| Summary: | 碩士 === 國立中正大學 === 化學所 === 94 === Liquid chromatography – mass spectrometer (LC/MS) is a conventional chemical analysis instrument to determine ultra-trace level (50 ng/mL or lower) samples. In this thesis, we modify popular signal processing methods such as matched filtering and second derivative filtering to minimize the noise intensities riding on weak LC/MS peaks. The signal-to-noise (S/N) ratios of LC/MS signals have been enhanced significantly with our modifications. Matched filtering that is a popular technique to improve chromatographic signals. Matched filtering, which consists of a cross correlation between one chromatographic analyte peak with a reference signal of similar peak shape. In our previous work, we found, because matched filter works as a low-pass filter in the frequency domain, it cannot effectively eliminate the low frequency components in flicker noises contributed by spike-like noises appeared on chromatograms. When the chromatograms are first through matched filter, we implement an multiplication computation against another artificial chromatogram added with thermal noises. As a result, low frequency noises remained in the filtered chromatograms are shifted toward higher frequency regime. When the noise-modified chromatogram is processed with the second matched filter, more noises are eliminated to accomplish higher S/N ratio enhancement. For example, the improvement ratio of despiramine analyte is about 2.3 with the first matched filtering procedures. Even two consecutive matched filters cannot provide any further improvement. In contrast, with the noise modification steps using multiplication computation, the improvement ratio is improved to 14.2 when the second matched filter is performed.
Second derivative filtering, of which the major application is to remedy the drifting problem of chromatogram base-line, is made into the reference peak with two differentials for the originally level and smooth reference peak and then cross-correlation with analyte chromatographic peaks. Using simulation studies we have found that needs to consider two parameters, the width and shaped of reference peak when second derivative filter is implemented. Especially, the S/N ratio improvement is dependent on the reference peak shape when the width of reference peak is less than that of analyte peak. Regardless the reference peak shape, the best improvement is found when the width of reference peak is two to two and half times of that of analyte peak. Finally, we also use the same multiplication computation as the above to modified the noises on the second derivative filtered chromatogram. Due to the noise frequency range shifting, when the second derivative fitler is performed one more time, the S/N ratio improvement is significantly enhanced.
|
|---|
