Evaluation of time-series registration methods in dynamic area telethermometry for breast cancer detection

Automated motion reduction in 3D dynamic infrared imaging is on demand in many applications. Few methods for registering time-series dynamic infrared frames have been proposed. Almost all such methods are feature based algorithms requiring manual intervention. We apply different automated registration methods based on spatial displacement to 11 datasets of Breast Dynamic Infrared Imaging (DIRI) and evaluate the results in terms of both the image similarity and anatomical consistency of the transformation. The aim is to optimize the registration strategy for breast DIRI in order to improve the spectral analysis of temperature modulation; thus facilitating the acquisition procedure in a Dynamic Area Telethermometry framework. The results show that symmetric diffeomorphic demons registration outperforms both warped frames similarity and smoothness of deformation fields; hence proving effective for time-series dynamic infrared registration.