Magnetic Resonance Signal Lifetime Spectra Analysis Using Artificial Neural Networks

Standing

Undergraduate

Type of Proposal

Oral Research Presentation

Challenges Theme

Open Challenge

Your Location

Windsor

Faculty

Faculty of Science

Faculty Sponsor

Dr. Dan Xiao

Proposal

Magnetic resonance imaging (MRI) is widely used as a non-invasive diagnostic technique to visualize the internal structure of biological systems. Quantitative analysis of magnetic resonance signal lifetime spectra will reveal molecular scale information, providing insights into medical diagnoses. Conventional methods to extract MR spectra are computationally intensive, requiring large amounts of data and generally lacking the spectrum peak widths information. A novel computationally efficient signal analysis method, based on artificial neural networks (ANN), has been developed to provide accurate real-time quantitative MR spectrum analysis.

Artificial neural networks are a series of densely connected, simple information processing nodes which cumulatively map a set of inputs to several features. ANN is a powerful tool to solve complex nonlinear problems and has been successfully employed in applications such as cancer classification, computer vision, and self-driving cars. The network structure and hyperparameters have been optimized to determine MR spectra. ANNs were trained on 400 000 simulated scans, with varying numbers of peaks, noise levels and input data sizes. The traditional inverse Laplace transform (ILT) method was employed as a comparison and the performance of both methods has been evaluated across a large parameter range. In addition to superior computation speed, higher accuracy was achieved compared to the traditional method and the width of the spectra could be determined within the noise limit.

This method could be easily translated to other areas with exponential analysis, such as fluorescence decay and radioactive decay. It could also be extended to solve ill-posed inversion problems in general.

Share

COinS
 

Magnetic Resonance Signal Lifetime Spectra Analysis Using Artificial Neural Networks

Magnetic resonance imaging (MRI) is widely used as a non-invasive diagnostic technique to visualize the internal structure of biological systems. Quantitative analysis of magnetic resonance signal lifetime spectra will reveal molecular scale information, providing insights into medical diagnoses. Conventional methods to extract MR spectra are computationally intensive, requiring large amounts of data and generally lacking the spectrum peak widths information. A novel computationally efficient signal analysis method, based on artificial neural networks (ANN), has been developed to provide accurate real-time quantitative MR spectrum analysis.

Artificial neural networks are a series of densely connected, simple information processing nodes which cumulatively map a set of inputs to several features. ANN is a powerful tool to solve complex nonlinear problems and has been successfully employed in applications such as cancer classification, computer vision, and self-driving cars. The network structure and hyperparameters have been optimized to determine MR spectra. ANNs were trained on 400 000 simulated scans, with varying numbers of peaks, noise levels and input data sizes. The traditional inverse Laplace transform (ILT) method was employed as a comparison and the performance of both methods has been evaluated across a large parameter range. In addition to superior computation speed, higher accuracy was achieved compared to the traditional method and the width of the spectra could be determined within the noise limit.

This method could be easily translated to other areas with exponential analysis, such as fluorescence decay and radioactive decay. It could also be extended to solve ill-posed inversion problems in general.