Multivariate Image Regression Research Articles

Microalgae Monitoring in Microscale Photobioreactors via Multivariate Image Analysis

Microscale photobioreactors for microalgae growth represent an interesting technology for fast data production and biomass characterization; however, the small scale poses severe monitoring challenges, as traditional methods cannot be used. Non-invasive techniques are therefore needed to quantify biomass concentration and other culture properties, for example, pigment composition. To this purpose, a soft sensing approach based on multivariate image regression is proposed to exploit RGB images and/or PAM-imaging chlorophyll fluorescence. Different PLS (Partial Least Squares) regression models are used to estimate: (a) biomass concentration from the features extracted by RGB indices and/or PAM-imaging chlorophyll fluorescence measurements; and (b) Chlorophyll a content per cell from the features extracted by RGB indices and biomass concentration measurements. Every single model is aimed at characterizing the microalgae culture at different light intensities during batch growth. Results show that the proposed monitoring approach is as accurate as traditional measurement approaches and may represent a promising methodology for fast and inexpensive monitoring of microscale photobioreactors.

Spatially resolved spectral determination of polysaccharides in hydrothermally carbonized biomass

Hyperspectral imaging and multivariate image regression can replace expensive and time-consuming laboratory analyses of biomass and hydrochar polysaccharides.

Determination of allura red dye in hard candies by using digital images obtained with a mobile phone and N-PLS

This paper describes the development of an optical sensor device using a smartphone and a homemade dark chamber built with recycled materials. This low cost instrument was employed in the development of multivariate image regression methods for the determination of the azo dye allura red in hard candies. To build the models, 238 candy samples of four flavors and different brands and batches were used. Firstly, a multivariate calibration model using RGB histograms and partial least squares (PLS) was built. This model provided high prediction errors, which were attributed to the presence of textural variations in the images. Then, a more complex image analysis methodology that incorporates spatial information, and consists of preprocessing by a two-dimensional fast Fourier transform followed by multi-way calibration with N-way PLS, provided better results, decreasing the prediction errors around 25–35%. The final model was submitted to a complete multivariate analytical validation, being considered precise, linear, sensitive and unbiased. The analytical range was established between 22.9 and 78.8 mg kg−1 of allura red. Root mean square errors of calibration (RMSEC) and prediction (RMSEP) of 4.8 and 6.1 mg kg−1 were estimated. The developed method is simple, rapid, and nondestructive.

Multivariate Image Regression for Quality Control of Natural Fiber Composites

This study reports on using near-infrared (NIR) hyperspectral imaging for nondestructive spectral-spatial characterization of natural fiber composites. To illustrate the approach, maleic anhydride grafted polyethylene (MAPE)/hemp fiber composites were produced with different filler contents between 0 and 60%. Two different chemometrics methods based on (1) the traditional multivariate PLS calibration and (2) multivariate image analysis and regression (MIA/MIR) were tested to predict tensile properties using NIR images. The results show good agreement between the measured properties and their predictions by both of these methods. The MIR ability to map chemical composition was compared to that of multivariate curve resolution (MCR) and the results were found similar. The proposed MIR approach was found very promising for quality control of polymer composites because of its combined ability to quantify filler content and dispersion within the material, to distinguish compositional variations from physical defects, and to predict the end-user properties of the product all with a single MIR model.

Relevant aspects of quantification and sample heterogeneity in hyperspectral image resolution

NIR-infrared chemical imaging (NIR-CI) is a powerful technique to provide spatial and spectral information about the samples analyzed. NIR-CI presents a big potential to obtain accurate and reliable information about the quality of end products and it is also an excellent tool for process control.Quantitative analysis and heterogeneity studies on NIR images are often carried out by multivariate calibration techniques, designed in this context as Multivariate Image Regression (MIR). Multivariate Curve Resolution—Alternating Least Squares (MCR-ALS) is another data analysis method, aimed mainly at recovering the pure spectra and distribution maps of images. This straightforward information can be potentially applied for quantitative purposes and heterogeneity description when image multiset structures are analyzed.In this work, the potential of MCR-ALS to provide quantitative and heterogeneity information has been explored paying special attention to particular aspects, such as the preprocessing used for resolution and the effect of several factors on the global quantitative results, namely: a) the design of the regions of interest (ROIs) included in the multiset image structure, b) the constraints applied in the resolution step and c) the calibration/validation strategy applied. Quantitative information at a pixel level has also been carried out to study the heterogeneity of the samples analyzed, stressing the difference between the constitutional heterogeneity (population statistics describing the quantitative distribution or pixel-to-pixel variability in concentration) and distributional heterogeneity (image maps expressing the variability in the spatial distribution of compounds in the image).To perform the work described above, real NIR images of non-homogeneous mixtures of acetylsalicilyc, caffeine and starch of different compositions have been used.

Multivariate image analysis: A review with applications

Abstract Nowadays, image analysis is becoming more important because of its ability to perform fast and non-invasive low-cost analysis on products and processes. Image analysis is a wide denomination that encloses classical studies on gray scale or RGB images, analysis of images collected using few spectral channels (sometimes called multispectral images) or, most recently, data treatments to deal with hyperspectral images, where the spectral direction is exploited in its full extension. Pioneering data treatments in image analysis were applied to simple images mainly for defect detection, segmentation and classification by the Computer Science community. From the late 80s, the chemometric community joined this field introducing powerful tools for image analysis, which were already in use for the study of classical spectroscopic data sets and were appropriately modified to fit the particular characteristics of image structures. These chemometric approaches adapt to images of all kinds, from the simplest to the hyperspectral images, and have provided new insights on the spatial and spectroscopic information of this kind of data sets. New fields open by the introduction of chemometrics on image analysis are exploratory image analysis, multivariate statistical process control (monitoring), multivariate image regression or image resolution. This paper reviews the different techniques developed in image analysis and shows the evolution in the information provided by the different methodologies, which has been heavily pushed by the increasing complexity of the image measurements in the spatial and, particularly, in the spectral direction.

Three dimensional image analysis for prediction of vinyl triethoxysilane concentration and porosity percent of methyl methacrylate/ethylene glycol dimethacrylate copolymers using scanning electron microscopy

In this work, scanning electron micrographs of copolymer beads based on methyl methacrylate (MMA), ethylene glycol dimethacrylate (EGDM) and vinyl triethoxysilane (VTES) were used to construct multivariate image regression models. Scanning electron microscopy (SEM) and swelling measurements indicated that increasing the silicone concentration (VTES) of the copolymers enhances the amount of pores and porosity percent. Scanning electron micrographs of the synthesized copolymers with different VTES concentrations were used to construct two soft models, which were able to relate the VTES concentration and the porosity percent of the copolymers with their SEM images. The constructed models were the three-way partial least squares analysis (N-PLS) and unfolded principal component regression (unfold-PCR). The predictive ability of the constructed models were evaluated by a test set. The results showed the models were able to effectively predict the VTES concentration and the porosity percent of the copolymers using their corresponding SEM images.

Principles of MIR, multivariate image regression: I: Regression typology and representative application studies

We present an introduction to Multivariate Image Regression (MIR) with a selection of illustrative application studies. Generalisation from two-way multivariate calibration to the three-way regimen leads to—at least—three alternative image regression cases depending on the nature of the available Y-data: IPLS-Y discrim; IPLS-Y grid; IPLS-Y total. A systematic image regression typology is briefly introduced. We here present the core of the principles of applied MIR. Two major MIR application studies are worked through, a food mass product industrial inspection study (IPLS-Y discrim) and a food product (fruit) storage stability image analytical monitoring (IPLS-Y grid). These exemplifications are presented as archetypes, representing a much wider range of potential industrial/technological application areas. Based on simple three-channel imagery (in order to simulate many industrial systems), they nevertheless represent all higher-dimensional multivariate image cases as well, since the pertinent MIR principles and software are invariant w.r.t. any number of channels/variables employed. The present paper represents one major element of our work towards establishing a complete, stand-alone facility for Multivariate Image Regression (MIR); the second paper in this series deals with the development, implementation and extensive exemplifications of a complementary cross-validation facility.

Mapping Lipid Oxidation in Chicken Meat by Multispectral Imaging of Autofluorescence

Multispectral imaging of autofluorescence was carried out to investigate the feasibility of mapping the degree of lipid oxidation in ground chicken meat. Meat samples from both breast and thighs were collected from 32 chickens, ground and freeze-stored for different time intervals. Sixteen samples were imaged at the time, making up two sets of multispectral images, A and B. Lipid oxidation was measured by a method using 2-thiobarbituric acid reactive substances (TBARS), and samples were in the range 0.15–3.23. Principal component analysis was performed on image set A, and variation in score images of the two first components corresponded well with lipid oxidation reference values. The multivariate image regression model based on image set A was tested on set B. Pixel-wise prediction gave large individual errors, but averaging predicted values within samples improved accuracy and resulted in a correlation of 0.98. Increasing the amount of spatial variation (number of pixel vectors) in the regression models led to more robust models with lower prediction errors. The technique has potential for nondestructive investigation of distribution and kinetics of lipid oxidation in food.

Multivariate image regression and analysis: Useful techniques for the evaluation of clinical magnetic resonance images

Multivariate image analysis (MIA) and multivariate image regression (MIR) techniques are useful tools in the extraction of information from magnetic resonance images. They aid the characterization of different tissues and can be used to describe their size and distribution. The obtained new information can be used to monitor growth, progression and effects of a treatment. The methodology is illustrated by a clinical example. The ongoing development of MIA and MIR, combining parameters from different imaging modalities, is commented on.

Strategies for multivariate image regression

Esbensen, K.H., Geladi, P.L. and Grahn, H.F., 1992. Strategies for multivariate image regression. Chemometrics and Inteligent Laboratory Systems, 14: 357–374. We present multivariate image regression (MIR) as a set of typically problem-dependent strategies for image decomposition guided by the nature of the Y variable and/or training data set delineation in the ( X, Y) image domains. Regression techniques common in chemometrics may be applied also to the image regimen (in this paper we treat mainly two-dimensional images). We present applications of both IMPCR and IMPLS-DISCRIM in an effort to delineate the various possibilities for image regression. IMPCR builds directly on our earlier bilinear multivariate image analysis projection approach, while IMPLS-DISCRIM is trained on scene space binary classification masking with subsequent off-screen partial least squares analysis; the results are back-projected as images in the original scene space. Regression may either be carried out for modelling purposes and/or for subsequent prediction purposes. In the image domain this duality is accompanied by several optional training data set delineations in the scene space and/or in the spectral domain. We try to cover as complete a survey as possible of typical, representative regression problem types. We illustrate some of these MIR strategies with an MR-imaging example as well as a simple didactic MIR calibration from analytical chemistry.