Some special topics in multivariate image analysis

A simple imaging system has been developed for acquiring multivariate images in order to characterise the heterogeneity of food materials. The objective of the present work is, first, to demonstrate the capability of this acquisition system to discriminate food products of different natures. Secondly, our goal is to apply Partial Least Squares regression on these multivariate images and to evaluate the interest of various strategies of classification. A data set containing 24 images (702 × 524) acquired at different wavelengths for four food products is analysed. After the establishment of the PLS2 models employed for predicting the indicator variables, four strategies of classification of observations are tested. The first classification is done by selecting the largest component of the indicator variables. The others are based on the measurement of distances to the barycentres of the qualitative groups. Distances calculated can be either Euclidian distances or Mahalonobis distances. Except the strategy based on the Euclidian distance on scores, the strategies are rather equivalent, with a slight advantage to the Euclidian distance on predicted indicators. Another possibility addressed by the use of linear discriminant analysis (LDA) on multivariate images is to represent the qualitative groups as artificial images. The largest confusion appears between both cereal products while others are well classified. Copyright © 2006 John Wiley & Sons, Ltd.

Application of partial least squares and radial basis function neural networks in multivariate imaging analysis-quantitative structure activity relationship: Study of cyclin dependent kinase 4 inhibitors

Multivariate image analysis (MIA) aims to extract useful information from multivariate image structures such as hypercubes. Multivariate image regression is particularly useful for the development of calibration models for quantitative analysis of hypercubes. The key step in multivariate calibration is model selection; this includes the selection of method (e.g. partial least squares or ridge regression) and setting of parameters within the selected method (e.g. number of latent variables or ridge parameter). Sub-optimal selection of method and or parameters generally leads to poor predictive performance on future samples. While conventional methods such as cross-validation are indispensible in this task, the inherent nature of hyperspectral images may also be exploited. Application of a given calibration model to each pixel in a hypercube results in a prediction map whose features may be used to evaluate the model itself. This paper explores the application of various image statistics extracted from prediction images as model performance metrics for MIA.

Multi-way methods in image analysis—relationships and applications

An Improved Multivariate Image Analysis Method for Quality Control of Nanofiber Membranes

2D-Discrete Fourier Transform: Generalization of the MIA-QSAR strategy in molecular modeling

Robustness of multivariate image analysis assessed by resampling techniques and applied to FDG-PET scans of patients with Alzheimer's disease

In several process industries including mineral processing, where the materials are solids or slurries, some important measurements cannot be obtained using standard instrumentation (e.g., flow, temperature, pressure, pH, power draw, etc.), but can be visually appraised, and could be automatically quantified using machine vision techniques. In general, the information to extract from process images is not well defined and is stochastic in nature compared with those typically encountered in the manufacturing industry for automatic inspection, which have well defined deterministic features. Multivariate image analysis (MIA) as well as Multiresolution analysis (MRA) have been shown to be very efficient methods for spectral/textural analysis of process images. The objective of this chapter is to illustrate these methods using three mineral processing problems: (1) on-line estimation of froth grade in a flotation process; (2) flotation froth health monitoring for appropriate reagent dosage; and (3) on-line estimation of run-of-mine ore lithotype composition on conveyor belts. In all cases, the extracted image information could be used for developing new vision sensors for advanced control of mineral processing plants.

We recently established a high-speed, label-free, spectral imaging method based on stimulated Raman scattering (SRS). This method enables examination of cellular features within relatively short periods, thus enabling new imaging applications in pathology. Previously, we reported on label-free visualization of mouse tissue using SRS spectral microscopy combined with multivariate image analysis, but the feasibility of applying this approach to diseased tissues with diverse morphology and irregular chemical species has not been examined. We, therefore, assessed acetaminophen-induced liver injury to evaluate the potential use of Raman spectral microscopy for visualizing histopathologic specimens. Acetaminophen-overdosed mouse liver was prepared and the pathologic changes including centrilobular necrosis were confirmed. Multi-colored images were reconstructed through principal component analysis (PCA) of a multi-band SRS dataset, which provided rich information compared with a monochrome single-band SRS dataset. A wide view of the multi-colored principal component (PC) images showed the distribution of cellular constituents, which was similar to that observed by fat staining. In addition, different types of cells in liver parenchyma were also demonstrated. In conclusion, the combination of SRS spectral microscopy and PCA has the potential to reveal both the morphological and chemical features of specimens and therefore has potential utility in diagnostic pathology.

Learning classifiers from imbalanced data is still one of challenging tasks in machine learning and data mining. Data difficulty factors referring to internal and local characteristics of class distributions deteriorate performance of standard classifiers. Many of these factors may be approximated by analyzing the neighbourhood of the learning examples and identifying different types of examples from the minority class. In this paper, we follow recent research on developing such methods for assessing the types of examples which exploit either k-nearest neighbours or kernels. We discuss the approaches to tune the size of both kinds of neighborhoods depending on the data set characteristics and evaluate their usefulness in series of experiments with real-world and synthetic data sets. Furthermore, we claim that the proper analysis of these neighborhoods could be the basis for developing new specialized algorithms for imbalanced data. To illustrate it, we study generalizations of over-sampling in pre-processing methods and neighbourhood based ensembles.

MIA and NIR Chemical Imaging for pharmaceutical product characterization

Integrated use of ultrasound imaging and multivariate image analysis for detecting bone fragments in poultry meat.

In this paper, we investigate the possibilities offered by the extension of the connected component trees (cc-trees) to multivariate images. We propose a general framework for image processing using the cc-tree based on the lattice theory and we discuss the possible applications depending on the properties of the underlying ordered set. This theoretical reflexion is illustrated by two applications in multispectral astronomical imaging: source separation and object detection.

Bulk video imaging based multivariate image analysis, process control chart and acoustic signal assisted nucleation detection

Application of multivariate image analysis in QSPR study of 13C chemical shifts of naphthalene derivatives: A comparative study