Identification of invariant anatomical markers for medical image registration using Gabor Filters

BALLARIN, VIRGINIA LAURA; MESCHINO, GUSTAVO JAVIER; ISA JARA, RAMIRO

Córdoba

Congreso; XXI Congreso Argentino de Bioingeniería; 2017

Sociedad Argentina de Bioingeniería (SABI)

Minimally invasive surgery, biopsies and radiation therapy planning increasingly depend on the combined information of different imaging modalities. Images need to be precisely aligned to reflect real relationships between pixels. Registration, defined as the alignment of different coordinate systems, is an essential step, which requires invariant anatomic markers. In this paper, we present a new approach to identify invariant anatomical markers, based on properties of Gabor filters. The use of features based on these filters allows extracting invariant regions in images using Zernike moments as feature descriptors. We applied this method in magnetic resonance brain images, to recognize invariant regions as anatomical markers, to use them for a later registration process. The method of feature identification consists in applying a Gabor filter set, with different scales and rotations in order to compute the energy response. Afterwards, distinctive points (maximum and minimum local pixels) are selected, computing contrast, scale factors and Zernike Moments. The selected anatomical markers allow feature correspondence between two images (without any deformation and with an isometric transformation applied). Preliminary results show that this approach is suitable to find reliable markers for a later image registration.