Probabilistic Non-Rigid Registration for Safe Brain Tumor Resection
Brain tumors strike people in the prime of life. Surgical resection is the initial treatment for nearly all brain tumors and aims at maximizing the extent of tumor resection while preserving the patient's cognitive function. To op...
ver más
¿Tienes un proyecto y buscas un partner? Gracias a nuestro motor inteligente podemos recomendarte los mejores socios y ponerte en contacto con ellos. Te lo explicamos en este video
Información proyecto SafeREG
Duración del proyecto: 27 meses
Fecha Inicio: 2024-04-17
Fecha Fin: 2026-07-31
Fecha límite de participación
Sin fecha límite de participación.
Descripción del proyecto
Brain tumors strike people in the prime of life. Surgical resection is the initial treatment for nearly all brain tumors and aims at maximizing the extent of tumor resection while preserving the patient's cognitive function. To optimize this tradeoff, neuronavigation systems have been developed to provide intraoperative guidance to surgeons. These systems allow for the visualization of the position of surgeons' surgical tools relative to the tumor and critical brain areas visible in preoperative Magnetic Resonance Imaging. However, these systems become inaccurate as the surgery progresses since they do not account for brain deformation and tissue resection occurring during surgery.
In an interdisciplinary effort, project SafeREG combines the researcher's background to the expertise of computational scientists from INRIA and clinicians from Parisian hospitals. Its objective is to invent a novel image registration methodology with intraoperative ultrasound that is rich enough to capture complex deformations occurring at the tumor and resection cavity boundaries, fast enough to be employable clinically, and interpretable enough for informed decision-making by neurosurgeons. This will be accomplished by pushing the envelope of scientific knowledge in (1) cross-modality domain adaptation for weakly- and unsupervised image segmentation; (2) modality-invariance representation learning using contrastive learning; (3) non-rigid registration with discrete probabilistic methods; (4) simulated-based variational inference for registration uncertainty quantification that leverages biomechanical knowledge. This research has the potential to deliver accurate and informed image-guided surgery, conferring a lower risk of new neurologic deficits and improved patient prognosis. Beyond neurosurgery, it has broad applications to additional areas of image-guided therapy, including spine, liver, and prostate surgery.