2017/06/14 by Giulio Siracusano, Aurelio La Corte, Siracusano, Giulio +5
Engineering · Medicine · #Advanced Chemical Sensor Technologies #Advanced MRI Techniques and Applications #FOS: Physical sciences #Medical Physics (physics.med-ph) #Microfluidic and Capillary Electrophoresis Applications
paper · pdf · doi:10.48550/arxiv.1706.04525
openalex publication_date 2017/06/14 · openalex created_date 2022/10/02 · openalex updated_date 2026/07/28
Multi-echo Chemical Shift Encoded methods for Fat-Water quantification are\ngrowing in clinical use due to their ability to estimate and correct some\nconfounding effects. State of the art CSE water-fat separation approaches rely\non a multi-peak fat spectrum with peak frequencies and relative amplitudes kept\nconstant over the entire MRI dataset. However, the latter approximation\nintroduces a systematic error in fat percentage quantification in patients\nwhere the differences in lipid chemical composition are significant, such as\nfor neuromuscular disorders, because of the spatial dependence of the peak\namplitudes. The present work aims to overcome this limitation by taking\nadvantage of an unsupervised clusterization-based approach offering a reliable\ncriterion to carry out a data-driven segmentation of the input MRI dataset into\nmultiple regions. The idea is to apply the clusterization for partitioning the\nmulti-echo MRI dataset into a finite number of clusters whose internal voxels\nexhibit similar distance metrics. For each cluster, the estimation of the fat\nspectral properties are evaluated with a self-calibration technique and finally\nthe fat-water percentages are computed via a non-linear fitting. The method is\ntested in ad-hoc and public datasets. The overall performance and results in\nterms of fitting accuracy, robustness and reproducibility are compared with\nother state-of-the-art CSE algorithms. This approach provides a more accurate\nand reproducible identification of chemical species, hence fat-water\nseparation, when compared with other calibrated and non-calibrated approaches.\n