The aim of the proposed research is to develop a highly accurate non- invasive determination of lead in bone by means of a sophisticated multistage signal processing technique, which is driven by case-based reasoning and a hybrid neural network. The validity and processing gain of the method shall be proven on Atlantex ln-Vivo Lead Spectrometer (INVILS) data. The objective is to allow INVILS to quantify lead spectral peaks more accurately with less data. This will reduce x-ray exposure and allow accurate measurement at low lead levels, where prior systems were unable to measure. The signal processing algorithm proposed is composed of four stages: a noise filter, a peak locator, a neural network peak identifier, and a case-based expert system for peak ratios enhancement (see Figure 2). The proposed research will also prototype a user friendly window-driven interface. This will enable future versions of INVILS to be employed at different locations such as hospitals, clinics, factories, etc. The device will supplement the use of blood-lead tests and will largely replace the use of provocative chelation testing. INVILS will be used for screening children at high risk of lead poisoning, Industrially exposed workers, and for certain classes of the general population, such as those suffering from renal disease.
