Deep learning network for NMR spectra reconstruction in time-frequency domain and quality assessment

Nature Communications, Journal Year: 2025, Volume and Issue: 16(1)

Published: March 8, 2025

High-quality nuclear magnetic resonance (NMR) spectra can be rapidly acquired by combining non-uniform sampling techniques (NUS) with reconstruction algorithms. However, current deep learning (DL) based methods focus only on single-domain (time or frequency domain), leading to drawbacks like peak loss and artifact peaks ultimately failing achieve optimal performance. Moreover, the lack of fully sampled makes it difficult, even impossible, determine quality reconstructed spectra, presenting challenges in practical applications NUS. In this study, a joint time-frequency domain network, referred as JTF-Net, is proposed. It effectively combines time features, exhibiting better performance protein across various dimensions compared traditional algorithms DL methods. addition, reference-free assessment metric, denoted REconstruction QUalIty assuRancE Ratio (REQUIRER), proposed base an established space field NMR spectral reconstruction. The metric capable evaluating without making more suitable for applications. Non-uniform reduced need Here, authors propose AI model enhancing via fusion, REQUIRER, assess quality.

Language: Английский

Relaxation time measurement in liquids using compact NMR

Technobius physics., Journal Year: 2025, Volume and Issue: 3(1), P. 0027 - 0027

Published: March 21, 2025

This study investigates the influence of experimental parameters on accurate determination longitudinal and transverse relaxation times in liquids using compact nuclear magnetic resonance relaxometry. Water glycerin were selected as representative samples due to their contrasting viscosities behaviors. The primary objective was evaluate how repetition time, echo number data points, time step affect precision T₁ T₂ measurements. Longitudinal determined a variable method, while measured via multi-echo spin sequence. Exponential fitting algorithms employed extract from recorded signal amplitudes. For water, found be approximately 3.0 s for 1.423 T₂. In contrast, exhibited significantly shorter times, with estimated at 0.126 0.094 s. results demonstrated that estimation requires carefully optimized acquisition settings. Specifically, must exceed three value ensure full recovery, short high echoes are essential reliable determination. findings address critical methodological gap relaxometry protocols offer practical recommendations enhancing measurement accuracy simple liquids.

Language: Английский