Original title: Deep Learning For Magnetic Resonance Spectroscopy Quantification: A Time-Frequency Analysis Approach
Authors: Shamaei, Amirmohammad
Document type: Papers
Language: eng
Publisher: Vysoké učení technické v Brně, Fakulta elektrotechniky a komunikačních technologií
Abstract: Magnetic resonance spectroscopy (MRS) is a technique capable of detecting chemical compounds from localized volumes in living tissues. Quantification of MRS signals is required for obtaining the metabolite concentrations of the tissue under investigation. However, reliable quantification of MRS is difficult. Recently deep learning (DL) has been used for metabolite quantification of MRS signals in the frequency domain. In another study, it was shown that DL in combination with time-frequency analysis could be used for artifact detection in MRS. In this study, we verify the hypothesis that DL in combination with time-frequency analysis can also be used for metabolite quantification and yields results more robust than DL trained with MR signals in the frequency domain. We used the complex matrix of absolute wavelet coefficients (WC) for the timefrequency representation of the signal, and convolutional neural network (CNN) implementation for DL. The comparison with DL used for quantification of data in the frequency domain is presented.
Keywords: magnetic resonance spectroscopy; quantification; deep learning; machine learning
Host item entry: Proceedings II of the 26st Conference STUDENT EEICT 2020: Selected papers, ISBN 978-80-214-5868-0

Institution: Brno University of Technology (web)
Document availability information: Fulltext is available in the Brno University of Technology Digital Library.
Original record: http://hdl.handle.net/11012/200638

Permalink: http://www.nusl.cz/ntk/nusl-447690


The record appears in these collections:
Universities and colleges > Public universities > Brno University of Technology
Conference materials > Papers