomography (CT) and magnetic resonance (MR) volume data.