Paper Title
Nonlinear Least Squares Regression Analysis And Diagnosis For Optical And Laser Measurement Systems

Abstract
Laser Doppler and Raman spectroscopy are widely adopted for biomedical diagnosis. Both Doppler ultrasound and Raman spectra are used to display normal and abnormal signature waveforms that are unique in collected samples, in order to recognize normal and abnormal regions in the spectral display respectively. Visualization of quantitative Doppler ultrasound data and Raman spectra provides a straightforward means to distinguish between normal and abnormal samples, however, mixing artifacts due to motion, instrumentation and background are in fact inevitable. The fluorescence artifact is the dominating factor whose formation mechanism varies. In general, fluorescence waveforms could be formulated as typical Gaussian, Sigmoidal or Lorentzian distribution, while single spectrum is seldom enough to represent the fluorescence waveforms completely to further extract intrinsic spectra with unique signatures. In this case, combination of Gaussian, Sigmoidal and Lorentzian waveforms could be the solution to model the fluorescence artifact. To identify a set of modeling parameters, a simple but powerful nonlinear least squares regression technique is introduced in this article. Case studies on both normal and abnormal sample data are conducted. Index Terms—Nonlinear Least Squares Regression, Raman Spectroscopy, Doppler Ultrasound