Dynamic infrared imaging for breast cancer diagnosis: a feature based registration approach

An essential, preprocessing step in dynamic infrared imaging applied to breast cancer diagnosis is a proper registration of the thermographic image sequence. The failure to do so may produce artifacts in the FFT power spectra of the temperature sequences due to breast movement which may constitute an important confounding factor potentially invalidating the data analysis. In this work we wish to draw attention to this issue, and show the influence of movement on a case sample of subjects, as well as the correction which may be introduced by realigning the image sequences with four different types of feature-based image registration. The registration process was preventively analyzed in the visible field and then applied to the dynamic thermography images. Results confirm the importance of the registration and encourage further work in the area.