Automated classification of food products using 2D low-field NMR

Rapid estimation approach for glycosylated serum protein of human serum based on the combination of deep learning and TD-NMR technology

Rapid and precise estimation of glycosylated serum protein (GSP) of human serum is of great importance for the treatment and diagnosis of diabetes mellitus. In this study, we propose a novel method for estimation of GSP level based on the combination of deep learning and time domain nuclear magnetic resonance (TD-NMR) transverse relaxation signal of human serum. Specifically, a principal component analysis (PCA)-enhanced one-dimensional convolutional neural network (1D-CNN) is proposed to analyze the TD-NMR transverse relaxation signal of human serum. The proposed algorithm is proved by accurate estimation of GSP level for the collected serum samples. Furthermore, the proposed algorithm is compared with 1D-CNN without PCA, long short-term memory network (LSTM) and some conventional machine learning algorithms. The results indicate that PCA-enhanced 1D-CNN (PC-1D-CNN) has the minimum error. This study proves that proposed method is feasible and superior to estimate GSP level of human serum using TD-NMR transverse relaxation signals.

NMR in Metabolomics: From Conventional Statistics to Machine Learning and Neural Network Approaches

NMR measurements combined with chemometrics allow achieving a great amount of information for the identification of potential biomarkers responsible for a precise metabolic pathway. These kinds of data are useful in different fields, ranging from food to biomedical fields, including health science. The investigation of the whole set of metabolites in a sample, representing its fingerprint in the considered condition, is known as metabolomics and may take advantage of different statistical tools. The new frontier is to adopt self-learning techniques to enhance clustering or classification actions that can improve the predictive power over large amounts of data. Although machine learning is already employed in metabolomics, deep learning and artificial neural networks approaches were only recently successfully applied. In this work, we give an overview of the statistical approaches underlying the wide range of opportunities that machine learning and neural networks allow to perform with accurate metabolites assignment and quantification.Various actual challenges are discussed, such as proper metabolomics, deep learning architectures and model accuracy.

Detecting Dye-Contaminated Vegetables Using Low-Field NMR Relaxometry

Dyeing vegetables with harmful compounds has become an alarming public health issue over the past few years. Excessive consumption of these dyed vegetables can cause severe health hazards, including cancer. Copper sulfate, malachite green, and Sudan red are some of the non-food-grade dyes widely used on vegetables by untrusted entities in the food supply chain to make them look fresh and vibrant. In this study, the presence and quantity of dye-based adulteration in vegetables are determined by applying ¹H-nuclear magnetic resonance (NMR) relaxometry. The proposed technique was validated by treating some vegetables in-house with different dyes and then soaking them in various solvents. The resulting solutions were collected and analyzed using NMR relaxometry. Specifically, the effective transverse relaxation time constant, T_2,eff, of each solution was estimated using a Carr–Purcell–Meiboom–Gill (CPMG) pulse sequence. Finally, the estimated time constants (i.e., measured signatures) were compared with a library of existing T_2,eff data to detect and quantify the presence of unwanted dyes. The latter consists of data-driven models of transverse decay times for various concentrations of each water-soluble dye. The time required to analyze each sample using the proposed approach is dye-dependent but typically no longer than a few minutes. The analysis results can be used to generate warning flags if the detected dye concentrations violate widely accepted standards for food dyes. The proposed low-cost detection approach can be used in various stages of a produce supply chain, including consumer household.

An easily reproducible, hand-held, single-sided, MRI sensor

Single-sided MRI sensors allow the imaging of samples that are larger than the magnet. Thus, they enable truly portable imagers with potential applications in medicine, quality assurance (QA), agriculture, material science, and other fields. However, despite recent advancements, single-sided MRI systems are relatively uncommon. This is partially due to the limited number of commercial products. Also, current implementations often require large and/or complex magnet arrays which require machining techniques such as milling or drilling. These techniques must be performed to tight tolerances to ensure accuracy of the B₀ field. Furthermore, these systems generally have hand-wound RF or gradient coils that are not trivial to construct. The main goals of this work are to reduce the size of single-sided MRI sensors while simultaneously making them more accessible for others to build. To this end, we present a hand-held, single-sided, MRI sensor that is constructed using an easy-to-assemble magnet array, a 3D-printed housing, and printed circuit boards (PCBs) that contain the RF coil, gradient coils, and matching network. By implementing all coils directly on PCBs, the geometry can be easily optimized and then manufactured at low cost. Both spin density-weighted and T₁-weighted images of various samples are presented to demonstrate the capabilities of the proposed sensor.

Low-field and benchtop NMR

NMR started at low field. Important discoveries like the first observation of NMR in condensed matter, the spin echo, NMR for chemical analysis, Fourier NMR spectroscopy, 2D NMR spectroscopy and magnetic resonance imaging happened at field strengths considered low today. With time the footprint of the NMR instruments at these field strengths shrunk from the laboratory floor to the tabletop. The first commercial tabletop NMR instruments were compact relaxometers for food analysis followed by mobile relaxometers for materials testing and oil-well exploration culminating in tabletop spectrometers for chemical analysis, capable of performing nearly the whole methodical portfolio of today's high-field instruments. The increasing sensitivity afforded by the lower noise of modern electronics and the unfolding richness of hyperpolarization scenarios along with detection schemes alternative to nuclear induction enable NMR at ultra-low field strengths down to zero applied field, where spin-spin couplings in local fields dominate the residual Zeeman interaction. Miniaturization and cost-reduction of NMR instruments outline current development goals along with the development of smart-phone-like apps to conduct standard NMR analyses.

Electropermanent magnets for variable-field NMR

In this work, a dynamically tunable B₀ field is used to perform variable-field NMR. The system consists of an array of electropermanent AlNiCo-5 magnets whose magnetizations are individually programmed using pulse-power control. This design allows the field strength to be varied for field-dispersion measurements. An ultra-broadband front-end is utilized that maintains efficient power transmission over a broad range of frequencies for robust operation without probe tuning. We perform T₁-T₂ correlation measurements at various B₀ field strengths (0.5-2 MHz) and demonstrate discrimination of different dairy products. We observe variation in the frequency dependence of the proton spin-lattice relaxation for the different products as a function of the degree of protein hydration. This variable-field technique provides a low-cost alternative to fast field-cycling NMR and could open possibilities for novel contrast measurements and spatial encoding in magnetic resonance imaging.