On the registrability of two CT volumes

Abstract

This paper describes a method to determine whether two CT volumes overlap in anatomy or not. This is an important problem because radiologists have to manually select which series should be registered together for follow-up exams. This task is becoming more and more tedious as the number of studies and series for each patient increases in large hospital settings, and meta-data is often erroneous, incomplete, or inconsistent, and therefore unreliable. We demonstrate on 40 patients and 100 possible matches that our tool is successful in identifying the overlapping (or registrable) cases automatically. We also show that this is not possible using the residual error after registration.