Skip to Main Content
An artificial-intelligence approach is proposed to differentiate various biomedical samples via Raman spectroscopy technology to obtain accurate medical diagnosis and decision making. The complete process consists of noise filtering, fluorescence identification, optimization and elimination, spectral normalization, multivariate statistical analysis, and data clustering, as well as the final decision making. Numerous modeling, intelligent control, and system-identification schemes have been employed. By means of fuzzy control, genetic algorithms, and principal component analysis (PCA), as well as system identification, a systematic intelligent-control approach is formulated, which is capable of classifying diversified biomedical samples. Raman spectra are weak signals whose features are sensitive to a variety of noises, which have to be reduced to an acceptable level. Fuzzy logic has been known to interpret uncertainty, imprecision, and vague phenomena. Thus, a fuzzy controller is used for noise filtering. On the other hand, background fluorescence acts as a secondary intensity component within a raw Raman spectrograph, so its spectral baseline should be determined. By removing background fluorescence, intrinsic Raman spectrum can be extracted in consequence. To optimize this detrend process, genetic algorithms have been implemented for baseline-function global optimization by selecting an optimal combination of individual spectroscopic functions. Normalization is performed by standard normal variate (SNV) afterwards to compensate for scattering effects. Normalized intrinsic spectra can be used for sample differentiation, where the PCA approach distinguishes some signatures from different samples in terms of dominant principal components. Eventually, various principal components are accumulated for clustering using scatter plots. The long-term objective of this intelligent-control approach is to create a real-time technique for sample analysis, using a Raman spectrometer directly mounted at the end-effectors of medical robots, which is to enhance the robotic surgery.
Automation Science and Engineering, IEEE Transactions on (Volume:2 , Issue: 1 )
Date of Publication: Jan. 2005