Deformable Thoracic CT Images Sequence Registration Using Strain Energy Minimization

Abstract

The idea of deformable image registration (DIR) has been explored for a thoracic CT (computed tomography) image database of ten subjects. Thoracic CT image acquisition for clinical interventions requires a well-defined procedure that has already been underlined on the basis of field expertise and past experiences. Despite strict adherence to the procedure, the acquired images are prone to distortions and artefacts. This might happen due to organ motion during the breathing process (at times even in breath-hold procedures), slight (even involuntary) movements or acquisition variations in supine and prone positions, etc. An intensity differences based energy minimization method has been proposed. The moving image is transformed in the process such that it gets maximum alignment with the fixed image. This is achieved by energy minimization of the moving image in an iterative process. It is a simpler and more practical method for thoracic CT image registration than the prevalent approaches. This has been shown by lower mean registration errors for the patient data; the errors were as such axial: 0.283 ± 0.08, coronal: 0.784 ± 0.32 and sagittal: 0.66 ± 0.2 pixels. This registration of moving image onto the fixed image in the sequence will help in minimizing the adverse effects of the otherwise present discrepancies, phase errors, and discontinuity artifacts that might have crept in during the acquisition.