Paul Geladi

Techniques and applications of hyperspectral image analysis

Preface. List of Contributors. List of Abbreviations. 1 Multivariate Images, Hyperspectral Imaging: Background and Equipment (Paul L. M. Geladi, Hans F. Grahn and James E. Burger). 2 Principles of Multivariate Image Analysis (MIA) in Remote Sensing, Technology and Industry (Kim H. Esbensen and Thorbjorn T. Lied). 3 Clustering and Classification in Multispectral Imaging for Quality Inspection of Postharvest Products (Jacco C. Noordam and Willie H. A. M. van den Broek). 4 Self-modeling Image Analysis with SIMPLISMA (Willem Windig, Sharon Markel and Patrick M. Thompson). 5 Multivariate Analysis of Spectral Images Composed of Count Data (Michael R. Keenan). 6 Hyperspectral Image Data Conditioning and Regression Analysis (James E. Burger and Paul L. M. Geladi). 7 Principles of Image Cross-validation (ICV): Representative Segmentation of Image Data Structures (Kim H. Esbensen and Thorbjorn T. Lied). 8 Detection, Classification, and Quantification in Hyperspectral Images Using Classical Least Squares Models (Neal B. Gallagher). 9 Calibration Standards and Image Calibration (Paul L. M. Geladi). 10 Multivariate Movies and their Applications in Pharmaceutical and Polymer Dissolution Studies (Jaap van der Weerd and Sergei G. Kazarian). 11 Multivariate Image Analysis of Magnetic Resonance Images: Component Resolution with the Direct Exponential Curve Resolution Algorithm (DECRA) (Brian Antalek, Willem Windig and Joseph P. Hornak). 12 Hyperspectral Imaging Techniques: an Attractive Solution for the Analysis of Biological and Agricultural Materials (Vincent Baeten, Juan Antonio Fernandez Pierna and Pierre Dardenne). 13 Application of Multivariate Image Analysis in Nuclear Medicine: Principal Component Analysis (PCA) on Dynamic Human Brain Studies with Positron Emission Tomography (PET) for Discrimination of Areas of Disease at High Noise Levels (Pasha Razifar and Mats Bergstrom). 14 Near Infrared Chemical Imaging: Beyond the Pictures (E. Neil Lewis, Janie Dubois, Linda H. Kidder and Kenneth S. Haber). Index.

Chemometrics in spectroscopy. Part 1. Classical chemometrics

Abstract An overview is given of chemometrics as it can be applied to spectroscopic and other multivariate data. Major chemometrics and data analysis techniques are described. An important aspect is the focus on soft modeling for situations that are too complicated for the traditional hard models to work. Also measurement noise is given due attention. A small example is used to illustrate some ways of working, mainly by using graphics. Selected literature references are given. Part 1 deals with classical chemometrics. Part 2 presents some newer developments and includes some more elaborated examples.

Maize kernel hardness classification by near infrared (NIR) hyperspectral imaging and multivariate data analysis.

The use of near infrared (NIR) hyperspectral imaging and hyperspectral image analysis for distinguishing between hard, intermediate and soft maize kernels from inbred lines was evaluated. NIR hyperspectral images of two sets (12 and 24 kernels) of whole maize kernels were acquired using a Spectral Dimensions MatrixNIR camera with a spectral range of 960-1662 nm and a sisuChema SWIR (short wave infrared) hyperspectral pushbroom imaging system with a spectral range of 1000-2498 nm. Exploratory principal component analysis (PCA) was used on absorbance images to remove background, bad pixels and shading. On the cleaned images, PCA could be used effectively to find histological classes including glassy (hard) and floury (soft) endosperm. PCA illustrated a distinct difference between glassy and floury endosperm along principal component (PC) three on the MatrixNIR and PC two on the sisuChema with two distinguishable clusters. Subsequently partial least squares discriminant analysis (PLS-DA) was applied to build a classification model. The PLS-DA model from the MatrixNIR image (12 kernels) resulted in root mean square error of prediction (RMSEP) value of 0.18. This was repeated on the MatrixNIR image of the 24 kernels which resulted in RMSEP of 0.18. The sisuChema image yielded RMSEP value of 0.29. The reproducible results obtained with the different data sets indicate that the method proposed in this paper has a real potential for future classification uses.

High quality biofuel pellet production from pre-compacted low density raw materials

In this study, pre-compaction was evaluated as a method to enhance stable reed canary grass pellet production. An experimental design of the factors raw material moisture content, steam addition, raw material bulk density, and die temperature was used to find production conditions for high quality pellets by multiple linear regression modelling of responses. Response variables being modelled were variability of pelletizer current (as a measurement of uneven production), pellet bulk density, and pellet durability. By pre-compacting the raw material from a bulk density of 150 kg/m3 to 270kg/m3, continuous production could be obtained at minimum raw material moisture content of 13.8%. Bulk density and durability were both highly correlated to raw material moisture content, but showed different optima. Multiple response optimization was used to target process settings for production of high quality reed canary grass pellets with bulk density >650kg/m3 and durability >97.5%.

Emission of Volatile Aldehydes and Ketones from Wood Pellets under Controlled Conditions

Different qualities of biofuel pellets were made from pine and spruce sawdust according to an industrial experimental design. The fatty/resin acid compositions were determined by gas chromatography-mass spectrometry for both newly produced pellets and those after 2 and 4 weeks of storage. The aldehydes/ketones compositions were determined by high performance liquid chromatography at 0, 2, and 4 weeks. The designs were analyzed for the response variables: total fatty/resin acids and total aldehydes/ketones. The design showed a strong correlation between the pine fraction in the pellets and the fatty/resin acid content but the influence decreased over storage time. The amount of fatty/resin acids decreased approximately 40% during 4 weeks. The influence of drying temperature on the aldehyde/ketone emission of fresh pellets was also shown. The amounts of emitted aldehydes/ketones generally decreased by 45% during storage as a consequence of fatty/resin acid oxidation. The matrices of individual concentrations were subjected to multivariate data analysis. This showed clustering of the different experimental runs and demonstrated the important mechanism of fatty/resin acid conversion.

Congener fingerprints of tetra- through octa-chlorinated dibenzo-p-dioxins and dibenzofurans in Baltic surface sediments and their relations to potential sources

Comprehensive congener fingerprints of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs), including non-2,3,7,8-substituted congeners, in 142 surface sediment samples from the Baltic Sea were characterized by Principal Component Analysis (PCA). The pattern analysis revealed source specific indicators of chlorophenol use, high temperature processes, chlorine bleach/chloralkali production and a source signature suggested to originate from pulp/paper or related production. Congener patterns in sediments from offshore and pristine coastal areas showed strong resemblance to patterns of atmospheric deposition and flue gases, indicating that these sources have high impact in areas that are not affected by point sources. Prominent contributors to the patterns of hotspot areas along the Swedish coast included chlorophenol indicators and a source characterized by hexa-CDDs while the contribution of the traditional chlorine bleach pattern was weaker. This study demonstrates the importance of comprehensive PCDD/F congener analysis for identifying links to candidate sources.

Tracking diffusion of conditioning water in single wheat kernels of different hardnesses by near infrared hyperspectral imaging.

The combination of near infrared (NIR) hyperspectral imaging and chemometrics was used to follow the diffusion of conditioning water over time in wheat kernels of different hardnesses. Conditioning was attempted with deionised water (dH(2)O) and deuterium oxide (D(2)O). The images were recorded at different conditioning times (0-36 h) from 1000 to 2498 nm with a line scan imaging system. After multivariate cleaning and spectral pre-processing (either multiplicative scatter correction or standard normal variate and Savitzky-Golay smoothing) six principal components (PCs) were calculated. These were studied visually interactively as score images and score plots. As no clear clusters were present in the score plots, changes in the score plots were investigated by means of classification gradients made within the respective PCs. Classes were selected in the direction of a PC (from positive to negative or negative to positive score values) in almost equal segments. Subsequently loading line plots were used to provide a spectroscopic explanation of the classification gradients. It was shown that the first PC explained kernel curvature. PC3 was shown to be related to a moisture-starch contrast and could explain the progress of water uptake. The positive influence of protein was also observed. The behaviour of soft, hard and very hard kernels was different in this respect, with the uptake of water observed much earlier in the soft kernels than in the harder ones. The harder kernels also showed a stronger influence of protein in the loading line plots. Difference spectra showed interpretable changes over time for water but not for D(2)O which had a too low signal in the wavelength range used. NIR hyperspectral imaging together with exploratory chemometrics, as detailed in this paper, may have wider applications than merely conditioning studies.

Indirect detection of Fusarium verticillioides in maize ( Zea maize L.) kernels by NIR hyperspectral imaging

Near infrared (NIR) hyperspectral imaging and hyperspectral image analysis were evaluated for their potential to distinguish between Fusarium verticillioides infected and sound whole maize (Zea mays L.) kernels. Hyperspectral images of infected and sound kernels were acquired using a Spectral Dimensions MatrixNIR camera with a spectral range of 960–1662nm and a sisuChema hyperspectral pushbroom imaging system with a spectral range of 1000–2498 nm. Background, bad pixels and shading were removed using exploratory principal component analysis (PCA) on absorbance images. PCA could be used effectively on the cleaned images to identify classes including infected and non-infected regions on individual kernels. A distinct difference between infected and sound kernels along principal component (PC) one with two distinguishable clusters was found. The loading line plot of the first PC of the sisuChema hypercube showed important absorbance peaks for the two classes, i.e. 1960nm and 2100nm for the infected class and U50 nm, 2300nm and 2350nm for the non-infected class. Partial least squares discriminant analysis (PLS-DA) was applied. The coefficient of determination was 0.73 for the MatrixNIR image and 0.86 for the sisuChema image, both after three PLS components. These PLS-DA models could be used to calculate predictions on a test set image. The predictions could be shown as prediction images and an acceptable root mean square error of prediction was obtained (0.23). NIR hyperspectral imaging has the potential to be used as a rapid, objective means of indentifying fungal infected maize kernels and infected regions.

Large Uptake of Titania and Iron Oxide Nanoparticles in the Nucleus of Lung Epithelial Cells as Measured by Raman Imaging and Multivariate Classification

It is a challenging task to characterize the biodistribution of nanoparticles in cells and tissue on a subcellular level. Conventional methods to study the interaction of nanoparticles with living cells rely on labeling techniques that either selectively stain the particles or selectively tag them with tracer molecules. In this work, Raman imaging, a label-free technique that requires no extensive sample preparation, was combined with multivariate classification to quantify the spatial distribution of oxide nanoparticles inside living lung epithelial cells (A549). Cells were exposed to TiO2 (titania) and/or α-FeO(OH) (goethite) nanoparticles at various incubation times (4 or 48 h). Using multivariate classification of hyperspectral Raman data with partial least-squares discriminant analysis, we show that a surprisingly large fraction of spectra, classified as belonging to the cell nucleus, show Raman bands associated with nanoparticles. Up to 40% of spectra from the cell nucleus show Raman bands associated with nanoparticles. Complementary transmission electron microscopy data for thin cell sections qualitatively support the conclusions.

Electrochemical impedance spectroscopy in label-free biosensor applications: multivariate data analysis for an objective interpretation.

Electrochemical impedance spectroscopy plays an important role in biosensor science thanks to the possibility of finding specific information from processes with different kinetics at a chosen electrode potential in one experiment. In this paper we briefly discuss label-free impedimetric biosensors described in the literature. A novel method for neutral interpretation of impedance data is presented that includes complex number chemometrics. Three examples are given based on impedance measurements on synthetic biomembranes, in this case a lipid monolayer deposited on a mercury electrode. The interaction of various compounds with the monomolecular lipid layer is illustrated with the following: (1) different concentrations of magainin (Geladi et al. in Proc. Int. Fed. Med. Biomed. Eng. 9:219–220, 2005); (2) different derivatives of gramicidin A (Lindholm-Sethson et al. in Langmuir 24:5029–5032, 2007), and (3) an antimicrobial peptide (Ringstad et al. in Langmuir 24:208–216, 2008).