Enhancing Fetal Brain Imaging: ALPS-FMEG Technique Achieves Accurate Signal Extraction by Mitigating Movement Artifacts.

PURPOSE: Fetal magnetoencephalography (fMEG) enables non-invasive monitoring of fetal brain function with high temporal resolution. However, how can we isolate low signal-to-noise ratio signals of the developing brain when disruptive artifacts arise from maternal and fetal movements? Addressing this challenge is critical for understanding brain development. We present Advanced Localization and Processing of fMEG Signals based on Maternal and Gross fetal body Movement Exclusion (ALPS-FMEG), a MATLAB-based framework that improves fetal brain signals by removing fetal and maternal movement artifacts.

METHODS

ALPS-FMEG integrates Independent Component Analysis for separation and reconstruction of fetal brain, fetal and maternal cardiac signal components in sensor space, Empirical Mode Decomposition for noise reduction, and a movement artifact detection-and-exclusion technique based on actogramCOG associated with heart rate patterns. This novel integration modifies the actogramCOG approach by pre-interpolating R waves for enhanced robustness and combines it with HRV-based logic gates, representing a first in fMEG processing to achieve artifact-free signals while preserving physiological latencies.

RESULTS

ALPS-FMEG was applied to 50 fMEG datasets from 28 to 39 weeks of gestation, enhancing signal quality. For group analysis, 45 datasets were retained after excluding recordings with auditory event-related field (fAEF) latencies < 70 ms. In these, it significantly improved signal-to-noise ratio and fAEF amplitudes (p < 0.0001), with preserved latencies. fAEF latency showed a significant negative correlation with gestational age (p < 0.001).

CONCLUSION

ALPS-FMEG improves fetal brain signal extraction by addressing movement artifacts. This method supports robust fetal brain analysis and may be adaptable to future fMEG systems, including optically pumped magnetometers, enhancing prenatal neurophysiology and clinical research, though manual steps currently limit scalability and could be addressed via automation for broader practical use.