The CIBM Breakfast and Science Seminars unite the scientific community on a regular basis (every last Tuesday of the month) to exchange knowledge, foster research collaborations, and create a vibrant research center of excellence in biomedical imaging. This seminar is composed of two 20-minute presentations on a specific research area in biomedical imaging, followed by a 5-10-minute Q&A discussion. The events will be concluded with an interactive sharing session where audience members have an opportunity to share news and updates with each other or simply network with members of the CIBM community.
Deep learning methods for fetal brain MRI tissue segmentation
Priscille Guerrier de Dumast, MIAL UNIL & CIBM SP CHUV-UNIL
The quantitative assessment of the developing human brain in utero is crucial to fully understand neurodevelopment. T2-weighted magnetic resonance imaging offers a good contrast between brain tissues, hence allowing to assess the brain growth and detect abnormalities in utero. In the clinical context, fetal brain MRI is performed with fast, 2D orthogonal series in order to minimize the effect of unpredictable fetal motion but results in low out-of-plane spatial resolution and significant partial volume effect. To combine these multiple series, advanced imaging techniques based on super-resolution (SR) algorithms allow the reconstruction of 3D high-resolution motion-free isotropic volumes. Such high-resolution volume opens up to the possibility of advanced quantitative analysis of improved accuracy. Accurate MR image segmentation, and more importantly a topologically correct delineation of the structure, is a key baseline to perform further morphometric and volumetric analysis of brain development. Nevertheless, the development of automatic machine learning based methods is hampered by the scarcity of the data and their multiple sources of variability. In this talk, I will discuss my work on the introduction of a topological constraint in the segmentation of the developing fetal cortical plate.
Fast in vivo assay of creatine kinase in human brain by 31P magnetic resonance fingerprinting
Mark Widmaier, LIFMET EPFL & CIBM MRI EPFL
31P MRS is a powerful tool for studying brain energy metabolism, however, it remains a challenging task. Especially, relaxometry and chemical exchange rates acquisitions, suffer from the inherent low sensitivity and long acquisition times. We introduce MT- 31P magnetic resonance fingerprinting, using a SAR efficient magnetization transfer (MT) approach. We extended the magnet resonance fingerprinting (MRF) framework to overcome obstacles of in vivo human brain 31P measurements. The result is an efficient way to measure relaxation parameters and the creatine kinase rate kCK, enabling ultra-fast kCK measurement in up to 2:15 min scan time. This talk will give a short journey into the challenges and possibilities of 31P-MRF.