Dual-domain mean-reverting diffusion model-enhanced temporal compressive coherent diffraction imaging.

Temporal compressive coherent diffraction imaging is a lensless imaging technique with the capability to capture fast-moving small objects. However, the accuracy of imaging reconstruction is often hindered by the loss of frequency domain information, a critical factor limiting the quality of the reconstructed images. To improve the quality of these reconstructed images, a method dual-domain mean-reverting diffusion model-enhanced temporal compressive coherent diffraction imaging (DMDTC) has been introduced. DMDTC leverages the mean-reverting diffusion model to acquire prior information in both frequency and spatial domain through sample learning. The frequency domain mean-reverting diffusion model is employed to recover missing information, while hybrid input-output algorithm is carried out to reconstruct the spatial domain image. The spatial domain mean-reverting diffusion model is utilized for denoising and image restoration. DMDTC has demonstrated a significant enhancement in the quality of the reconstructed images. The results indicate that the structural similarity and peak signal-to-noise ratio of images reconstructed by DMDTC surpass those obtained through conventional methods. DMDTC enables high temporal frame rates and high spatial resolution in coherent diffraction imaging.

A Combined UWB/IMU Localization Method with Improved CKF.

Aiming at the problem that ultra-wide band (UWB) cannot be accurately localized in environments with large noise variations and unknown statistical properties, a combinatorial localization method based on improved cubature (CKF) is proposed. First, in order to overcome the problem of inaccurate local approximation or even the inability to converge due to the initial value not being set near the optimal solution in the process of solving the UWB position by the least-squares method, the Levenberg-Marquardt algorithm (L-M) is adopted to optimally solve the UWB position. Secondly, because UWB and IMU information are centrally fused, an adaptive factor is introduced to update the measurement noise covariance matrix in real time to update the observation noise, and the fading factor is added to suppress the filtering divergence to achieve an improvement for the traditional CKF algorithm. Finally, the performance of the proposed combined localization method is verified by field experiments in line-of-sight (LOS) and non-line-of-sight (NLOS) scenarios, respectively. The results show that the proposed method can maintain high localization accuracy in both LOS and NLOS scenarios. Compared with the Extended Kalman filter (EKF), unbiased Kalman filter (UKF), and CKF algorithms, the localization accuracies of the proposed method in NLOS scenarios are improved by 25.2%, 18.3%, and 11.3%, respectively.

Er3+/Yb3+ Co-Doped Fluorotellurite Glass Fiber with Broadband Luminescence.

In order to address the 'capacity crisis' caused by the narrow bandwidth of the current C band and the demand for wide-spectrum sensing sources and tunable fiber lasers, a broadband luminescence covering the C + L bands using Er3+/Yb3+ co-doped fluorotellurite glass fiber is investigated in this paper. The optimal doping concentrations in the glass host were determined based on the intensity, lifetime, and full width at half maximum (FWHM) of the fluorescence centered at 1.5 µm, which were found to be 1.5 mol% Er2O3 and 3 mol% Yb2O3. We also systematically investigated this in terms of optical absorption spectra, absorption and emission cross-sections, gain coefficients, Judd-Ofelt parameters, and up-conversion fluorescence. The energy transfer (ET) mechanism between the high concentrations of Er3+ and Yb3+ was summarized. In addition, a step-indexed fiber was prepared based on these fluorotellurite glasses, and a wide bandwidth of ~112.5 nm (covering the C + L bands from 1505.1 to 1617.6 nm) at 3 dB for the amplified spontaneous emission (ASE) spectra has been observed at a fiber length of 0.57 m, which is the widest bandwidth among all the reports based on tellurite glass. Therefore, this kind of Er3+/Yb3+ co-doped fluorotellurite glass fiber has great potential for developing broadband C + L band amplifiers, ultra-wide fiber sources for sensing, and tunable fiber lasers.

[Study on the total flavonoid content and antioxidant activity of Morus nigra Linn. from different origins in Xinjiang].

OBJECTIVE: Optimization of the extraction process of total flavonoids from Morus nigra Linn. based on response surface design, to compare the differences in total flavonoid content of Morus nigra Linn. from different origins in Xinjiang and its relationship with antioxidant activity. METHODS: A one-way test was used to investigate the effects of ethanol volume fraction, material-liquid ratio, ultrasonic power and ultrasonic time on total flavonoid content analysis of Morus nigra Linn. , Box-Behnken response surface design optimisation was used to derive the optimal extraction process parameters. Meanwhile, the in vitro antioxidant activity of Morus nigra Linn. was evaluated by antioxidant activity indexes such as hydroxyl radical scavenging capacity and ferric ion reducing capacity. RESULTS: The optimal extraction condition was 45% ethanol by volume, 1∶20 g/mL material-liquid ratio, 300W ultrasound power and 60 min ultrasound time. The total flavonoid content of Morus nigra Linn. s from six origins was detected and the result showed that Kuche >Kashgar >Kuche Wuqia Town>Hetian >Ying jisha >Yutian, in which the total flavonoid content of mulberry in Kucha city was(27.159±0.091)mg/g. The difference in the total flavonoid content of Morus nigra Linn. s in different origicns was statistically significant(P<0.05). In vitro antioxidant activity analysis showed that the antioxidant activity of mulberry from Kuche and Hetian was stronger, and the hydroxyl radical scavenging capacity, ferric ion reducing capacity, and total antioxidant capacity of mulberry from various origins were statistically significant(P<0.05). The correlation analysis showed that the total flavonoid content of Morus nigra Linn. was correlated with the antioxidant capacity to improve the scavenging of hydroxyl radicals. CONCLUSION: The ultrasonic-assisted method of extracting total flavonoids from Morus nigra Linn. is simple, and the model constructed has a high degree of fit, which can better compare the total flavonoids content of Morus nigra Linn. from different origins in Xinjiang.

[Analysis on vitamin B_1 and vitamin B_2 proficiency testing assessment in China disease control and prevention system].

OBJECTIVE: To evaluate the detection ability of vitamin B_1 and vitamin B_(2 )in rice flour in the laboratories of disease control and prevention system, by conducting the proficiency testing(PT)activity. METHODS: Before the vitamin B_1 and vitamin B_2 quality control samples were distributed to the laboratories of disease control and prevention system, the uniformity and stability of samples were analyzed by one-way ANOVO respectively. High performance liquid chromatography(HPLC) method was required to determine vitamin B_1(GB 5009.84-2016: determination of vitamin B_1 in food, first method as reference). HPLC method was also required to determine vitamin B_2(GB 5009.85-2016: determination of vitamin B_2 in food, first method as reference). Robust statistics analysis of proficiency testing result was conducted to evaluate laboratory testing ability through Z score. RESULTS: A total of 43 laboratories completed the proficiency testing. In all of the laboratories participated in the determination of vitamin B_(1 )and vitamin B_2, the total satisfactory rate of vitamin B_1 was 88.4%, while vitamin B_2 was 86.0%. CONCLUSION: The ability of vitamin B_1 and vitamin B_2 detection in disease control and prevention system in China is better than expected, and the testing ability of a few laboratory needs to be improved.

[Determination of 13 kinds of free and bound phenolic acids in fruits by high-performance liquid chromatography-mass spectrometry].

OBJECTIVE: To establish a high-performance liquid chromatography-mass spectrometry(HPLC-MS/MS) method for detecting 13 kind of free and bound phenolic acids(chlorogenic acid, protocatechuic acid, ferulic acid, p-coumaric acid, gallic acid, gentisic acid, vanillic acid, caffeic acid, syringic acid, sinapic acid, rosmarinic acid, salicylic acid, p-hydroxybenzoic acid) in fruits, and optimize the pre-treatment conditions to meet the detection requirements for phenolic acid content in various types of fruits. METHODS: Free phenolic acids in fruits were extracted using methanol through ultrasonic extraction. Conjugated phenolic acids in the centrifuged residue were released by alkaline hydrolysis and extracted with ethyl acetate. The two extracts were combined, concentrated, and analyzed using HPLC-MS/MS. Separation was achieved using an Agilent ZORBAX SB-C_(18) chromatography column(3.0 mm×100 mm, 3.5 µm), and detection was performed in multiple reaction monitoring(MRM) mode. RESULTS: All 13 standard phenolic acids achieved complete separation within 10 minutes, with linear correlation coefficients greater than 0.998 and detection limits ranging from 0.172 to 3.471 ng/mL. After optimization of the pre-treatment method, the recovery rates of the method for four types of fruits-apples, strawberries, oranges, and peaches-ranged from 80.0% to 119.4%, and the precision were lower than 7.00%(n=6). The result of testing on four categories of twelve types of fruits demonstrated significant variations in the content of phenolic acids among different fruits, and within the same category, the composition of phenolic acids did not exhibit consistency. CONCLUSION: The HPLC-MS/MS method exhibits high sensitivity, precision, and accuracy. It is suitable for the detection of both free and bound phenolic acids in various types of fruits.

[Determination of vitamin A, D and E in dishes by two-dimensional liquid chromatography].

OBJECTIVE: To establish a method for the simultaneous determination of vitamins A, D and E in dishes by two-dimensional liquid chromatography. METHODS: The samples were saponified and extracted with a mixture of ethyl acetate and n-hexane(3∶2, V/V)under the protection of antioxidants, and determined by two-dimensional liquid chromatography. The baseline separation of retinol, α-tocopherol, ß-tocopherol, γ-tocopherol, δ-tocopherol, and α-tocotrienols was achieved by using Alphasil pentafluorophenyl column(PFP, 150 mm×4.6 mm, 5 µm) as the one-dimensional column, and using water and methanol as mobile phases for gradient elution under the fluorescence detector. The separation detection of ergocalciferol and cholecalciferol from other impurities was achieved on the UV detector by Alphasil VC-C_(18) column(150 mm×4.6 mm, 3.5 µm) as the two-dimensional column. RESULTS: The baseline separation detection of retinol, ergocalciferol, cholecalciferol and five tocopherols was achieved. The compounds were linearly correlated within the set range, and the correlation coefficients were >0.999. The recovery rate of the method was between 85.4% and 106.4%. The detection limit of all-trans retinol was 0.7 µg/100 g, and the limit of quantitation was 2.4 µg/100 g. The limits of detection and quantification of tocopherols ranged from 1.1 to 2.5 µg/100 g and 3.6 to 8.3 µg/100 g. The detection limit of ergocalciferol and cholecalciferol was 0.3 µg/100 g, and the limit of quantification was 1.0 µg/100 g. In the end, finished dishes such as dry fried hairtail, braised mushroom, steamed egg with soy sauce, sweet and sour ribs were repeatedly measured by this method for six times, and the relative standard deviation was less than 10%. CONCLUSION: The method has the characteristics of simple operation, good repeatability, high accuracy and friendly to operators and ecological environment. It can realize the simultaneous typing detection of vitamins A, D and E in finished dishes.

[Determination of 21 polyphenols in fruits by high performance liquid chromatography].

OBJECTIVE: To establish a high performance liquid chromatography(HPLC) method for simultaneous determination of flavonoids(epicatechin, isoorientin, eriocitrin, hyperoside, narirutin, naringin, hesperidin, neohesperidin, quercetin, didymin, naringenin and hesperetin), phenolic acids(gallic acid, chlorogenic acid, caffeic acid, ferulic acid, sinapic acid and p-coumaric acid), terpenoids(limonin), stilbenes(piceatannol and resveratrol) in fruits. METHODS: Add ethanol-water(4∶1, V/V) solution to the fruit pulp sample, ultrasonic extraction at 55 ℃ for 10 min, after centrifugation, the supernatant was dried and re-dissolved, and then eluted with ZORBAX Eclipse XDB-C_(18)(4.6 mm×100 mm, 3.5 µm) as chromatographic column, polyphenols are determined by HPLC under switching wavelength, and quantified by external standard method. RESULTS: The 21 polyphenols can be detected within 45 min and well separated from baseline, with linear correlation coefficients greater than 0.999, the detection limit of the method was 0.03-0.31 µg/g, and the limit of quantitation of the method was 0.09-1.03 µg/g. The intra-day precision was 0.5%-9.3%, the inter-day precision was 2.0%-9.6%. The recoveries of more than 90% polyphenols in the three fruits at three levels were between 80.0% and 119.8%. The relative standard deviation of spiked recoveries of three samples was less than 10.8%. The detection result of citrus, mango, blueberry, persimmon and other samples proved that the composition of polyphenols in different fruits was different. CONCLUSION: The method is sensitive, accurate and reproducible, and is suitable for the determination of polyphenols in fruits.

[Development of nutrient profiling model include nutrients and food groups evaluation].

OBJECTIVE: To develop the nutrient profiling model based on the nutrition and health status of Chinese residents, and comprehensively evaluate the nutritional quality of food and dietary pattern. METHODS: Nutrient-and food group-indexes closely related to the main health problems of Chinese residents were screened and determined based on the level of evidence. Different food and dietary patterns were taken as examples to calculate the nutrient dense score. RESULTS: The nutrient indexes of this model included protein, dietary fiber, vitamin D, sodium, potassium, calcium, saturated fatty acid and sugar. Food group indexes included whole grain and legume, fruit, vegetable, soybean and nut. CONCLUSION: Nutrient profiling model included both nutrient-and food group-indexes could provide a more comprehensive evaluation on nutrient density of food and diet, and the rationality of diet align with dietary guidelines.

[Determination of vitamin K in animal foods].

OBJECTIVE: To establish a high performance liquid chromatography method with fluorescence detection for a variety of vitamin K and to assess the content of vitamin K in animal foods. METHODS: Animal foods were hydrolyzed by 0.2 g lipase and 0.1 g protease in pH 8 for 4 hours, extracted with isooctane followed by rotary evaporation and reconstitution. The mobile phase was 900 mL methanol and 100 mL tetrahydrofuran which contained 5 mmol glacial acetic acid, 11 mmol zinc chloride and 6 mmol anhydrous sodium acetate. The content of vitamin K_1, menaquinone-4(MK-4), and menaquinone-7(MK-7)were separated on Atlantis T3 column(4.6 mm×250 mm, 5 µm)by high-performance liquid chromatography. The fluorescence detector was set at a wavelength of 320 nm for excitation and 410.3 nm for emission. RESULTS: The linear range of the method was 0.01-0.40 µg/mL, and coefficient of determination was > 0.999. The spiked recoveries were 84.4%-124.2% with relative standard deviation was <6%(n=6). MK-4 was the main form of vitamin K in pork and chicken. The highest content of vitamin K_1 was found in beef, and MK-7 could be detected in aquatic products. CONCLUSION: The proposed method is successfully applied for the determination of vitamin K in animal foods. A variety of vitamin K are distributed differently in distinct animals.

Bioremediation of heavy metal contaminated aqueous solution by using red algae Porphyra leucosticta.

Bioremediation is an effective process for the removal and recovery of heavy metal ions from aqueous solutions. In this study, red algae Porphyra leucosticta was examined to remove Cd(II) and Pb(II) ions from wastewater through biological enrichment and biological precipitation. The experimental parameters that affect the bioremediation process such as pH, contact time and biomass dosage were studied. The maximum bioremediation capacity of metal ions was 31.45 mg/g for Cd(II) and 36.63 mg/g for Pb(II) at biomass dosage 15 g/L, pH 8.0 and contact time 120 minutes containing initial 10.0 mg/L of Cd(II) and 10.0 mg/L of Pb(II) solution. Red algae Porphyra leucosticta biomass was efficient at removing metal ions of 10.0 mg/L of Cd(II) and 10.0 mg/L of Pb(II) solution with bioremediation efficiency of 70% for Cd(II) and 90% for Pb(II) in optimal conditions. At the same time, the removal capacity for real industrial effluent was gained at 75% for 7.6 mg/L Cd(II) and 95% for 8.9 mg/L Pb(II). In conclusion, it is demonstrated that red algae Porphyra leucosticta is a promising, efficient, cheap and biodegradable sorbent biomaterial for reducing heavy metal pollution in the environment and wastewater.

[Study of DHA algal oil compound preparation on improving memory in mice].

OBJECTIVE: To investigate the effects of docosahexaenoic acid (DHA) algal oil compound preparation on improving memory in mice. METHODS: A total of 144 KM mice weighed 18 - 22 gram were classified into three groups randomized for the step - down test, passive avoidance test and water maze test respectively. For each test, we select 48 mice initially, then allocated the mice into 4 groups at random, each with 12 mice. According to the recommended dose of DHA algae oil soft capsule (600 mg/d) for human, each adult weight calculating on 60 kg meters, then we calibrate the dose to 50, 100 and 300 mg/kg BW (the equivalent doses of 5, 10 and 30 times to human) for three groups, respectively. The dose for intragastric injection administration is 10 ml/kg BW DHA algae oil, equivalent amount of edible vegetable oil for the control group, once per day for continuous 30 days. Just after the last lavage, we conduct the step - down test, passive avoidance test and water maze test respectively. RESULTS: DHA algal oil compound preparation has obvious effect on improving the reappearance ability of passive avoidance and the response of the avoiding darkness (P < 0.05). The indicators of water maze test were different statistically between DHA algal oil compound preparation groups and control group (P < 0.05). CONCLUSION: DHA algal oil compound preparation may play an important role in memory improvement in mice.

A concise update on prostate pathology.

Prostate carcinoma is the most common non-cutaneous malignancy in men in developed countries and the incidence has been steadily rising in the developing countries. Active research in recent years has led to tremendous progress in our understanding of the biology and genetics, and marked improvement in diagnosis and treatment of prostate cancer. Gleason grading has remained as the cornerstone for management of patients with prostate cancer. However, the grading system has continuously evolving since its inception in response to changes in the clinical practice of diagnosis and treatment of prostate cancer. The modification of Gleason grading system implemented by the International Society of Urological Pathology in 2005 has profoundly changed the way prostate cancer is graded and consequently how patients are managed. Several prostate cancer histological types with distinct clinical and pathological features have been rediscovered or redefined. Finally, elucidations of the molecular and genetic mechanism helps not only better understand the pathogenesis of prostate cancer, but also identify biomarkers for improved diagnosis, risk stratification and clinical management. This article briefly reviews the most recent advances in the Gleason grading system, new histological types and molecular genetics of prostate cancer.

Cross-layer importance evaluation for neural network pruning.

Filter pruning has achieved remarkable success in reducing memory consumption and speeding up inference for convolutional neural networks (CNNs). Some prior works, such as heuristic methods, attempted to search for suitable sparse structures during the pruning process, which may be expensive and time-consuming. In this paper, an efficient cross-layer importance evaluation (CIE) method is proposed to automatically calculate proportional relationships among convolutional layers. Firstly, every layer is pruned separately by grid sampling way to obtain the accuracy of the model for all sampling points. And then, contribution matrices are built to describe the importance of each layer to model accuracy. Finally, the binary search algorithm is used to search the optimal sparse structure under a target pruned value. Extensive experiments on multiple representative image classification tasks demonstrate that proposed method acquires better compression performance under a little time cost compared to existing pruning algorithms. For instance, it reduces more than 50% FLOPs with only a small loss of 0.93% and 0.43% in the top-1 and top-5 accuracy for ResNet50, respectively. At the cost of only 0.24% accuracy loss, the pruned VGG19 model parameters are successfully compressed by 27.23× and the throughput has increased by 2.46×. On the whole, CIE has an excellent effect on the deployment and application of the CNNs model in edge device in terms of efficiency and accuracy.

Metabolite profiling of camel milk and the fermentation bacteria agent TR1 fermented two types of sour camel milk using LC-MS in relation to their probiotic potentials.

Camel milk is a nutrient-rich diet and fermentation affects its nutritional value and probiotic function. In this study, sour camel milk and oat jujube sour camel milk were prepared using fermentation bacteria agent TR1, and the metabolites of camel milk, sour camel milk and oat jujube sour camel milk were detected using a non-targeted metabolomics approach using liquid chromatography-mass spectrometry (LC-MS).The results showed that the partial least squares discriminant analysis (PLS-DA) with 100 % accuracy and good predictive power detected 343 components in positive ion mode and 220 components in negative ion mode. The differential metabolites were mainly organic acids, amino acids, esters, vitamins and other substances contained in camel milk.It showed that there were significant differences in the metabolites of camel milk, sour camel milk and oat jujube sour camel milk. Based on the pathway enrichment analysis of the three dairy products in the KEGG database, 12 metabolic pathways mainly involved in the positive ion mode and 20 metabolic pathways mainly involved in the negative ion mode were identified. The main biochemical metabolic pathways and signal transduction pathways of the differential metabolites of the three dairy products were obtained. This study provides theoretical support for improving the nutritional quality and probiotic function of camel milk and fermented camel milk products and provides a basis for the development of relevant processing technologies and products for camel milk and fermented camel milk.

Characterization of phospholipidome in milk, yogurt and cream, and phospholipid differences related to various dairy processing methods.

Milk phospholipids have multiple health benefits, but the deficiency of detailed phospholipid profiles in dairy products brings obstacles to intake calculation and function evaluation of dairy phospholipids. In present study, 306 phospholipid molecular species were identified and quantified among 207 milk, yogurt and cream products using a HILIC-ESI-Q-TOF MS and a HILIC-ESI-QQQ MS. The phospholipid profiles of five mammals' milk show that camel milk contains the most abundant phosphatidylethanolamine, phosphatidylserine and sphingomyelin; cow, yak and goat milk have similar phospholipidomes, while buffalo milk contains abundant phosphatidylinositol. Fewer plasmalogens but more lyso-glycerolphospholipids were found in ultra-high-temperature (UHT) sterilized milk than in pasteurized milk, and higher proportions of lyso-glycerolphospholipid/total phospholipid were observed in both cream and skimmed/semi-skimmed milk than whole milk, indicating that UHT and skimming processes improve glycerolphospholipid degradation and phospholipid nutrition loss. Meanwhile, more diacyl-glycerolphospholipids and less of their degradation products make yogurt a better phospholipid resource than whole milk.

A game theoretical model for location of terror response facilities under capacitated resources.

This paper is concerned with the effect of capacity constraints on the locations of terror response facilities. We assume that the state has limited resources, and multiple facilities may be involved in the response until the demand is satisfied consequently. We formulate a leader-follower game model between the state and the terrorist and prove the existence and uniqueness of the Nash equilibrium. An integer linear programming is proposed to obtain the equilibrium results when the facility number is fixed. The problem is demonstrated by a case study of the 19 districts of Shanghai, China.

[Simultaneous determination of 11 polyphenols in ratafee by reversed-phase high performance liquid chromatography].

OBJECTIVE: To develop a novel quantitative method for simultaneous analysis of 11 polyphenols in ratafee by reversed-phase high performance liquid chromatography. METHODS: The separation was performed on a Agilent Eclipse XDB C18 (4.6 mmx250 mm, 51 m) and the column temperature was set at 30 degrees C. The mobile phase was consisted of 0. 1% phosphoric acid (A) and methanol (B), with a linear gradient elution at a flow rate of I. 0 ml/min. The gradient program was as follow: 10% B isocratic (0-10 min), 10%-20% B liner (10-15 min), 20% B isocratic (15-25 min), 20%-60% B liner (25-65 min), 60% B isocratic (65-75 min). And detection wavelength was 210, 270, 320, 327 and 360 nm. RESULTS: The correlation coefficients of the 11 calibration curves were all higher than 0.9995. The scopes of the recovery rate was within the range of 85.1% to 112.0%, with RSDs no more than 2%. The limit of quantification (LOQ) and the detection limit (LOD) were 0.32-1.59 ng and 0.10-0.48 ng, respectively. CONCLUSION: The method is accurate and reliable, and can successfully analyze different ratafee samples from different sources.

Analysis of metabolites of bactrain camel milk in Alxa of China before and after fermentation with fermenting agent TR1 based on untargeted LC-MS/MS based metabolomics.

Camel milk produces many beneficial functional compounds and affects body health through metabolism. The differential metabolites of bactrain camel milk in Alxa before and after fermentation were identified by liquid chromatography-tandem mass spectrometry based metabolomics (LC-MS/MS). The differential metabolite pathway types were also identified in this paper. We obtained the following results that 148 and 82 differential metabolites were detected in positive and negative ion mode respectively, 85 differential metabolites were shown a significant upward trend and 63 with downward trend after fermentation in positive ion mode. Meanwhile, 32 differential metabolites characterized upward trend and 50 characterized downward trend in negative ion mode. The differential metabolites were mainly organic acids, amino acids, esters, vitamins and other substances contained in camel milk. Among them, most up-regulated substances had the functions of lowering blood pressure, lowering blood sugar, treatment of inflammation, antibiosis and other effects. Many harmful substances were significantly down-regulated after camel milk fermentation. However, there were also some metabolites whose prebiotic functions have been weakened by camel milk fermentation, which may provide reference values for healthcare function, exploitation and application of camel milk.

MBFQuant: A Multiplier-Bitwidth-Fixed, Mixed-Precision Quantization Method for Mobile CNN-Based Applications.

Deploying Convolutional Neural Network (CNN)-based applications to mobile platforms can be challenging due to the conflict between the restricted computing capacity of mobile devices and the heavy computational overhead of running a CNN. Network quantization is a promising way of alleviating this problem. However, network quantization can result in accuracy degradation and this is especially the case with the compact CNN architectures that are designed for mobile applications. This paper presents a novel and efficient mixed-precision quantization pipeline, called MBFQuant. It redefines the design space for mixed-precision quantization by keeping the bitwidth of the multiplier fixed, unlike other existing methods, because we have found that the quantized model can maintain almost the same running efficiency, so long as the sum of the quantization bitwidth of the weight and the input activation of a layer is a constant. To maximize the accuracy of a quantized CNN model, we have developed a Simulated Annealing (SA)-based optimizer that can automatically explore the design space, and rapidly find the optimal bitwidth assignment. Comprehensive evaluations applying ten CNN architectures to four datasets have served to demonstrate that MBFQuant can achieve improvements in accuracy of up to 19.34% for image classification and 1.12% for object detection, with respect to a corresponding uniform bitwidth quantized model.