DiffNMR: Advancing Inpainting of Randomly Sampled Nuclear Magnetic Resonance Signals

2025-05-26

AI Summary

To address severe artifacts and low fidelity in non-uniformly sampled (NUS) nuclear magnetic resonance (NMR) spectrum reconstruction, this work introduces denoising diffusion probabilistic models (DDPMs) to NMR signal recovery for the first time. We propose a time–frequency joint sparse representation framework incorporating complex-domain preprocessing, adaptive noise scheduling, and Fourier-domain constraints—significantly outperforming conventional time–time domain modeling. On the Artina benchmark, our method effectively suppresses NUS-induced artifacts, improves signal-to-noise ratio and peak shape accuracy, and achieves state-of-the-art spectral fidelity. Moreover, reconstruction time is reduced by 3–5× compared to existing approaches. This study not only establishes the efficacy of diffusion models for NMR spectral reconstruction but also pioneers a novel time–frequency collaborative modeling paradigm, opening new avenues for deep generative modeling in analytical spectroscopy.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Nuclear Magnetic Resonance (NMR) spectroscopy leverages nuclear magnetization to probe molecules' chemical environment, structure, and dynamics, with applications spanning from pharmaceuticals to the petroleum industry. Despite its utility, the high cost of NMR instrumentation, operation and the lengthy duration of experiments necessitate the development of computational techniques to optimize acquisition times. Non-Uniform sampling (NUS) is widely employed as a sub-sampling method to address these challenges, but it often introduces artifacts and degrades spectral quality, offsetting the benefits of reduced acquisition times. In this work, we propose the use of deep learning techniques to enhance the reconstruction quality of NUS spectra. Specifically, we explore the application of diffusion models, a relatively untapped approach in this domain. Our methodology involves applying diffusion models to both time-time and time-frequency NUS data, yielding satisfactory reconstructions of challenging spectra from the benchmark Artina dataset. This approach demonstrates the potential of diffusion models to improve the efficiency and accuracy of NMR spectroscopy as well as the superiority of using a time-frequency domain data over the time-time one, opening new landscapes for future studies.

Problem

• Enhancing reconstruction quality of Non-Uniform Sampling NMR spectra
• Reducing artifacts and improving spectral quality in NMR
• Exploring diffusion models for time-time and time-frequency NUS data

Innovation

• Deep learning enhances NUS spectra reconstruction
• Diffusion models applied to time-time and time-frequency data
• Time-frequency domain data outperforms time-time domain