A Learning Method to Estimate Multi-compartmental T2 Distributions with low Data Requirements
Daniel Vallejo-Aldana1, Arturo Gonzalez-Vega2, Victor H Hernandez2, Valeria Piazza3, Milvia Alata3, Jonathan Rafael-Patiño4,5, Thomas Yu4,6, Luis Concha7, and Alonso Ramirez-Manzanares8
1Mathematics Department, Universidad de Guanajuato, Guanajuato, Mexico, 2Department of Chemical, Electronic and Biomedical Engineering, Division of Sciences and Engineering, University of Guanajuato, Leon, Mexico, 3Center of Research in Optics, Leon, Mexico, 4Signal Processing Lab 5 (LTS5), Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, 5Radiology Department, Centre Hospitalier Universitaire Vaudois and University of Lausanne, Lausanne, Switzerland, 6Medical Image Analysis Laboratory, Center for Biomedical Imaging (CIBM), University of Lausanne, Lausanne, Switzerland, 7Institute of Neurobiology, Universidad Nacional Autonoma de Mexico, Juriquilla, Mexico, 8Computer Science, Centro de Investigación en Matemáticas A.C., Guanajuato, Mexico
The estimation of intravoxel distributions of T2 values based on multi-echo MR data is a challenging task. Interestingly, the information above is quite useful to detect damage on brain tissue, e.g. to estimate myelin-water-fraction changes associated with demyelination processes. Currently available methods typically require a long train of echoes, which are not always feasible to acquire. In this work we tackle this problem using state-of-the-art supervised learning convolutional networks to build a robust prediction model on very limited data ( 5 echoes and 4 TR). The methodology identifies myelin abnormalities in a rodent model of a neurological disorder with demyelination.
This abstract and the presentation materials are available to members only;
a login is required.