Accurate localization of the target is essential to reduce morbidity during brain tumor removal interventions. Yet, image-guided neurosurgery faces an important issue for large skull openings where brain soft-tissues can exhibit large deformations in the course of surgery. As a consequence of this "brain-shift" the pre-operatively acquired images no longer correspond to reality and subsequent neuronavigation is therefore strongly compromised. In this article we present a neuronavigator which addresses this issue and offers passive help to the surgeon by displaying the position of the guided tools with respect to the corrected location of the anatomical features. This low-cost system relies on localized 2D Doppler ultrasound imaging of the brain which makes it possible to track the vascular tree deformation throughout the intervention. An elastic registration procedure is used to match the shifted tree with its pre-operative structure identified within Magnetic Resonance Angiography images. A patient speci c Finite Element biomechanical model of the brain further extends the resulting sparse deformation field to the overall organ volume. Finally, the estimated global deformation is applied to all pre-operatively available volumetric images or data, such as tumor contours, and the corrected planning is displayed to the surgeon. The system, tested on a patient presenting a large meningioma, was able to compensate within seconds for the intraoperatively observed brain-shift, reducing the mean error on tumor margin localization from 3.5 mm (max=7.6 mm, RMS=3.7 mm) to 0.9 mm (max=1.7 mm, RMS=1.0 mm).