Multispectral Signals Research Articles

Multispectral Thermometry Method Based on Optimisation Ideas.

Multispectral thermometry is based on the law of blackbody radiation and is widely used in engineering practice today. Temperature values can be inferred from radiation intensity and multiple sets of wavelengths. Multispectral thermometry eliminates the requirements for single-spectral and spectral similarity, which are associated with two-colour thermometry. In the process of multispectral temperature inversion, the solution of spectral emissivity and multispectral data processing can be seen as the keys to accurate thermometry. At present, spectral emissivity is most commonly estimated using assumption models. When an assumption model closely matches an actual situation, the inversion of the temperature and the accuracy of spectral emissivity are both very high; however, when the two are not closely matched, the inversion result is very different from the actual situation. Assumption models of spectral emissivity exhibit drawbacks when used for thermometry of a complex material, or any material whose properties dynamically change during a combustion process. To address the above problems, in the present study, we developed a multispectral thermometry method based on optimisation ideas. This method involves analysing connections between measured temperatures of each channel in a multispectral temperature inversion process; it also makes use of correlations between multispectral signals at different temperatures. In short, we established a multivariate temperature difference correlation function based on the principles of multispectral radiometric thermometry, using information correlations between data for each channel in a temperature inversion process. We then established a high-precision thermometry model by optimising the correlation function and correcting any measurement errors. This method simplifies the modelling process so that it becomes an optimisation problem of the temperature difference function. This also removes the need to assume the relationships between spectral emissivity and other physical quantities, simplifying the process of multispectral thermometry. Finally, this involves correction of the spectral data so that any impact of measurement error on the thermometry is reduced. In order to verify the feasibility and reliability of the method, a simple eight-channel multispectral thermometry device was used for experimental validation, in which the temperature emitted from a blackbody furnace was identified as the standard value. In addition, spectral data from the 468-603 nm band were calibrated within a temperature range of 1923.15-2273.15 K, resulting in multispectral thermometry based on optimisation principles with an error rate of around 0.3% and a temperature calculation time of less than 3 s. The achieved level of inversion accuracy was better than that obtained using either a secondary measurement method (SMM) or a neural network method, and the calculation speed achieved was considerably faster than that obtained using the SMM method.

Application of the Savitzky-Golay filter in multi-spectral signal processing

Multi-spectral signals are the result of the interaction between electromagnetic energy and the test material, which is then displayed by the signal fluctuation pattern of the test material. Signal fluctuations are inaccuracies in the peak amplitude of a signal caused by noise in the data. This fluctuation pattern reflects the properties of the test material, especially in this case H2O. To overcome this problem, it is necessary to use the right filter to smooth the signal and reduce the noise in the data so that the fluctuation pattern obtained is clearer and more accurate. This research involves the segmentation of HF fluctuation patterns, followed by the application of a Savitzky-Golay filter for signal smoothing. Signal quality is assessed objectively by calculating the Signal to Noise Ratio (SNR) and Mean Square Error (MSE). The research results show that the Savitzky-Golay filter succeeded in reducing noise and producing clearer fluctuation patterns. The SNR value varies, with the largest value reaching 16.6146 dB, and the smallest value being 3.0171 dB. This research contributes to a new method, namely the Savitzky-Golay adaptive filter, to identify multi-spectral signal fluctuation patterns more effectively, thereby enabling more accurate identification of fluctuation patterns. Apart from that, this research also provides insight into the characteristics of H2O which can be identified through fluctuation patterns, especially in certain segments with high amplitude. This method has potential for applications in various fields, especially in precise multi-spectral signal analysis.

Influence of full penetration welding on the multispectral signal of the process radiation during laser beam welding of austenitic stainless steel

A 100% visual inspection of the welds carried out is not always possible for economic reasons. Different methods of process monitoring are therefore used for non-destructive testing of process and weld seam quality. In the following, the recorded multispectral signals of the reflective process radiation are evaluated using descriptive statistics and characteristics of a full penetration weld during laser beam welding of austenitic stainless steel are worked out. The aim is to detect and localize unintentional full penetration welding at an early stage and thus avoid rejects. For this purpose, adapted sample geometries with varying material thicknesses are prepared for the test series. As a result, the need to vary the welding process parameters to achieve the different welding states (full and partial penetration weld) is avoided. By evaluating the measured values of the wavelength-selective signals, a significant influence on the signal intensity and the signal frequency could be confirmed.

Prediction of Stroke Onset Time With Combined Fast High-Resolution Magnetic Resonance Spectroscopic and Quantitative T2 Mapping.

The purpose of this work is to develop a multispectral imaging approach that combines fast high-resolution 3D magnetic resonance spectroscopic imaging (MRSI) and fast quantitative T2 mapping to capture the multifactorial biochemical changes within stroke lesions and evaluate its potentials for stroke onset time prediction. Special imaging sequences combining fast trajectories and sparse sampling were used to obtain whole-brain maps of both neurometabolites (2.0 × 3.0 × 3.0 mm3) and quantitative T2 values (1.9 × 1.9 × 3.0 mm3) within a 9-minute scan. Participants with ischemic stroke at hyperacute (0-24 h, n = 23) or acute (24 h-7d, n = 33) phase were recruited in this study. Lesion N-acetylaspartate (NAA), lactate, choline, creatine, and T2 signals were compared between groups and correlated with patient symptomatic duration. Bayesian regression analyses were employed to compare the predictive models of symptomatic duration using multispectral signals. In both groups, increased T2 and lactate levels, as well as decreased NAA and choline levels were detected within the lesion (all p < 0.001). Changes in T2, NAA, choline, and creatine signals were correlated with symptomatic duration for all patients (all p < 0.005). Predictive models of stroke onset time combining signals from MRSI and T2 mapping achieved the best performance (hyperacute: R2 = 0.438; all: R2 = 0.548). The proposed multispectral imaging approach provides a combination of biomarkers that index early pathological changes after stroke in a clinical-feasible time and improves the assessment of the duration of cerebral infarction. Developing accurate and efficient neuroimaging techniques to provide sensitive biomarkers for prediction of stroke onset time is of great importance for maximizing the proportion of patients eligible for therapeutic intervention. The proposed method provides a clinically feasible tool for the assessment of symptom onset time post ischemic stroke, which will help guide time-sensitive clinical management.

A Geologic Map of Vesta Produced Using a Hybrid Method for Incorporating Spectroscopic and Morphologic Data

We have constructed a global geologic map of the minor planet Vesta at 1:300,000-scale using Dawn spacecraft imaging, spectroscopic, topographic, and elemental data. In this effort, we used a mapping method that requires creating two maps independently: the first map uses morphology and topography to define map units, while the second map relies on multispectral data (“color”) to define units. The two are then combined into a hybrid product that retains the maximum amount of unique information from both maps in a readable format. This effort has revealed that for bodies where cratering is the dominant unit-forming process, and where there is not a close correlation between morphological feature types and multispectral signal, a hybrid mapping method better retains unique information carried by multispectral data during the mapping process than traditional morphology-based methods alone. Conversely, relying too heavily on color data risks placing too much emphasis on information drawn from the top few microns of the surface. To ensure both consistency and retention of unique information, we created a decision tree for determining which data would be primary in choosing where to draw unit boundaries. Also due to the significant amount of information borne by spectral data, we repurposed traditional mapping nomenclature so that subscripts carry color information. We recommend using this mapping methodology on bodies where (a) morphologic feature boundaries are commonly subtle, gradational, or both, and (b) spectral data carries a significant amount of unique data for identifying, characterizing, and interpreting geologic units.

Smart soils track the formation of pH gradients across the rhizosphere

AimsOur understanding of the rhizosphere is limited by the lack of techniques for in situ live microscopy. Current techniques are either destructive or unsuitable for observing chemical changes within the pore space. To address this limitation, we have developed artificial substrates, termed smart soils, that enable the acquisition and 3D reconstruction of chemical sensors attached to soil particles.MethodsThe transparency of smart soils was achieved using polymer particles with refractive index matching that of water. The surface of the particles was modified both to retain water and act as a local sensor to report on pore space pH via fluorescence emissions. Multispectral signals were acquired from the particles using a light sheet microscope, and machine learning algorithms predicted the changes and spatial distribution in pH at the surface of the smart soil particles.ResultsThe technique was able to predict pH live and in situ within ± 0.5 units of the true pH value. pH distribution could be reconstructed across a volume of several cubic centimetres around plant roots at 10 μm resolution. Using smart soils of different composition, we revealed how root exudation and pore structure create variability in chemical properties.ConclusionSmart soils captured the pH gradients forming around a growing plant root. Future developments of the technology could include the fine tuning of soil physicochemical properties, the addition of chemical sensors and improved data processing. Hence, this technology could play a critical role in advancing our understanding of complex rhizosphere processes.

Multi‐spectral flare designs using both red‐ and green‐light emitters: Progress towards wavelength selectable signals

AbstractPyrotechnic flares are used extensively, with each color relating to specific information such as landing zone, enemy location, medical evacuation, or submarine location. While all current military flares are restricted to a single illumination color, progress towards a wavelength selectable signal and/or multispectral signal is of interest. In this effort, we use additive manufacturing, via direct ink write approach, to print three unique configurations with green/red light emitter compositions (GLE/RLE). Varying the infill design provides the opportunity to fire either of the base colors individually (red‐ and/or green‐light illuminant) or to fire them simultaneously to achieve a multispectral result. In this work, architecture is shown to play a critical role, and when the GLE composition is approx. 70 wt% of the overall, then a reasonable multi‐spectral result is achieved. While early, this work demonstrates that there is significant promise to optimize wavelength selectable signals via architecture achievable with additive manufacturing.

Glucose diagnosis system combining machine learning and NIR photoacoustic multispectral using a low power CW laser.

Non-invasive, portable, economical, dynamic blood glucose monitoring device has become a functional requirement for diabetes in his regulating entire life. In a photoacoustic (PA) multispectral near-infrared diagnosis system, the glucose in aqueous solutions was excited by low power (order of milliwatts) CW laser whose wavelengths were from 1500 to 1630 nm. The glucose in aqueous solutions to be analyzed was contained within the photoacoustic cell (PAC). The PA multispectral signals were measured using a piezoelectric detector, and then the voltage signals from the piezoelectric detector were amplified with a precision Lock-in Amplifier (MFLI500K). The continuously tunable lasers were used to verify the various influencing factors of the PA signal, and the PA spectrum of the glucose solution was examined. Subsequently, six wavelengths with high power were selected at approximately equal intervals from 1500 to 1630 nm, and the gaussian process regression of the quadratic rational kernel was used to collect data through these wavelengths to predict the glucose concentration. The experimental results showed that the near-infrared PA multispectral diagnosis system could be engineered for the prediction of the glucose level (more than 92%, zone A of Clarke Error Grid). Subsequently, the model trained with glucose solution was used to predict serum glucose. With the increase of serum glucose content, the prediction results of the model also showed a high linear relationship, indicating that the photoacoustic method was sensitive to the detection of glucose concentration changes. The results of our study have the potential to not only better develop the PA blood glucose meter but also extend the viability into the detection of otherwise blood components.

DOP12 Non-invasive assessment of intestinal inflammatory activity in Ulcerative Colitis by Multispectral Optoacoustic Tomography (MSOT)

Abstract Background In order to guide therapy in Ulcerative Colitis (UC), repeated determination of intestinal inflammatory activity is essential. Endoscopy is the standard procedure to assess inflammation in UC. However innovative methods for non-invasive, uncomplicated and risk free estimation of inflammatory activity are needed as bowel preparation, patients discomfort and risk of procedural complications limit the (frequent) use of colonoscopy. Multispectral optoacoustic tomography (MSOT) is a promising new method to measure inflammation in UC. Using short and harmless impulses of NIR-lasers, it allows for determination of a specific hemoglobin-signal in the bowel-wall and therefore inflammatory activity in affected bowel segments. However, its informative value in UC has not been evaluated so far. Methods In 34 patients with confirmed UC, clinical activity parameters (e.g. clinical Mayo-Subscore, B-mode-sonography, C-reactive protein, white blood count) were collected and MSOT of the sigmoid was performed within 2 weeks before/after endoscopy. For MSOT, a commercially available clinical MSOT-system (Acuity Echo, iThera Medical, Munich) was used with sequential analysis of collected data on an external desktop PC. Finally, clinical data, ultrasound findings (Limberg) and MSOT-parameters (single wavelenghts 760 nm, 800 nm, 900 nm; multispectral signals hb, hbO2, hbT) were correlated with endoscopic findings (Mayo endoscopic Subscore, MES). Results We found strong and significant correlation between MES and MSOT parameters 800 nm (Spearman r = 0,6599; p &amp;lt; 0,0001) and HbO2 (Spearman r = 0,6695; p &amp;lt; 0,0001), superior to sonographic evaluation of the inflammatory activity in affected bowel segments (Spearman r = 0,4914; p = 0,0023). Simultaneously these MSOT parameters demonstrated excellent sensitivity and specifity in distinguishing moderately to highly active (MES 2,3) from inactive and mild disease (MES 0,1) (800nm: AUROC 0,9063 (p &amp;lt; 0,0001); sensitivity = 93,75 %, specificity = 88,89 %; HbO2: AUROC 0,9063 (p &amp;lt; 0,0001); sensitivity = 100 %, specificity = 88,89 %). Conclusion MSOT is a promising approach to non-invasively assess intestinal inflammation in UC and therefore monitor anti-inflammatory therapy in these patients. Further studies are required to validate these findings.

Vertically Stacked Full Color Quantum Dots Phototransistor Arrays for High-Resolution and Enhanced Color-Selective Imaging.

Color-selective multifunctional and multiplexed photodetectors have attracted considerable interest with the increasing demand for color filter-free optoelectronics which can simultaneously process multispectral signal via minimized system complexity. The low efficiency of color-filter technology and conventional laterally pixelated photodetector array structures often limit opportunities for widespread realization of high-density photodetectors. Here, low-temperature solution-processed vertically stacked full color quantum dot (QD) phototransistor arrays are developed on plastic substrates for high-resolution color-selective photosensor applications. Particularly, the three different-sized/color (RGB) QDs are vertically stacked and pixelated via direct photopatterning using a unique chelating chalcometallate ligand functioning both as solubilizing component and, after photoexposure, a semiconducting cement creating robust, insoluble, and charge-efficient QD layers localized in the a-IGZO transistor region, resulting in efficient wavelength-dependent photo-induced charge transfer. Thus, high-resolution vertically stacked full color QD photodetector arrays are successfully implemented with the density of 5500 devices cm-2 on ultrathin flexible polymeric substrates with highly photosensitive characteristics such as photoresponsivity (1.1 × 104 AW-1 ) and photodetectivity (1.1 × 1018 Jones) as well as wide dynamic ranges (>150dB).

Non-invasive assessment of exfoliated kidney cells extracted from urine using multispectral autofluorescence features

Optimally preserved urinary exfoliated renal proximal tubule cells were assessed by multispectral imaging of cell autofluorescence. We demonstrated different multispectral autofluorescence signals in such cells extracted from the urine of patients with healthy or diseased kidneys. Using up to 10 features, we were able to differentiate cells from individuals with heathy kidneys and impaired renal function (indicated by estimated glomerular filtration rate (eGFR) values) with the receiver operating characteristic area under the curve (AUC) of 0.99. Using the same method, we were also able to discriminate such urine cells from patients with and without renal fibrosis on biopsy, where significant differences in multispectral autofluorescence signals (AUC = 0.90) were demonstrated between healthy and diseased patients (p < 0.05). These findings show that multispectral assessment of the cell autofluorescence in urine exfoliated proximal tubule kidney cells has the potential to be developed as a sensitive, non-invasive diagnostic method for CKD.

Increase the contrast of the image of objects in the active optoelectronic system with dynamic spectral processing of optical emitting

The article discusses a method for increasing the contrast of images in an optoelectronic system based on active dynamic spectral matched filtering. The principles of constructing active optoelectronic systems with matched filtering are based on the fact that the optical system uses a set of amplitude-controlled radiation sources operating in different parts of the spectral range as emitting sources. It is essential that the energy composition of the light emission control signals is formed on the basis of a priori information about the characteristics of the target and the background, so as to reduce the value of the spectral components of the optical signal reflected from the surface belongs to the background and with minimal attenuation of the signal intensity belonging to the object. The method assumes the presence of a set of a priori information about the spectral characteristics of the background and the object to form the instrumental function for controlling the amplitude of emitting sources. The analysis of the mathematical and physical aspects of systems with dynamic spectral processing of active type optical emitting is made. It is shown that an active optoelectronic system with dynamic spectral processing can be considered as an analog processor for calculating the dot product of a vector by a vector. One of the factors is the reflection coefficient from a surface with a priori known characteristics, and the second is a dimmable multispectral signal. A block diagram of an optoelectronic system with dynamic spectral processing of optical emitting with active formation of the information field in order to increase the contrast of the object image has been developed. The goal of the article is to develop the mathematical and physical foundations for constructing an active optoelectronic system with dynamic spectral processing of optical emitting in order to increase the image contrast.

Artifact Suppression with Geodesic Kernel Filter for Defocused Images Restored by Wiener Filtering

Abstract—The Wiener-filter restoration of images degraded by defocus blur usually causes specific low frequency artifacts. In this work, we propose to suppress such low frequency artifacts with a filter based on the geodesic distance affinity. A multispectral signal model is considered and the main idea of the proposed algorithm is based on assumption that low frequency outliers and additive noise in different image channels are uncorrelated. Thus, the affinity space formed by opposite channels can be used to efficiently suppress the Wiener filter restoration artifacts. The performance of the proposed filter is analyzed and compared in terms of the PSNR accuracy.

Dipole modeling of multispectral signal for detecting metallic biopsy markers during MRI-guided breast biopsy: a pilot study.

During MRI-guided breast biopsy, a metallic biopsy marker is deployed at the biopsy site to guide future interventions. Conventional MRI during biopsy cannot distinguish such markers from biopsy site air, and a post-biopsy mammogram is therefore performed to localize marker placement. The purpose of this pilot study is to develop dipole modeling of multispectral signal (DIMMS) as an MRI alternative to eliminate the cost, inefficiency, inconvenience, and ionizing radiation of a mammogram for biopsy marker localization. DIMMS detects and localizes the biopsy marker by fitting the measured multispectral imaging (MSI) signal to the MRI signal model and marker properties. MSI was performed on phantoms containing titanium biopsy markers and air to illustrate the clinical challenge that DIMMS addresses and on 20 patients undergoing MRI-guided breast biopsy to assess DIMMS feasibility for marker detection. DIMMS was compared to conventional MSI field map thresholding, using the post-procedure mammogram as the reference standard. Biopsy markers were detected and localized in 20 of 20 cases using MSI with automated DIMMS post-processing (using a threshold of 0.7) and in 18 of 20 cases using MSI field mapping (using a threshold of 0.65 kHz). MSI with DIMMS post-processing is a feasible technique for biopsy marker detection and localization during MRI-guided breast biopsy. With a 2-min MSI scan, DIMMS is a promising MRI alternative to the standard-of-care post-biopsy mammogram.

Spectral Super-Resolution with Optimized Bands

Hyperspectral (HS) sensors sample reflectance spectrum in very high resolution, which allows us to examine material properties in very fine details. However, their widespread adoption has been hindered because they are very expensive. Reflectance spectra of real materials are high dimensional but sparse signals. By utilizing prior information about the statistics of real HS spectra, many previous studies have reconstructed HS spectra from multispectral (MS) signals (which can be obtained from cheaper, lower spectral resolution sensors). However, most of these techniques assume that the MS bands are known apriori and do not optimize the MS bands to produce more accurate reconstructions. In this paper, we propose a new end-to-end fully convolutional residual neural network architecture that simultaneously learns both the MS bands and the transformation to reconstruct HS spectra from MS signals by analyzing large quantity of HS data. The learned band can be implemented in hardware to obtain an MS sensor that collects data that is best to reconstruct HS spectra using the learned transformation. Using a diverse set of real-world datasets, we show how the proposed approach of optimizing MS bands along with the transformation can drastically increase the reconstruction accuracy. Additionally, we also investigate the prospects of using reconstructed HS spectra for land cover classification.

Hierarchical Metamaterials for Multispectral Camouflage of Infrared and Microwaves

AbstractCamouflage is an emerging application of metamaterials owing to their exotic electromagnetic radiative properties. Based on the use of a selective emitter and an absorber as the metamaterials, most reported articles have suggested the use of single‐band camouflage, however, multispectral camouflage is a challenging issue owing to a difference of several orders of magnitude in the unit cell structure. Herein, hierarchical metamaterials (HMMs) for multispectral signal control when dissipating the absorbed energy of microwaves through the selective emission of infrared (IR) waves from the unit cell structure of the HMM are demonstrated. Integrating an IR selective emitter (IRE) with a microwave selective absorber, multispectral signal control with the large‐sized unit cell structures of up to 10 cm are realized. With an IRE, the emissive power from the HMM toward 5–8 µm is 1570% higher than the Au surface, which is preventing the occurrence of thermal instability. Furthermore, we determine that the signature levels of targeted IR waves (8–12 µm) and microwaves (2.5–3.8 cm) are reduced by up to 95% and 99%, respectively, when applying the HMM.

Application of Multiband Acoustooptical Filtering of Radiation to Solution of the Object Remote Identification Problems by the Method of Programmed Linear Separation of Multispectral Optical Signals in the Case of Reflection of a Wideband Light from Objects with A Priori Known Reflection Coefficients

A new approach to solution of the problem of remote determination of spectrally complex optical signals and identification on this base of two objects out of a set of possible spatially inseparable objects located in one small ground pixel is considered. Computer simulation is performed; it is shown that the results of measurements of the mean powers of filtered signals at the outputs of two (or more) selected reference multiband acoustooptical filters (MAOFs) can ensure subsequent programmed (digital) separation of spectrally complex optical signals and identification of the objects corresponding to these signals. By the example of an arbitrarily chosen group of ten types of objects and three types of the background, computer verification of the new method is carried out with the use of realizable parameters of modern MAOFs. It has been found that, in conditions of the actual accuracy of setting and/or measuring of the energy of the main components of the problem (~1%), this method can provide unambiguous correct decision about the type of the object located on part of a pixel with a probability of ~99.9%.

Shortwave IR Adaption of the Mid-Infrared Radiance Method of Fire Radiative Power (FRP) Retrieval for Assessing Industrial Gas Flaring Output

The radiative power (MW) output of a gas flare is a useful metric from which the rate of methane combustion and carbon dioxide emission can be inferred for inventorying purposes and regular global surveys based on such assessments are now being used to keep track of global gas flare reduction efforts. Several multispectral remote sensing techniques to estimate gas flare radiative power output have been developed for use in such surveys and single band approaches similar to those long used for the estimation of landscape fire radiative power output (FRP) can also be applied. The MIR-Radiance method, now used for FRP retrieval within the MODIS active fire products, is one such single band approach—but its applicability to gas flare targets (which are significantly hotter than vegetation fires) has not yet been assessed. Here we show that the MIR-Radiance approach is in fact not immediately suitable for retrieval of gas flare FRP due to their higher combustion temperatures but that switching to use data from a SWIR (rather than MWIR) spectral channel once again enables the method to deliver unbiased FRP retrievals. Over an assumed flaring temperature range of 1600–2200 K we find a maximum FRP error of ±13.6% when using SWIR observations at 1.6 µm and ±6.3% when using observations made at 2.2 µm. Comparing these retrievals to those made by the multispectral VIIRS ‘NightFire’ algorithm (based on Planck Function fits to the multispectral signals) we find excellent agreement (bias = 0.5 MW, scatter = 1.6 MW). An important implication of the availability of this new SWIR radiance method for gas flare analysis is the potential to apply it to long time-series from older and/or more spectrally limited instruments, unsuited to the use of multispectral algorithms. This includes the ATSR series of sensors operating between 1991–2012 on the ERS-1, ERS-2 and ENVISAT satellites and such long-term data can be used with the SWIR-Radiance method to identify key trends in global gas flaring that have occurred over the last few decades.

Optimal Waveband of Multispectral for Infrared Target Detection

Aiming at the low signal to noise ratio (SNR) and weak anti-interference problem of single band detection, we use a method called Joint Skewness-Kurtosis figure (JSFK) to select the optimal waveband. Firstly, the expression of infrared signal detection is given and the coupling and random characteristics of the detection process are analyzed. Then based on the statistical theory and utilizing the non-Gaussian distribution characteristic of JSKF method, the optimal wavebands of multispectral are selected. From the selection results, the wavebands selected by the JSKF method distribute dispersively in the band 2μm~15μm and have strong complementarity, which can better reflect the multispectral characteristics of target radiation. Finally, the wavebands selected by the IABC, ECA and JSKF methods are respectively fused by the PCA method. The experiment results show that the fusion effect of JSKF method is superior to the IABC and ECA methods. Besides, the JSKF method can realize real-time processing of multispectral signals detected by the airborne, which is significant for infrared target detection.

The research of near-infrared blood glucose measurement using particle swarm optimization and artificial neural network

Existing near-infrared non-invasive blood glucose detection modelings mostly detect multi-spectral signals with different wavelength, which is not conducive to the popularization of non-invasive glucose meter at home and does not consider the physiological glucose dynamics of individuals. In order to solve these problems, this study presented a non-invasive blood glucose detection model combining particle swarm optimization (PSO) and artificial neural network (ANN) by using the 1 550 nm near-infrared absorbance as the independent variable and the concentration of blood glucose as the dependent variable, named as PSO-2ANN. The PSO-2ANN model was based on two sub-modules of neural networks with certain structures and arguments, and was built up after optimizing the weight coefficients of the two networks by particle swarm optimization. The results of 10 volunteers were predicted by PSO-2ANN. It was indicated that the relative error of 9 volunteers was less than 20%; 98.28% of the predictions of blood glucose by PSO-2ANN were distributed in the regions A and B of Clarke error grid, which confirmed that PSO-2ANN could offer higher prediction accuracy and better robustness by comparison with ANN. Additionally, even the physiological glucose dynamics of individuals may be different due to the influence of environment, temper, mental state and so on, PSO-2ANN can correct this difference only by adjusting one argument. The PSO-2ANN model provided us a new prospect to overcome individual differences in blood glucose prediction.