Chemometric Methods for Biomedical Raman Spectroscopy and Imaging - Emerging Raman Applications and Techniques in Biomedical and Pharmaceutical Fields - page 189

Fig. 8.4. A priori probability density functions (pdfs) of two classes and their result-
ing assignments after weighting them with their corresponding class probabilities.
For a given feature value along the x -axis, the higher of the corresponding y -axis
values decides the class for that value. Two types of errors are possible with this
scheme, namely the misclassification of class 1 as class 2 ( horizontal stripes )andthe
misclassification of class 2 as class 1 ( vertical stripes/shaded ). The colored regions
indicate the relative probabilities of such errors. Errors can be explicitly understood
and the contribution of each feature to classification can be quantitatively measured
8.3.3 Artificial Neural Networks (ANN)
Neural network methods [90] are among the most well-known nonlinear clas-
sification techniques. They involve a “layer” of input nodes and a layer of
output nodes along with one or more “hidden” layers of elements called neu-
rons (Fig. 8.5). Each neuron sums input from points in different layers and
changes its output to these inputs in a nonlinear fashion. The major advan-
tage of artificial neural network (ANN) methods is that they are general:
they can handle problems with large numbers of parameters and classify ob-
jects well, even when distributions in the N -dimensional parameter space
are very complex. ANNs are good at capturing information and recogniz-
ing patterns, especially in modeling human pattern observation and recogni-
tion capabilities (i.e., repetitive tasks). The nonlinear nature of the classifier
makes it exceptionally powerful and there are few problems that the classi-
fier will not address to high accuracy. ANNs have been extensively employed
in Raman spectroscopy and imaging. Representative examples include bac-
terial identification [91, 92], histopathology of skin [93, 94] and vessels [95],
aqueous humor glucose [96], plant materials [97], and for fundamental peak
assignments [98].
A major criticism of ANNs is that features that were critical to decision
making are not explicitly available. Consequently, it is dicult to determine