| Summary: | 碩士 === 國立中正大學 === 化學所 === 95 === Abstract
Second derivative filter is one kind of band pass filter. When the signal profile is differentiated twice consecutively, the constant and linear part of the background is eliminated; the result is a rather sharp peak rising above a background around zero. We have used second derivative filter- based signal processing techniques to solve the problem of chromatogram baseline drifting. When the chromatogram is first through second derivative filter, we add one artificial chromatogram containing only thermal noise. Then we multiply these two chromatograms in order to adjust noise spectrum range in frequency domain. Finally, the multiplied chromatogram through the other second derivative filter to enhance the signal to noise (S/N) ratio of LC-MS peaks.
We determine quality control samples to validate the above techniques. Comparison of the difference between the estimated concentrations using original LC-MS signal intensity and using post- signal processing signal intensity. In low concentrations, we find that standard deviation of post-signal processing peak is smaller than that of original peak. Finally, we have used standard addition method to confirm that analyte concentration determination is more accurate when LC-MS peaks are processed. Therefore we can utilize second derivative filtering to quantify real drug samples.
Besides, we use matched filter and median filter and Savitzky-Golay smoothing in the first filtering step respectively, and combine multiplication computation to modify the noise frequency spectrum. After that we use the other matched filter as the second filter to process LC-MS-MS chromatograms. Comparisons of the S/N ratio improvement and reproducibility are studied. We have found that the S/N ratio improvement using the procedures of matched filter- multiplication computation-matched filter is up to 12 fold. We obtained optimum quantification results with standard addition methods, when the procedures of matched filter- multiplication computation-matched filter is used.
|
|---|
