Repeat CT scanning consists of acquiring multiple CT scans of the same patient at various times. It is frequently used in many clinical situations, e.g., to assess disease progression, to evaluate response to treatment and to track the patient and the needle during interventional procedures. The main image analysis goal, which is to identify the changes between the baseline and the follow up scans, is time-consuming, error prone, and requires radiological expertise. Moreover, since each scan adds cumulative radiation that may be harmful to the patient, it is highly desirable to optimize the radiation dose of each scan. However, lower doses reduce image quality and thereby difficult even more the image interpretation task.
We have developed a new computational paradigm for on-line radiation dose optimization and automatic change detection in repeat CT scanning. The key principle of our approach is to perform sparse repeat scanning to significantly reduce the radiation dose and to obtain the missing information from the baseline scan without image quality loss. Our approach is unique in that it formulates the problem as sparse sinogram comparison problem in 3D Radon space instead of an image reconstruction problem in image space. We will describe novel methods for registration of the baseline and the repeat scan, for the automatic identification of regions where the changes have occurred, and for image-less needle tracking in interventional radiology. Our experimental results show that these tasks can be accomplished with a dose reduction of about x10.
Joint work with Guy Medan, Naomi Shamul and Zeev Adelman, PhS students.