A System for Performing Automated Measurements on Large Bone Databases

Abstract

Many medical applications benefit from gathering anatomical data of large numbers of patients into a centralized database. Current systems are easily able to handle several thousand datasets in diverse representations, such as CT, segmented bone surfaces etc., annotated with patient metadata (e.g. age, ethnic group or height). In this paper, we present a method for performing automated measurements of quantities such as distances, angles or circle diameters on a large number of pre-segmented bone samples. The quantity to be measured is specified on a template shape, and then identified on every sample respectively through correspondence matching. Additionally, the system allows the inclusion of automatically extracted anatomical landmarks in the measurements. By filtering the input dataset based on patient attributes, the measured quantities can be statistically analyzed with respect to specific subpopulations (e.g. differentiating by sex or ethnic group, to be used in the design of optimized implants). The quality of the automated measurements was assessed through a comparison with an evaluation performed manually.