Rapid and Simultaneous Measurement of Fat and Moisture Contents in Pork by Low-Field Nuclear Magnetic Resonance

In order to improve the efficiency of Soxhlet extraction and oven drying, low-field nuclear magnetic resonance (LF-NMR) technology was used to detect fat and moisture contents in pork. The transverse relaxation time (T2) distribution curves were constructed by Carr–Purcell–Meiboom–Gill (CPMG) experiments. In addition, the optimal conditions of adding MnCl2 aqueous solution was explored to separate water and fat signal peaks. Finally, the reliability of this method for the determination of fat and moisture contents in pork was verified. The present study showed that adding 1.5 mL of 20% MnCl2 aqueous solution solution at 50 °C can isolate and obtain a stable peak of fat. The lard and 0.85% MnCl2 aqueous solution were used as the standards for fat and moisture measurements, respectively, and calibration curves with R2 = 0.9999 were obtained. In addition, the repeatability and reproducibility of this method were 1.71~3.10%. There was a significant correlation (p < 0.05) between the LF-NMR method and the conventional methods (Soxhlet extraction and oven drying), and the R2 was 0.9987 and 0.9207 for fat and moisture, respectively. All the results proved that LF-NMR could determine fat and moisture contents in pork rapidly and simultaneously.


Introduction
Pork is rich in many nutrients, of which moisture and fat are not only the nutritional components but also the important indicators for evaluating the eating quality [1][2][3] and the safety quality of pork. Fat is one of the important nutrients, but excessive fat intake can cause adverse health effects. In addition, the content of fat directly affects the flavor, tenderness, and juiciness of the meat. Many biochemical reactions are closely related to water content, and the content of water also directly affects the quality of meat. Some studies showed that higher moisture levels could lead to more salmonella in meat [4]. Now, Soxhlet extraction and oven drying are conventional methods for the determination of fat and moisture contents, respectively. These methods are time-consuming and laborious, and large amounts of hazard organic reagents, such as ether or petroleum ether, are used in the evaluation process [5]. In order to avoid these short-comings, rapid non-destructive methodologies such as near-infrared spectroscopy [6,7], Raman spectroscopy [8,9], and hyperspectral imaging [10,11] technologies are used to evaluate pork quality. These technologies can accurately measure the moisture and fat contents of pork online, but stoichiometry methods and model building need to be applied for predictions with abundant samples. Moreover, the complicated data processing and high equipment cost are not conducive to popularization and application. Therefore, it is essential to establish a new method for the simultaneous and accurate determination of fat and moisture contents with high efficiency and low pollution.
In recent years, low-field nuclear magnetic resonance (LF-NMR) has been widely used to analyze and quantify the moisture and fat of fruits [12], grain [13], aquatic products [14],

Fat and Moisture Contents Measurement
The Soxhlet extraction method (AOAC Official Method 991.36) and oven drying method (AOAC Official Method 950.46) were used to measure fat and moisture contents in pork samples, respectively, as the reference values.

LF-NMR Relaxation Measurements
The pork samples and standard samples were measured using an LF-NMR analyzer (NMI20-40V-I, Suzhou Niumag Analytical Instrument Co., Ltd., Suzhou, China), with a proton resonance of 20 MHz at 32 • C and a sample tube of 25 mm in diameter. The spin-spin transverse relaxation was measured using a sequence based on the Carr-Purcell-Meiboom -Gill (CPMG) sequence with the following parameters: 90 • pulse width (P1) = 5.52 µs, 180 • pulse width (P2) = 12.48 µs, waiting time (TW) = 2000 ms, and number of scans (NS) = 32. The T 2 distribution curves were acquired by multi-exponential fitting with the simultaneous iterative reconstruction technique (SIRT) algorithm. The relaxation times, peak areas, and proportions of peak areas were recorded.

Effect of Echo Time (TE) on LF-NMR Relaxation
An amount of 4.5 mL of 20% MnCl 2 aqueous solution was added into 2.0 g of minced meat and mixed, and then the sample was heated at 50 • C in a metal bath for 30 min. LF-NMR measurements were performed at different TEs of 0.1, 0.15, 0.20, 0.25, 0.30, and 0.35 ms.

Effect of Number of Echoes (NS) on LF-NMR Relaxation
An amount of 1.5 mL of 20% MnCl 2 aqueous solution was added into 2.0 g of minced meat and mixed, and then the sample was heated at 50 • C in a metal bath for 30 min. LF-NMR measurements were taken at different NSs of 16, 32, 64, and 128 times.

Statistical Analysis
All experiments were carried out in triplicate. The statistical analyses were performed with SPSS 25.0 (SPSS Inc., Chicago, IL, USA). The differences among samples were analyzed by analysis of variance (ANOVA) and Duncan's multiple range tests. Differences at p < 0.05 were considered significant. The results were presented as the mean ± standard deviation. Figures were created using Origin 2021 (OriginLab, Corporation, Northampton, MA, USA).

Effect of MnCl 2 Aqueous Solution on T 2 Distribution Curves of Lard and Pork Meat
The peak of the T 2 curve within the relaxation time of 36-843 ms was ascribed to fat, whose proportions of the peak areas were between 87.37-98.92%. The iterations with different numbers were selected in multi-exponential fitting [24], but these peaks with different iteration numbers did not show obvious bimodal distribution as in other studies [25]. After adding 40% MnCl 2 aqueous solution, the T 2 curves were almost the same as those before adding the solution, and the peak areas of fat did not change significantly (p > 0.05). Since manganese salts are insoluble in hydrocarbon compounds, the addition of MnCl 2 aqueous solution had no effect on the relaxation characteristics of hydrogen protons in the oil phase [26,27]. The T 2 curve of pork was expressed by the black blocks in Figure 1. Three discontinuous peaks from left to right were ascribed to bound water (T 21 ), immobile water (T 22 ), and free water (T 23 ), respectively, which was in accordance with other studies on the distribution of T 2 in meat [15,28]. However, the peaks of fat and moisture were superimposed, and it was also impossible to be distinguished by T 2 curves [29].

Effect of MnCl2 Aqueous Solution on T2 Distribution Curves of Lard and Pork Meat
The peak of the T2 curve within the relaxation time of 36-843 ms was ascribed to fat, whose proportions of the peak areas were between 87.37-98.92%. The iterations with different numbers were selected in multi-exponential fitting [24], but these peaks with different iteration numbers did not show obvious bimodal distribution as in other studies [25]. After adding 40% MnCl2 aqueous solution, the T2 curves were almost the same as those before adding the solution, and the peak areas of fat did not change significantly (p > 0.05). Since manganese salts are insoluble in hydrocarbon compounds, the addition of MnCl2 aqueous solution had no effect on the relaxation characteristics of hydrogen protons in the oil phase [26,27]. The T2 curve of pork was expressed by the black blocks in Figure 1. Three discontinuous peaks from left to right were ascribed to bound water (T21), immobile water (T22), and free water (T23), respectively, which was in accordance with other studies on the distribution of T2 in meat [15,28]. However, the peaks of fat and moisture were superimposed, and it was also impossible to be distinguished by T2 curves [29]. The T2 curve of pork with Mn 2+ was represented by the gray dotted line in Figure 1. Two peaks, T21 and T22, were attributed to the hydrogen protons of water and fat, respectively. It can be seen that the T23 disappeared, T22 decreased greatly, and T21 increased, compared with the T2 curve of pork. Mn 2+ can permeate into the sample through ion transport, resulting in direct contact between Mn 2+ and the hydrogen protons of water. In addition, there was a strong spin-spin interaction between the spin magnetic moments of hydrogen protons and those of electrons in ions, which accelerated the decay of hydrogen protons in water [30,31]. Hence, the peaks of immobile water and free water decreased or even disappeared. Meanwhile, the hydrogen protons introduced by the MnCl2 aqueous solution enhanced the T21 peak area. Compared with the T2 curve of lard, the distribution of T22 was basically the same. It proved that adding the MnCl2 aqueous solution was feasible to solve the overlapping problem of moisture and fat signal peaks in pork. This method has been used to distinguish water-oil signals in crude oil [23]. The T 2 curve of pork with Mn 2+ was represented by the gray dotted line in Figure 1. Two peaks, T 21 and T 22 , were attributed to the hydrogen protons of water and fat, respectively. It can be seen that the T 23 disappeared, T 22 decreased greatly, and T 21 increased, compared with the T 2 curve of pork. Mn 2+ can permeate into the sample through ion transport, resulting in direct contact between Mn 2+ and the hydrogen protons of water. In addition, there was a strong spin-spin interaction between the spin magnetic moments of hydrogen protons and those of electrons in ions, which accelerated the decay of hydrogen protons in water [30,31]. Hence, the peaks of immobile water and free water decreased or even disappeared. Meanwhile, the hydrogen protons introduced by the MnCl 2 aqueous solution enhanced the T 21 peak area. Compared with the T 2 curve of lard, the distribution of T 22 was basically the same. It proved that adding the MnCl 2 aqueous solution was feasible to solve the overlapping problem of moisture and fat signal peaks in pork. This method has been used to distinguish water-oil signals in crude oil [23]. With the increase in concentration, the T 21 moved to the left, and the peak area was reduced gradually. When the concentration reached 10%, the distribution of T 21 did not change significantly, but the peak area still gradually decreased. It was reported that the amount of paramagnetic substance had a great influence on spin relaxation [16]. The higher the concentration of Mn 2+ , the faster the decay of hydrogen protons in water, and the greater the relaxation signals loss [26]. Moreover, the T 22 also manifested a significant left shift with the increase in MnCl 2 aqueous solution concentration. When the concentration was lower than 10%, the starting relaxation time of T 22 was 43.354 ms, which was significantly different from that of lard. When concentration was 20%, the shape of the peak and distribution of T 22 were similar to that of lard. As the concentration continued to increase, the initial relaxation time of T 22 phased down, and the shape of the peak gradually became wider and shorter. However, the peak area did not change significantly. While the concentration was 20%, the stability of the T 22 peak area was good, and the relative standard deviation (RSD) was only 2.38%. It was suggested that 20% MnCl 2 aqueous solution could separate water and fat signals better. manifested a significant left shift with the increase in MnCl2 aqueous solution concentration. When the concentration was lower than 10%, the starting relaxation time of T22 was 43.354 ms, which was significantly different from that of lard. When concentration was 20%, the shape of the peak and distribution of T22 were similar to that of lard. As the concentration continued to increase, the initial relaxation time of T22 phased down, and the shape of the peak gradually became wider and shorter. However, the peak area did not change significantly. While the concentration was 20%, the stability of the T22 peak area was good, and the relative standard deviation (RSD) was only 2.38%. It was suggested that 20% MnCl2 aqueous solution could separate water and fat signals better.

MnCl2 Aqueous Solution Volume
The solution volume directly affects the quantity of Mn 2+ . Figure 2b illustrates T2 curves corresponding to the adding of different volumes of MnCl2 aqueous solutions. The peak area of T21 decreased with the augment of volume, and the shape of the T22 peak gradually became shorter and wider. The results were consistent with the influence of MnCl2 aqueous solution concentration. The distribution and peak area of T22 did not change significantly when the volume was more than 1.5 mL, and the peak area was significantly higher than the areas of the samples treated with 0.5 or 1.0 mL MnCl2 aqueous solution. Furthermore, the RSD of the T22 peak area decreased from 5.05% to 2.18%, indicating that the stability of the peak area of high-volume (≥1.5 mL) treated groups was better than low-volume (<1.5 mL) treated groups. Although the proportion of T21 reached 97.06% when adding a small volume addition (0.5 mL), there was a large deviation of T22 peak in multi-exponential fitting, which could lead to the uneven distribution of Mn 2+ in the sample. Therefore, the solution volume was determined to be 1.5 mL, and then a stable peak area of fat was able to be obtained.

Heat Treatment Temperature
Temperature is an important factor affecting the relaxation characteristics of samples. As shown in Figure 2c, the peak area of the T21 increased with the increase in temperature, and the T22 shifted to the right with the rise in temperature. At 40 ℃, the T22 peak area was significantly higher than that of other temperatures (p < 0.05). Temperature can affect internal molecular structure and phase state. Lard was solid at room temperature, and it gradually melted into liquid after heating. In this process, the arrangement of the internal molecules of fat changed from order to disorder, and the binding force between molecules was weakened, resulting in a prolonged relaxation time. As the temperature continued to rise, the thermal movement of the system intensified, and the molecular fluidity increased [32], which led to the extension of T2. Lard can be completely melted at approximately

MnCl 2 Aqueous Solution Volume
The solution volume directly affects the quantity of Mn 2+ . Figure 2b illustrates T 2 curves corresponding to the adding of different volumes of MnCl 2 aqueous solutions. The peak area of T 21 decreased with the augment of volume, and the shape of the T 22 peak gradually became shorter and wider. The results were consistent with the influence of MnCl 2 aqueous solution concentration. The distribution and peak area of T 22 did not change significantly when the volume was more than 1.5 mL, and the peak area was significantly higher than the areas of the samples treated with 0.5 or 1.0 mL MnCl 2 aqueous solution. Furthermore, the RSD of the T 22 peak area decreased from 5.05% to 2.18%, indicating that the stability of the peak area of high-volume (≥1.5 mL) treated groups was better than low-volume (<1.5 mL) treated groups. Although the proportion of T 21 reached 97.06% when adding a small volume addition (0.5 mL), there was a large deviation of T 22 peak in multi-exponential fitting, which could lead to the uneven distribution of Mn 2+ in the sample. Therefore, the solution volume was determined to be 1.5 mL, and then a stable peak area of fat was able to be obtained.

Heat Treatment Temperature
Temperature is an important factor affecting the relaxation characteristics of samples. As shown in Figure 2c, the peak area of the T 21 increased with the increase in temperature, and the T 22 shifted to the right with the rise in temperature. At 40 • C, the T 22 peak area was significantly higher than that of other temperatures (p < 0.05). Temperature can affect internal molecular structure and phase state. Lard was solid at room temperature, and it gradually melted into liquid after heating. In this process, the arrangement of the internal molecules of fat changed from order to disorder, and the binding force between molecules was weakened, resulting in a prolonged relaxation time. As the temperature continued to rise, the thermal movement of the system intensified, and the molecular fluidity increased [32], which led to the extension of T 2 . Lard can be completely melted at approximately 48.4 • C. Sørland et al. [17] heated minced meat to 70 • C to ensure that the fat phase was liquid. While vegetable oil existed in a liquid form at room temperature, generally, the sample temperature of vegetable oil was kept consistent with that of the magnet (32 • C) [33,34] in order to avoid the presence of both solid and liquid fat in the sample. A temperature of 50 • C was chosen as the sample temperature, at which the stability (RSD = 3.52%) of the fat peak area was good.
3.3. Effect of Performance Parameters on LF-NMR Relaxation 3.3.1. Effect of TE on LF-NMR Relaxation TE, as the interval time between echoes and should be quite short to minimize the diffusion effects. Figure 3 shows the T 2 distribution curves of the samples detected at different TEs, where T 21 represents water and T 22 represents fat. The increase in TE caused T 21 to move to the right, and the peak area obviously decreased. However, the T 2 and peak area of T 22 did not change significantly. This can be explained by the fact that increasing TE will cause molecular diffusion, which will lead to prolonged T 2 [35]. Moreover, the diffusion of water is stronger than that of fat, and the hydrogen proton decay rate of moisture in Mn 2+ -treated samples is faster. Therefore, the relaxation behavior of T 21 changed significantly, but the relaxation behavior of T 22 did not. temperature of 50 °C was chosen as the sample temperature, at which the stability (RSD = 3.52%) of the fat peak area was good.

Effect of Performance Parameters on LF-NMR Relaxation
3.3.1. Effect of TE on LF-NMR Relaxation TE, as the interval time between echoes and should be quite short to minimize the diffusion effects. Figure 3 shows the T2 distribution curves of the samples detected at different TEs, where T21 represents water and T22 represents fat. The increase in TE caused T21 to move to the right, and the peak area obviously decreased. However, the T2 and peak area of T22 did not change significantly. This can be explained by the fact that increasing TE will cause molecular diffusion, which will lead to prolonged T2 [35]. Moreover, the diffusion of water is stronger than that of fat, and the hydrogen proton decay rate of moisture in Mn 2+ -treated samples is faster. Therefore, the relaxation behavior of T21 changed significantly, but the relaxation behavior of T22 did not.
The variation of TE within 0.10~0.35 ms affected the attenuation degree and relaxation interval of CPMG echo decay curves. When TE increased from 0.10 ms to 0.30 ms, the signal amplitude decreased from 257.42 to 57.31. As the TE continued to increase, the signal amplitude did not change significantly, which indicated that the signal had been completely attenuated. In addition, when TE was 0.30 ms, the RSD of the fat peak area was 3.06%. Therefore, in this study, TE was chosen as 0.30 ms.  Figure 4a shows the influence of different NSs on T2 distribution curves. With the increase in NS, the areas of the two relaxation peaks of T21 and T22 gradually increased. The peak area of T22 and NS had a good linear relationship. The equation was y = 64.41x-5.0905, and R2 was 0.9999. The NS was proportional to the signal noise ratio (SNR), as shown in Figure 4b. When the NS was set from 16 to 128, the SNR increased from 184.82 to 290.44. There was no significant difference in SNR when NS was 32 or 64 times. However, if NS is further increased, the time of signal acquisition will increase. Meanwhile, the continuous excitation of a RF pulse will increase the sample temperature, which will affect the relaxation characteristics of the samples [36]. Therefore, the NS was set to 32 times, and the areas of T22 showed good stability (RSD = 1.28%). The variation of TE within 0.10~0.35 ms affected the attenuation degree and relaxation interval of CPMG echo decay curves. When TE increased from 0.10 ms to 0.30 ms, the signal amplitude decreased from 257.42 to 57.31. As the TE continued to increase, the signal amplitude did not change significantly, which indicated that the signal had been completely attenuated. In addition, when TE was 0.30 ms, the RSD of the fat peak area was 3.06%. Therefore, in this study, TE was chosen as 0.30 ms. Figure 4a shows the influence of different NSs on T 2 distribution curves. With the increase in NS, the areas of the two relaxation peaks of T 21 and T 22 gradually increased. The peak area of T 22 and NS had a good linear relationship. The equation was y = 64.41x − 5.0905, and R2 was 0.9999. The NS was proportional to the signal noise ratio (SNR), as shown in Figure 4b. When the NS was set from 16 to 128, the SNR increased from 184.82 to 290.44. There was no significant difference in SNR when NS was 32 or 64 times. However, if NS is further increased, the time of signal acquisition will increase. Meanwhile, the continuous excitation of a RF pulse will increase the sample temperature, which will affect the relaxation characteristics of the samples [36]. Therefore, the NS was set to 32 times, and the areas of T 22 showed good stability (RSD = 1.28%).  Columns marked with different letters are significantly different according to Duncan's multiple range test at p < 0.05

Calibration Curves of Fat
In this study, lard was used as the standard sample to measure the fat content of pork. The T2 distribution curves of different weights (0.0114-0.6413 g) of lard were measured. As shown in Figure 5, the T22 peak areas were enlarged linearly with the increased weight of lard over a range of 0.0114-0.6413 g (equivalent to 0.55-39.99% of the fat content of 2 g of minced pork). There was a significant positive correlation between lard weight and T22 peak areas (p < 0.05), and the regression equation was: y = 34,313x + 71.172 (R 2 = 0.9999). The signal generation in LF-NMR was related to the amount of hydrogen protons, which positively correlated with the lipid or moisture content in the samples [13,37]. We calculated the hydrogen proton density of lard by measuring fat peak areas, and then we measured the fat contents.

Calibration Curves of Moisture
It was reported that using deionized water as a standard sample would result in greatly reduced detection efficiency, due to the excessively long relaxation time of water

Calibration Curves of Fat
In this study, lard was used as the standard sample to measure the fat content of pork. The T 2 distribution curves of different weights (0.0114-0.6413 g) of lard were measured. As shown in Figure 5, the T 22 peak areas were enlarged linearly with the increased weight of lard over a range of 0.0114-0.6413 g (equivalent to 0.55-39.99% of the fat content of 2 g of minced pork). There was a significant positive correlation between lard weight and T 22 peak areas (p < 0.05), and the regression equation was: y = 34,313x + 71.172 (R 2 = 0.9999). The signal generation in LF-NMR was related to the amount of hydrogen protons, which positively correlated with the lipid or moisture content in the samples [13,37]. We calculated the hydrogen proton density of lard by measuring fat peak areas, and then we measured the fat contents. Columns marked with different letters are significantly different according to Duncan's multiple range test at p < 0.05

Calibration Curves of Fat
In this study, lard was used as the standard sample to measure the fat content of pork. The T2 distribution curves of different weights (0.0114-0.6413 g) of lard were measured. As shown in Figure 5, the T22 peak areas were enlarged linearly with the increased weight of lard over a range of 0.0114-0.6413 g (equivalent to 0.55-39.99% of the fat content of 2 g of minced pork). There was a significant positive correlation between lard weight and T22 peak areas (p < 0.05), and the regression equation was: y = 34,313x + 71.172 (R 2 = 0.9999). The signal generation in LF-NMR was related to the amount of hydrogen protons, which positively correlated with the lipid or moisture content in the samples [13,37]. We calculated the hydrogen proton density of lard by measuring fat peak areas, and then we measured the fat contents.

Calibration Curves of Moisture
It was reported that using deionized water as a standard sample would result in greatly reduced detection efficiency, due to the excessively long relaxation time of water

Calibration Curves of Moisture
It was reported that using deionized water as a standard sample would result in greatly reduced detection efficiency, due to the excessively long relaxation time of water [23]. As a result, MnCl 2 aqueous solution [38], a rehydratable freeze-dried meat cube [39], and CuSO 4 aqueous solution [40] were used as standard samples for water quantification to improve detection efficiency. In this study, the MnCl 2 aqueous solution was used as the standard sample to measure moisture. The fat peak area had been available from the T 2 distribution curve of samples treated with 20% MnCl 2 aqueous solution. Then, the moisture peak area was calculated using the following equation: where Aw is the moisture peak area, A is the total peak area of minced fresh pork, Mf is the mass of the minced pork that was treated with 20% MnCl 2 aqueous solution, M is the mass of minced fresh pork, and Af is the fat peak area. The calculated hydrogen proton density of moisture in pork was 33,501. Figure 6a shows the correlation curve between the concentration of the MnCl 2 aqueous solution and the peak area of moisture. With the increase in concentration, the peak areas decreased exponentially. According to the correlation curve in Figure 6a, 0.85% MnCl 2 aqueous solution was required to simulate the hydrogen proton density of moisture in pork. Therefore, this solution was applied as the standard sample to construct the calibration curve of moisture. [23]. As a result, MnCl2 aqueous solution [38], a rehydratable freeze-dried meat cube [39], and CuSO4 aqueous solution [40] were used as standard samples for water quantification to improve detection efficiency. In this study, the MnCl2 aqueous solution was used as the standard sample to measure moisture. The fat peak area had been available from the T2 distribution curve of samples treated with 20% MnCl2 aqueous solution. Then, the moisture peak area was calculated using the following equation: where Aw is the moisture peak area, A is the total peak area of minced fresh pork, Mf is the mass of the minced pork that was treated with 20% MnCl2 aqueous solution, M is the mass of minced fresh pork, and Af is the fat peak area. The calculated hydrogen proton density of moisture in pork was 33501. Figure 6a shows the correlation curve between the concentration of the MnCl2 aqueous solution and the peak area of moisture. With the increase in concentration, the peak areas decreased exponentially. According to the correlation curve in Figure 6a, 0.85% MnCl2 aqueous solution was required to simulate the hydrogen proton density of moisture in pork. Therefore, this solution was applied as the standard sample to construct the calibration curve of moisture. The T 2 distribution curves of different weights (0.4958-2.9867 g) of 0.85% MnCl 2 aqueous solution were measured and are shown in Figure 6b. Only one peak (T 21 ) appeared before 10 ms, and its peak point relaxation time was 0.73 ms. As the mass of solution increased, the peak area multiplied at the same rate. The solution weight was highly linear with the moisture peak area, and the linear equation was: y = 34,467.93x + 98.647 (R 2 = 1.00). The T 2 distribution of the standard sample (0.85% MnCl 2 aqueous solution) was different from that of moisture in pork due to the complex states that existed of moisture in pork. However, it was suggested that the change of relaxation time did not affect the hydrogen proton density of water by comparing five different kinds of meat of livestock and poultry [41]. Therefore, it was feasible to use 0.85% MnCl 2 aqueous solution as the standard sample.

Method Validation
The repeatability and reproducibility of three different samples (n = 6) were verified using the above method. The T 2 distribution curves of pork with different fat contents were shown in Figure 7. The fat peak (T 22 ) was basically coincident, and peak areas were stable. The RSDs of peak area were 2.04%, 1.57%, and 1.71%, respectively ( Table 1).
The T2 distribution curves of different weights (0.4958-2.9867 g) of 0.85% MnCl2 aqueous solution were measured and are shown in Figure 6b. Only one peak (T21) appeared before 10 ms, and its peak point relaxation time was 0.73 ms. As the mass of solution increased, the peak area multiplied at the same rate. The solution weight was highly linear with the moisture peak area, and the linear equation was: y = 34467.93x + 98.647 (R 2 = 1.00). The T2 distribution of the standard sample (0.85% MnCl2 aqueous solution) was different from that of moisture in pork due to the complex states that existed of moisture in pork. However, it was suggested that the change of relaxation time did not affect the hydrogen proton density of water by comparing five different kinds of meat of livestock and poultry [41]. Therefore, it was feasible to use 0.85% MnCl2 aqueous solution as the standard sample.

Method Validation
The repeatability and reproducibility of three different samples (n = 6) were verified using the above method. The T2 distribution curves of pork with different fat contents were shown in Figure 7. The fat peak (T22) was basically coincident, and peak areas were stable. The RSDs of peak area were 2.04%, 1.57%, and 1.71%, respectively (Table 1).  Then, the intra-day and inter-day reproducibility of the method were validated in the morning, at noon, and in the evening on the same day and different days (day 1, day 3, and day 5). As shown in Table 2, the RSDs were 3.03% and 3.10%, respectively. The above results suggested that the method had good precision.   Then, the intra-day and inter-day reproducibility of the method were validated in the morning, at noon, and in the evening on the same day and different days (day 1, day 3, and day 5). As shown in Table 2, the RSDs were 3.03% and 3.10%, respectively. The above results suggested that the method had good precision.

Determination of Fat and Moisture Content in Pork
The fat and moisture contents of 17 pork samples were determined using the Soxhlet extraction method and the oven drying method, respectively, with reference values contained at 1.94-36.61% and 47.33-73.02%. The peak areas of these samples were tested by LF-NMR, and then fat and moisture contents were calculated. Figure 8 shows the comparison of fat and moisture contents between the conventional method and the LF-NMR method, and the linear regression equation and R 2 are exhibited. There was no significant difference in results between the two methods (p > 0.05). It can be observed that the accuracy of fat content (R 2 = 0.9987) was better than moisture content (R 2 = 0.9207) when using the LF-NMR method. The fat content measured by conventional method was higher than that measured by the LF-NMR method because the conventional method required prolonged heating, which led to the increase in fatty acid polarity and significantly increased fat weight, manifested as the increase in the peak area of fat [42]. Samples that were pretreated with MnCl 2 aqueous solution were not affected by high temperature, and the measured values of fat by LF-NMR were more accurate.
the LF-NMR method. The fat content measured by conventional method was higher than that measured by the LF-NMR method because the conventional method required prolonged heating, which led to the increase in fatty acid polarity and significantly increased fat weight, manifested as the increase in the peak area of fat [42]. Samples that were pretreated with MnCl2 aqueous solution were not affected by high temperature, and the measured values of fat by LF-NMR were more accurate.
However, the moisture content measured by the LF-NMR method was lower than the oven drying method. It could be attributed to the high sample temperature (50 °C) in our study, which resulted in the loss of partial moisture relaxation signals of fresh pork [43]. Han et al. reported that the total peak area of the samples decreased, and the peaks of immobile water moved left with the rise in temperature [44]. The binding force of moisture in different samples was varied, which led to a certain deviation in the determination of moisture content [45,46]. Therefore, compared with the conventional methods, the results of the LF-NMR method were more accurate, and the detection time was shortened to less than 3 min for a single sample.

Conclusions
In this study, LF-NMR was used to determine fat and moisture contents. In particular, the problem of overlapping signal peaks of moisture and fat in pork was solved by adding 20% MnCl2 aqueous solution with a volume of 1.5 mL at 50 °C. Lard and 0.85% MnCl2 aqueous solution were used as standard samples for determining fat and moisture contents, respectively. Furthermore, the results of the LF-NMR method showed a good correlation with conventional methods, with R 2 being 0.9987 (fat) and 0.9207 (moisture), However, the moisture content measured by the LF-NMR method was lower than the oven drying method. It could be attributed to the high sample temperature (50 • C) in our study, which resulted in the loss of partial moisture relaxation signals of fresh pork [43]. Han et al. reported that the total peak area of the samples decreased, and the peaks of immobile water moved left with the rise in temperature [44]. The binding force of moisture in different samples was varied, which led to a certain deviation in the determination of moisture content [45,46]. Therefore, compared with the conventional methods, the results of the LF-NMR method were more accurate, and the detection time was shortened to less than 3 min for a single sample.

Conclusions
In this study, LF-NMR was used to determine fat and moisture contents. In particular, the problem of overlapping signal peaks of moisture and fat in pork was solved by adding 20% MnCl 2 aqueous solution with a volume of 1.5 mL at 50 • C. Lard and 0.85% MnCl 2 aqueous solution were used as standard samples for determining fat and moisture contents, respectively. Furthermore, the results of the LF-NMR method showed a good correlation with conventional methods, with R 2 being 0.9987 (fat) and 0.9207 (moisture), respectively. The present study indicated that LF-NMR could detect the fat and moisture contents of pork simultaneously and accurately.

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author.