A CNN Based Software Gradiometer for Electromagnetic Background Noise Reduction in Low Field MRI Applications

Area of Science:

• Medical Imaging
• Biophysics
• Signal Processing

Background:

• Electromagnetic noise significantly degrades signal-to-noise ratio (SNR) in ultra-low field magnetic resonance imaging (UFL-MRI).
• Traditional electromagnetic shielding enclosures are costly and impede the portability of UFL-MRI systems.
• Developing effective noise suppression methods is crucial for advancing UFL-MRI applications.

Purpose of the Study:

• To introduce a CNN-based software gradiometer for suppressing ambient electromagnetic noise in UFL-MRI.
• To demonstrate the efficacy of this software approach in enhancing SNR without physical shielding.
• To explore the potential for making UFL-MRI systems more portable and accessible.

Main Methods:

• Utilized three ambient noise monitoring coils positioned separately from the UFL-MRI signal detector.
• Employed a convolutional neural network (CNN) to synthesize and subtract ambient noise from the signal detector.
• Developed mathematical foundations to support the noise suppression framework.

Main Results:

• Achieved up to 20-fold noise suppression using an optimized CNN and simultaneous noise measurements.
• Demonstrated successful suppression of inductively coupled electromagnetic background noise.
• Validated the CNN-based software gradiometer's performance in enhancing UFL-MRI signals.

Conclusions:

• The CNN-based software gradiometer offers a powerful and cost-effective alternative to traditional shielding for UFL-MRI.
• This approach significantly improves SNR, making UFL-MRI measurements more reliable.
• The proposed method holds substantial potential for realizing truly portable UFL-MRI systems.