Mechanisms for the conversion of hyperspectral images to chemical images

Focal plane array detection systems for measuring optical spectra of spatial images in the visible, near-IR and mid-IR regions have led to a dramatic increase in the amount of spectroscopic information available on a sample. Instead of a single spectrum from an object, we can simultaneously obtain spectra from multiple locations on its surface. Hyperspectral images contain information on the chemical components present in the object under investigation. Generally, the spectrum at each pixel is that of a mixture of a few components. The problem of dealing with so much information can be made more manageable and meaningful by converting the hyperspectral images to chemical images so that the image data can be expressed in terms of the individual chemical components. During the last decade, several algorithms have been proposed for extracting pure component spectra from spectra of mixtures. We have evaluated and combined the features of three different methods to develop an algorithm for rapidly processing hyperspectral images. The hyperspectra were initially processed with Principal Component Analysis to find the appropriate number of independent components and abstract spectral representations (loadings). Key Set Factor Analysis (KSFA) and SIMPLISMA (SIMPle-to-use Interactive Self-modeling Mixture Analysis) methods were combined to find 'pure' wavelengths for the components from the loadings. These 'pure' wavelengths were used to produce initial guesses for the relative concentrations of the components, and these concentrations were used to predict the pure component spectra. The spectra were further refined by using the method of Alternating Least Squares. The methodology is demonstrated on infrared spectra of a simple, three-component chemical mixture and on a hyperspectral infrared image of cartilage tissue.