Meeting Banner
Abstract #0885

Automatic tissue-type classification of 1H-MRSI spectra in patients with glioblastoma

Nuno Pedrosa de Barros1, Raphael Meier2, Martin Pletscher1, Urspeter Knecht1, Mauricio Reyes2, Roland Wiest1, and Johannes Slotboom1

1SCAN / Neuroradiology, University Hospital Bern (Inselspital), Bern, Switzerland, 2Institute for Surgical Technology & Biomechanics, University of Bern, Bern, Switzerland

Clinical-routine MRSI-data analysis is commonly performed through the visual inspection of multiple metabolite and metabolite-ratio maps, and aims at translating the different spectroscopic patterns into known tissue-types, such as, necrosis, solid tumour, tumour-infiltration, normal-brain-tissue, etc. Such translation/segmentation requires solid expertise in MR-spectroscopy, which most clinicians do not have. Bad-quality-data, as well as frequency-dependant-selection-profiles further complicate proper interpretation of MRSI-data. Therefore, to facilitate the clinical-use of MRSI, we present an automatic MRSI-tissue-type segmentation algorithm, that includes automatic-quality-filtering and selection-profile-correction. The method was tested in glioblastoma and the tissue-types were compared against an MRI-based tumour-segmentation-method.

This abstract and the presentation materials are available to members only; a login is required.

Join Here