Magnetic Resonance Analysis Using Artificial Neural Networks

Standing

Undergraduate

Type of Proposal

Visual Presentation (Poster, Installation, Demonstration)

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 lifetimes will reveal molecular scale information, providing insights into medical diagnoses. Conventional methods to extract such information require a large amount of data with high signal-to-noise ratio (SNR), leading to impractical measurement times. A new signal analysis method, robust against noise, will enable quantitative MR measurements in a feasible time scale. Artificial neural networks (ANN) 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 nonlinear problems. Recently, neural networks have gained significant recognition and have been employed in applications such as cancer classification, computer vision, and self-driving cars. We propose to use ANN’s to recover magnetic resonance signal lifetimes. The ANN’s were trained on 25 sets of 7000 simulated scans with 5 different SNR’s, from low to high, and with 5 different input sizes, to recover magnetic resonance signal lifetimes. The traditional least squares fitting (LSF) method was employed as a comparison. ANN outperformed LSF in all the simulation datasets. In particular, with very low input data SNR, ANN produced reliable results while LSF failed.

Location

WIndsor

Grand Challenges

Viable, Healthy and Safe Communities

Share

COinS
 

Magnetic Resonance Analysis Using Artificial Neural Networks

WIndsor

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 lifetimes will reveal molecular scale information, providing insights into medical diagnoses. Conventional methods to extract such information require a large amount of data with high signal-to-noise ratio (SNR), leading to impractical measurement times. A new signal analysis method, robust against noise, will enable quantitative MR measurements in a feasible time scale. Artificial neural networks (ANN) 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 nonlinear problems. Recently, neural networks have gained significant recognition and have been employed in applications such as cancer classification, computer vision, and self-driving cars. We propose to use ANN’s to recover magnetic resonance signal lifetimes. The ANN’s were trained on 25 sets of 7000 simulated scans with 5 different SNR’s, from low to high, and with 5 different input sizes, to recover magnetic resonance signal lifetimes. The traditional least squares fitting (LSF) method was employed as a comparison. ANN outperformed LSF in all the simulation datasets. In particular, with very low input data SNR, ANN produced reliable results while LSF failed.