Explainable artificial intelligence based analysis for interpreting infant fNIRS data in developmental cognitive neuroscience.

In the last decades, non-invasive and portable neuroimaging techniques, such as functional near infrared spectroscopy (fNIRS), have allowed researchers to study the mechanisms underlying the functional cognitive development of the human brain, thus furthering the potential of Developmental Cognitive Neuroscience (DCN). However, the traditional paradigms used for the analysis of infant fNIRS data are still quite limited. Here, we introduce a multivariate pattern analysis for fNIRS data, xMVPA, that is powered by eXplainable Artificial Intelligence (XAI). The proposed approach is exemplified in a study that investigates visual and auditory processing in six-month-old infants. xMVPA not only identified patterns of cortical interactions, which confirmed the existent literature; in the form of conceptual linguistic representations, it also provided evidence for brain networks engaged in the processing of visual and auditory stimuli that were previously overlooked by other methods, while demonstrating similar statistical performance.

Rapid determination of total phenolic content of whole wheat flour using near-infrared spectroscopy and chemometrics.

Phenolics in whole wheat products provide many health benefits. Wheat breeders, producers, and end-users are becoming increasingly interested in wheats with higher total phenolic content (TPC). Whole wheat flour with higher phenolics may have greater marketing value in the future. However, conventional methods determining TPC are costly and labor-intensive, which are not practical for wheat breeders to analyze several thousands of lines within a limited timeframe. We presented a novel application of near-infrared spectroscopy for TPC prediction in whole wheat flour. The optimal regression model demonstrated R2 values of 0.92 and 0.90 for the calibration and validation sets, and a residual prediction deviation value of 3.4. The NIR method avoids the tedious extraction and TPC assay procedures, making it more convenient and cost-effective. Our result also demonstrated that NIR can accurately quantify phenolics even at low concentration (less than 0.2%) in the food matrix such as whole wheat flour.

Influence of environmental factors on the detection of blood in sheep faeces using visible-near-infrared spectroscopy as a measure of Haemonchus contortus infection.

BACKGROUND: Existing diagnostic methods for the parasitic gastrointestinal nematode, Haemonchus contortus, are time consuming and require specialised expertise, limiting their utility in the field. A practical, on-farm diagnostic tool could facilitate timely treatment decisions, thereby preventing losses in production and flock welfare. We previously demonstrated the ability of visible-near-infrared (Vis-NIR) spectroscopy to detect and quantify blood in sheep faeces with high accuracy. Here we report our investigation of whether variation in sheep type and environment affect the prediction accuracy of Vis-NIR spectroscopy in quantifying blood in faeces. METHODS: Visible-NIR spectra were obtained from worm-free sheep faeces collected from different environments and sheep types in South Australia (SA) and New South Wales, Australia and spiked with various sheep blood concentrations. Spectra were analysed using principal component analysis (PCA), and calibration models were built around the haemoglobin (Hb) wavelength region (387-609 nm) using partial least squares regression. Models were used to predict Hb concentrations in spiked faeces from SA and naturally infected sheep faeces from Queensland (QLD). Samples from QLD were quantified using Hemastix® test strip and FAMACHA© diagnostic test scores. RESULTS: Principal component analysis showed that location, class of sheep and pooled versus individual samples were factors affecting the Hb predictions. The models successfully differentiated 'healthy' SA samples from those requiring anthelmintic treatment with moderate to good prediction accuracy (sensitivity 57-94%, specificity 44-79%). The models were not predictive for blood in the naturally infected QLD samples, which may be due in part to variability of faecal background and blood chemistry between samples, or the difference in validation methods used for blood quantification. PCA of the QLD samples, however, identified a difference between samples containing high and low quantities of blood. CONCLUSION: This study demonstrates the potential of Vis-NIR spectroscopy for estimating blood concentration in faeces from various types of sheep and environmental backgrounds. However, the calibration models developed here did not capture sufficient environmental variation to accurately predict Hb in faeces collected from environments different to those used in the calibration model. Consequently, it will be necessary to establish models that incorporate samples that are more representative of areas where H. contortus is endemic.

A simple design for the validation of a FT-NIR screening method: Application to the detection of durum wheat pasta adulteration.

The demand for the development of fast, easy-to-use and low-cost analytical methods for food adulteration analysis has being increasing in the last years. Although infrared spectroscopic techniques offer these advantages, the validation of screening methods requiring the application of multivariate data treatment is less frequently described in literature thus limiting their use as routine tools in control laboratories for food fraud monitoring. In this paper, an EU-validation procedure for screening methods was successfully applied to a multivariate FT-NIR spectroscopic method for the screening of durum wheat pasta samples adulterated with common wheat at the screening target concentration of 3%. Good results in terms of the cut-off value (2.32% mass fraction of soft wheat) and false suspect rates (0.1% for blanks; 13% at 1% mass fraction) demonstrated that the present validation approach would be a proof-of-strategy to be used for multivariate infrared methods applied for screening purposes.

Using near-infrared hyperspectral imaging with multiple decision tree methods to delineate black tea quality.

The evaluation of tea quality tended to be subjective and empirical by human panel tests currently. A convenient analytical approach without human involvement was developed for the quality assessment of tea with great significance. In this study, near-infrared hyperspectral imaging (HSI) combined with multiple decision tree methods was utilized as an objective analysis tool for delineating black tea quality and rank. Data fusion that integrated texture features based on gray-level co-occurrence matrix (GLCM) and short-wave near-infrared spectral features were as the target characteristic information for modeling. Three different types of supervised decision tree algorithms (fine tree, medium tree, and coarse tree) were proposed for the comparison of the modeling effect. The results indicated that the performance of models was enhanced by the multiple perception feature fusion. The fine tree model based on data fusion obtained the best predictive performance, and the correct classification rate (CCR) of evaluating black tea quality was 93.13% in the prediction process. This work demonstrated that HSI coupled with intelligence algorithms as a rapid and effective strategy could be successfully applied to accurately identify the rank quality of black tea.

Prediction of monounsaturated and polyunsaturated fatty acids of various processed pork meats using improved hyperspectral imaging technique.

Freezing, heating, and pickling are common processes for pork meats. Unsaturated fatty acids including monounsaturated fatty acids and polyunsaturated fatty acids are indispensable nutrition beneficial to human's health and growth. However, Unsaturated fatty acids are affected by processing methods. Hyperspectral imaging is a novel technique widely used for food quality and safety evaluation. In the current study, the contents of monounsaturated and polyunsaturated fatty acids were assessed by Hyperspectral imaging. Optimal wavelengths were selected by the regression coefficients curves of partial least squares regression models. The least-squares support vector machine models established achieved a better coefficient of determination in the Monte Carlo validation set than the partial least squares regression models developed and the R2MV values for the least squares - support vector machine models based on selected optimal wavelengths were higher than 0.81. Finally, colour maps of the contents of monounsaturated and polyunsaturated fatty acids were developed.

A rapid analysis method of safflower (Carthamus tinctorius L.) using combination of computer vision and near-infrared.

The quality of safflower (Carthamus tinctorius L.) in the market is uneven due to the problems of dyeing and adulteration of safflower, and there is no perfect standard for the classification of quality grade of safflower at present. In this study, computer vision (CV) and near-infrared (NIR) were combined to realize the rapid and nondestructive analysis of safflower. First, the partial least squares discrimination analysis (PLS-DA) model was used to qualitatively identify the dyed safflower from 150 samples. Then the partial least squares (PLS) model was used for quantitative analysis of the hydroxy safflower yellow pigment A (HSYA) and water extract of undyed safflower. Herein, the discrimination rate of PLS-DA model reached 100%, and the residual predictive deviation (RPD) values of the PLS models for HSYA and water extract were 2.5046 and 5.6195, respectively. It indicated that the rapid analysis method combining CV and NIR was reliable, and its results can provide important reference for the formulation of safflower quality classification standards in the market.

Reduction of repeatability error for analysis of variance-Simultaneous Component Analysis (REP-ASCA): Application to NIR spectroscopy on coffee sample.

A method to reduce repeatability error in multivariate data for Analysis of variance-Simultaneous Component Analysis (REP-ASCA) has been developed. This method proposes to adapt the acquisition protocol by adding a set containing repeated measures for describing repeatability error. Then, an orthogonal projection is performed in the row-space to reduce the repeatability error of the original dataset. Finally, ASCA is performed on the orthogonalized dataset. This method was evaluated on NIR spectral data of coffee beans. This study shows that the repeatability error due to physical variations between measurements can alter results of the analysis of variance. These effects are predominant in factors analysis and can be seen on spectra as constant or non-constant baselines. By reducing repeatability error with REP-ASCA, baselines are removed and factor analysis provides more information about chemical content of the factors of interest.

A modified mid-level data fusion approach on electronic nose and FT-NIR data for evaluating the effect of different storage conditions on rice germ shelf life.

Evaluating the possibility of extending shelf life of rice germ (a by-product of rice milling process) by reducing water activity in combination with storage atmosphere packaging, without any heat treatment, is the aim of the present study. Samples at different water activities (0.55, 0.45 and 0.36) were packed in air, argon or under vacuum, and stored at 27 °C for 150 days. To the aim, a non-targeted approach was applied by means of an FT-NIR spectrometer in reflectance with a rotating sample holder and a portable electronic nose, equipped with 10 non-specific sensors. For understanding the impact of the factors under study on the rice germ shelf life, a modified mid-level data fusion approach was applied to enhance the information most correlated with time. Moreover, Principal Component Analysis was applied on fused data to follow samples evolution during storage and identify different clusters according to the storage conditions. The rice germ case study allowed to better understand the information captured by the non-specific sensors: a 2D correlation map was developed combining the e-nose data with the NIR spectral information, highlighting relationships among NIR absorption bands and classes of chemical compounds inducing e-nose responses. A data fusion approach highlighted the importance of water activity on rice germ storage, while no interesting differences were ascribable to storage atmosphere packaging systems. In terms of correlation, the sensors could be divided in two groups, negatively inter-correlated: sensors ascribable to aromatic compounds (WC) and correlated with the NIR band around 4800-4900 cm-1 (N-H bending of primary amides, typical for peptides coming from protein hydrolysis); broad-range response sensors (WS), linked with the NIR band at 5128 cm-1 (second overtone of CO stretching of esters).

A new application of NIR spectroscopy to describe and predict purees quality from the non-destructive apple measurements.

The potential of NIRS was investigated on both apples and purees to (i) examine factors involving quality variability (variety, agricultural practice, cold storage, puree mechanical refining level) and (ii) establish the link between quality traits before and after processing in order to predict the quality characteristics of purees from spectral information of raw apples. Apples and purees were well-classified at over 82% and 88% according to varieties and storage times respectively. The PLS models showed a good ability to estimate puree characteristics from spectra acquired on corresponding apples such as viscosity (R2 > 0.82), cell wall content (R2 > 0.81) and also dry matter (R2 > 0.83), soluble solids content (R2 > 0.80) and titratable acidity (R2 > 0.80). NIR technique should be a useful tool for industry insofar as it can give a reliable assessment of texture and taste of the final products based on the non-destructive fresh materials evaluation.

Potential biomonitoring of atmospheric carbon dioxide in Coffea arabica leaves using near-infrared spectroscopy and partial least squares discriminant analysis.

The potencial of Coffea arabica leaves as bioindicators of atmospheric carbon dioxide (CO2) was evaluated in a free-air carbon dioxide enrichment (FACE) experiment by using near-infrared reflectance (NIR) spectroscopy for direct analysis and partial least squares discriminant analysis (PLS-DA). A supervised classification model was built and validated from the spectra of coffee leaves grown under elevated and current CO2 levels. PLS-DA allowed correct test set classification of 92% of the elevated-CO2 level leaves and 100% of the current-CO2 level leaves. The spectral bands accounting for the discrimination of the elevated-CO2 leaves were at 1657 and 1698 nm, as indicated by the variable importance in the projection (VIP) score together with the regression coefficients. Seven months after suspension of enriched CO2, returning to current-CO2 levels, new spectral measurements were made and subjected to PLS-DA analysis. The predictive model correctly classified all leaves as grown under current-CO2 levels. The fingerprints suggest that after suspension of elevated-CO2, the spectral changes observed previously disappeared. The recovery could be triggered by two reasons: the relief of the stress stimulus or the perception of a return of favorable conditions. In addition, the results demonstrate that NIR spectroscopy can provide a rapid, nondestructive, and environmentally friendly method for biomonitoring leaves suffering environmental modification. Finally, C. arabica leaves associated with NIR and mathematical models have the potential to become a good biomonitoring system.

Quality evaluation of decoction pieces of Rhizoma Atractylodis Macrocephalae by near infrared spectroscopy coupled with chemometrics.

OBJECTIVE: To establish a fast, simple and reliable method for quality evaluation of decoction pieces of Rhizoma Atractylodis Macrocephalae (referred as BZ below) by near infrared spectroscopy coupled with chemometrics. METHOD: Twelve batches of raw medicinal materials of BZ were collected from three main producing location in China. According to the Pharmacopoeia of the People's Republic of China, these raw decoction pieces were stir-fried in wheat bran using a stir-frying machine for 3, 6, 9, 12 and 15 min, respectively. The resulted 60 samples were categorized into three classes (i.e., light, moderate and dark) by experienced pharmacists according to their surface color. After that, these slices were smashed to acquire near infrared spectra and to determine the contents of atractylenolide I, II and III by HPLC method. Qualitative and quantitative models were constructed to relate the spectra to the color labels and to the contents of three atractylenolides. Various chemometrics methods, including calibration methods like principal component analysis, partial least squares discriminant analysis (PLS-DA) and partial least squares regression (PLSR), spectra pretreatment methods like standard normal variate, multiplicative scatter correction, derivation and smoothing, feature selection methods like particle swarm optimization, genetic algorithm (GA) and other fourteen methods were compared in detail. The PLS-DA models were evaluated by jackknife tests with calculating parameters such as error rate (ERR), true positive rate (TPR), true negative rate (TNR) and F1 score, meanwhile the PLSR models were evaluated by five fold cross-validation tests with calculating parameters such as coefficients of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and residual predictive deviation (RPD). RESULTS: The PLS-DA models with spectra pretreated by 1D5S or 1D9S and wavelengths selected by InfFS, Relief-F, MutInfFS, fisher or CFS performed best, yielding 0.00 of ERR, 1.00 of TPR, 1.00 of TNR, and 1.00 of F1 for all three classes. As for quantitative models, the PLSR models by 1D5S spectra pretreatment and GA wavelengths selection performed best, where R2C and R2P were all >0.95, RMSEC and RMSEP were all <0.04%, MAEC and MAEP were all <0.04%, and RPD were all >5. CONCLUSION: The present qualitative and quantitative models can be successfully used to distinguish the degree of suitability of processed BZ, and to determine the contents of three atractylenolides, which thus are of great help for quality evaluation and control of processed BZ and other decoction pieces.

Attention based residual network for medicinal fungi near infrared spectroscopy analysis.

As an effective technology, near infrared spectroscopy (NIRS) can be widely applied to analysis of active ingredients in medicinal fungi. Multiple regression methods are used to compute the relationship between spectral vectors and ingredient contents. In this paper, an autonomous feature extraction method by using attention based residual network (ABRN) to model original NIRS vectors is introduced. Attention module in ABRN is employed to enhance feature wave bands and to decay noise. Different from traditional NIRS analysis methods, ABRN does not require any preprocessing of artificial feature selections which rely on expert experience. The experiments test ABRN by analyzing original spectrums of medicinal fungi (Antrodia Camphorata and Matsutake), which are from 800 nm to 2500 nm, and predicting active ingredients within them. We compare ABRN with other popular NIRS analysis methods. The root mean square error of Antrodia Camphorata training set (RMSET) and validation set (RMSEV) are 0.0229 g⋅g⁻¹ and 0.0349 g⋅g⁻¹ for polysaccharide, and 0.0173 g⋅g⁻¹ and 0.0189 g⋅g⁻¹ for triterpene. The RMSET and RMSEV of Matsutake are 0.1343 g⋅g⁻¹ and 0.2472 g⋅g⁻¹ for polysaccharide, and 0.0328 g⋅g⁻¹ and 0.0445 g⋅g⁻¹ for ergosterol. The (coefficient of determination) of these four ingredients are 0.711, 0.753, 0.847 and 0.807. The results indicate that ABRN has better performance in autonomously extracting feature wave bands from original NIRS vectors, which can decrease the loss of tiny feature peaks.

Express detection of expired drugs based on near-infrared spectroscopy and chemometrics: A feasibility study.

Levofloxacin is a third-generation fluoroquinolone antimicrobials drug that inhibits bacterial DNA replication. Driven by huge profit, one kind of particular counterfeit, e.g., repackaged expired tablets, becomes very common especially in developing countries. The feasibility of identifying expired levofloxacin tablets by combining NIR spectroscopy with chemometrics was investigated. Five kinds of levofloxacin samples from different manufacturers were collected for experiment. Two types of expired mode were considered and a simple model-independent algorithm was used for feature selection. Principal component analysis (PCA) was used for exploratory analysis and simple discriminant analysis. Taking seventy samples as the target class, a final one-class model based on Data Driven Soft Independent Modeling by Class Analogy with abbreviation DD-SIMCA was constructed, which achieved 97% sensitivity and 100% specificity on the independent set composed of 34 unexpired and 128 expired tablets. These results confirm that the combination of NIR spectroscopy, feature selection and class-modeling is feasible for identifying the expired levofloxacin tablets. Such a method can be extended to the analysis of similar drugs.

Comparison of Near-Infrared Spectroscopy with Needle Indentation and Histology for the Determination of Cartilage Thickness in the Large Animal Model Sheep.

The suitability of near-infrared spectroscopy (NIRS) for non-destructive measurement of cartilage thickness was compared with the gold standard needle indentation. A combination of NIRS and biomechanical indentation (NIRS-B) was used to address the influence of varying loads routinely applied for hand-guided NIRS during real-life surgery on the accuracy of NIRS-based thickness prediction. NIRS-B was performed under three different loading conditions in 40 osteochondral cylinders from the load-bearing area of the medial and lateral femur condyle of 20 cadaver joints (left stifle joints; female Merino sheep; 6.1 ± 0.6 years, mean ± standard error of the mean). The cartilage thickness measured by needle indentation within the region analyzed by NIRS-B was then compared with cartilage thickness prediction based on NIRS spectral data using partial least squares regression. NIRS-B repeat measurements yielded highly reproducible values concerning force and absorbance. Separate or combined models for the three loading conditions (the latter simulating load-independent measurements) resulted in models with optimized quality parameters (e.g., coefficients of determination R2 between 92.3 and 94.7) and a prediction accuracy of < 0.1 mm. NIRS appears well suited to determine cartilage thickness (possibly in a hand-guided, load-independent fashion), as shown by high reproducibility in repeat measurements and excellent reliability compared with tissue-destructive needle indentation. This may provide the basis for non-destructive, intra-operative assessment of cartilage status quo and fine-tuning of repair procedures.

Quantification of mineral composition of Brazilian bee pollen by near infrared spectroscopy and PLS regression.

Bee pollen consumption has increased in the last years, mainly due to its nutritional value and therapeutic applications. The quantification of mineral constituents is of great importance in order to evaluate both, the toxicity and the beneficial effect of essential elements. The purpose of this work was to quantify the essential elements, Ca, Mg, Zn, P and K, by diffuse reflectance spectra in the near infrared region (NIR) combined with partial least squares regression (PLS), which is a clean and fast method. Reference method used was ICP OES. The determination coefficients for calibration models (R2) were above 0.87 and the mean percent calibration error varied from 5 to 10%. For external validation R2 values were higher than 0.76. The results indicated that NIR spectroscopy can be useful for an approximate quantification of these minerals in bee pollen samples and can be used as a faster alternative to the standard methodologies.

Rapid determination of phytosterols by NIRS and chemometric methods.

Phytosterols have been extensively studied because it plays essential roles in the physiology of plants and can be used as nutritional supplement to promote human health. We use a rapid method by coupling near-infrared spectroscopy (NIRS) and chemometric techniques to quickly and efficiently determine three essential phytosterols (ß-sitosterol, campesterol and stigmasterol) in vegetable oils. Continuous wavelet transform (CWT) method was adopted to remove the baseline shift in the spectra. The quantitative analysis models were constructed by partial least squares (PLS) regression and randomization test (RT) method was used to further improve the models. The optimized models were used to calculate the phytosterol contents in prediction set in order to evaluate their predictability. We have found that the phytosterol contents obtained by the optimized models and Gas Chromatography/Mass Spectrometry (GC/MS) analysis are almost consistent. The root mean square error of prediction (RMSEP) and ratio of prediction to deviation (RPD) for the three phytosterols are 525.7590, 212.2245, 65.1611 and 4.0060, 4.7195 and 3.5441, respectively. The results have proved the feasibility of the proposed method for rapid and non-destructive analysis of phytosterols in edible oils.

Quantifying several adulterants of notoginseng powder by near-infrared spectroscopy and multivariate calibration.

The authentication of traditional Chinese medicine (TCM) is critically important for public-health and economic terms. Notoginseng, a classical TCM of high economic and medical value, could be easily adulterated with Sophora flavescens powder (SFP), corn flour (CF) or other analogues of low-grade (ALG) because of their similar tastes, appearances and much lower cost. The main objective of this study was to evaluate the feasibility of applying of near-infrared (NIR) spectroscopy and multivariate calibration for identifying and quantifying several common adulterants in notoginseng powder. Two datasets were prepared for experiment. The competitive adaptive reweighted sampling (CARS) was used to select informative variables. Two different schemes were used for sample set partition. Model population analysis (MPA) was made. The results showed that, the constructed partial least squares (PLS) model using a reduced set of variables from CARS can provide superior performance to the full-spectrum PLS model. Also, the sample set partition is very of great importance. It seems that the combination of NIR spectroscopy, CARS and PLS is feasible to quantify common adulterants in notoginseng powder.

Feasibility study of smartphone-based Near Infrared Spectroscopy (NIRS) for salted minced meat composition diagnostics at different temperatures.

This research work evaluates the feasibility of a smartphone-based spectrometer (740-1070 nm) for salted minced meat composition diagnostics at industrial scale. A commercially available smartphone-based spectrometer and a benchtop NIR spectrometer (940-1700 nm) were used for acquiring 1312 spectra from meat samples stored at four different temperatures ranging from -14 °C to 25 °C. Thereafter, for each spectrometer, PLS and Random Forest regression models specific for each temperature and global models were created to predict the fat, moisture and protein contents. Fat and moisture can be estimated with the global model in a wide range of temperatures by using the smartphone-based spectrometer, which has an acceptable accuracy for quality control purposes (RPD > 7) and comparable to the accuracy of a benchtop spectrometer.

Tea types classification with data fusion of UV-Vis, synchronous fluorescence and NIR spectroscopies and chemometric analysis.

The potential of selected spectroscopic methods - UV-Vis, synchronous fluorescence and NIR as well a data fusion of the measurements by these methods - for the classification of tea samples with respect to the production process was examined. Four classification methods - Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), Regularized Discriminant Analysis (RDA) and Support Vector Machine (SVM) - were used to analyze spectroscopic data. PCA analysis was applied prior to classification methods to reduce multidimensionality of the data. Classification error rates were used to evaluate the performance of these methods in the classification of tea samples. The results indicate that black, green, white, yellow, dark, and oolong teas, which are produced by different methods, are characterized by different UV-Vis, fluorescence, and NIR spectra. The lowest error rates in the calibration and validation data sets for individual spectroscopies and data fusion models were obtained with the use of the QDA and SVM methods, and did not exceed 3.3% and 0.0%, respectively. The lowest classification error rates in the validation data sets for individual spectroscopies were obtained with the use of RDA (12,8%), SVM (6,7%), and QDA (2,7%), for the UV-Vis, SF, and NIR spectroscopies, respectively. NIR spectroscopy combined with QDA outperformed other individual spectroscopic methods. Very low classification errors in the validation data sets - below 3% - were obtained for all the data fusion data sets (SF + UV-Vis, SF + NIR, NIR + UV-Vis combined with the SVM method). The results show that UV-Vis, fluorescence and near infrared spectroscopies may complement each other, giving lower errors for the classification of tea types.