Extended Normalized Mutual Information for Lung SPECT - CT Registration

In this paper an extension of the normalized mutual information is proposed to register standalone obtained low dose CT with already superimposed lung inhalation SPECT and perfusion SPECT images. In order to validate our method, superimposed inhalation SPECT, perfusion SPECT and low dose CT image triples were collected obtained by a hybrid SPECT/CT camera at the same time. A known transformation was applied to the low dose CT to simulate a misalignment, followed by an optimal transformation search performing our extended normalized mutual information-based (eNMI) method. For comparison and evaluation, the low dose CT was also registered to both inhalation and perfusion images one-by-one applying a dual-normalized mutual information-based (dNMI) method. Comparative results have shown that our eNMI method worked with minimal registration error and number of iterations, hence it can be successfully applied to stand alone performed low dose CT - SPECT registrations. I. INTRODUCTION