Automated hemochromatosis spectra analysis using neutron stimulated emission tomography

We identified and diagnosed hemochromatotic cases with an automatic technique based on peak recognition and chemometrics. Hemochromatosis is a disease characterized by an accumulation of iron in body organs. Neutron-stimulated emission computed tomography (NSECT) has demonstrated its ability to detect elevated iron concentrations in the liver through a non-invasive, low dose scan. Fast neutrons are used to generate gamma-ray emission from atomic nuclei in the liver, and the spectral energies of the emitted gamma photons are used to identify the elements of interest. The ability to analyze all gamma lines in the spectra, belonging either to an individual element or to different elements, significantly enhances the overall sensitivity, accuracy, and effectiveness of the diagnosis. We developed a novel peak-finding/peak-fitting algorithm, which rapidly processes all spectra collected on a large scale (i.e. all peaks within each spectrum in a set of spectra), and classifies the samples into healthy and diseased categories. The technique finds, deconvolves, and characterizes peaks based on their position, height, full width at half maximum (FWHM), and area, classifying the samples automatically with two methods, incriminant (novel) and discriminant analysis. We demonstrated that the algorithm classified a population of 64 healthy and 120 diseased simulated patients into healthy and hemochromatotic groups with clinically significant accuracy.