Selection of Reference Channels Based on Mutual Information for Frequency-Dependent Subtraction Method Applied to...

Area of Science:

• Biomagnetism
• Signal Processing
• Medical Physics

Background:

• Fetal magnetoencephalography (fMEG) recordings are crucial for understanding fetal brain development.
• Maternal magnetocardiograms (mMCG) and fetal magnetocardiograms (fMCG) can interfere with fMEG signals.
• Current methods for interference attenuation may require manual selection of reference sensors.

Purpose of the Study:

• To develop and evaluate an automated method for selecting reference sensors for interference reduction in fMEG.
• To improve the accuracy and efficiency of the frequency-dependent subtraction (SUBTR) method.
• To enhance the quality of fetal brain activity measurements.

Main Methods:

• Utilized minimal redundancy and maximal relevance (mRMR) criteria based on mutual information to select reference sensors.
• Applied the selected references to the SUBTR method for attenuating maternal (mMCG) and fetal (fMCG) signals.
• Evaluated the performance of the mRMR-based SUBTR method on 38 real-world datasets by measuring MCG amplitude reduction.
• Compared the mRMR approach against random sensor selection.

Main Results:

• The mRMR-based reference selection demonstrated significant improvements in interference removal compared to random selection.
• Quantifiable differences in signal quality were observed based on the number of references chosen by mRMR.
• The study confirmed the effectiveness of mRMR in identifying optimal reference sensors.

Conclusions:

• Minimal redundancy and maximal relevance (mRMR) offers an effective automated approach for selecting optimal reference sensors.
• This method enhances the utility of the SUBTR technique for biomagnetic signal processing.
• The mRMR approach is adaptable to various sensor array data applications beyond biomagnetism.