Method for Filtering a Chromatogram

Abstract

Low-complexity, application-independent fixation of a chromatogram is achieved by a) determining a limit frequency under the assumption that the shape of the peaks in the chromatogram corresponds approximately to a Gaussian function having a standard deviation and the Fourier transform of the Gaussian function describes the frequency spectrum of a peak, at which limit frequency the Fourier transform has decreased to a predetermined limit value, b) determining the height, width, and retention time of each individual peak from the chromatogram, or a chromatogram taken previously under the same conditions, c) determining a constant factor based on a first predetermined relationship, d) determining the functional dependency of the limit frequency on the retention time as a variable quantity based on a second predetermined relationship, and e) filtering the chromatogram with the limit frequency depending on the retention time as the variable quantity using a low-pass filter.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The invention relates to digital signal processing and, more particularly, to a method for filtering a chromatogram.

2. Description of the Related Art

In chromatography, a sample of a substance mixture to be analyzed is passed through a chromatographic separating device. Because of different migration rates through the separating device, the analytes, i.e., the individual substances of the substance mixture, reach the output of the separating device at different times and are successively detected at this point by a suitable detector. The time that the analytes require to migrate through the separating device is referred to as the retention time. As its measurement signal, the detector generates a chromatogram that consists of a baseline and a number of peaks corresponding to the separated substances. In practice, the chromatogram is affected by noise, with the individual peaks standing out more or less clearly from the signal noise. The detection limit of an analyte is defined as a predetermined multiple of the noise. That is, the peak height measured from the noise-free baseline, i.e., from the average value of the noise, must be at least the predetermined multiple of the noise.

With well-resolved peaks, the peak area above the noise-free baseline is proportional to the concentration of the analyte. The peak area, in contrast to the peak height, provides accurate measurement results even for nonsymmetrical peaks.

In order to isolate the analytical information, i.e., the peaks, the chromatogram is smoothed by lowpass filtering. Smoothing algorithms suitable for this purpose are, for example, a moving average or the Savitzky-Golay filter. The lower the limit frequency of the lowpass filter or the greater the filter length of the Finite Impulse Response (FIR) filter used, the better the smoothing that can be obtained. With increasing smoothing, however, the peaks may also be deformed so that the measurement accuracy is reduced. Depending on which substance mixtures are to be analyzed, the measurement applications, for example, different separating columns with different interconnection, and measurement conditions within an application, for example, different temperature and pressure profiles in the separating device, may be very different and lead to correspondingly different chromatograms, which necessitates differently dimensioned filters for smoothing the chromatograms.

SUMMARY OF THE INVENTION

It is therefore an object of the invention to permit low-complexity, application-independent and universal filtering of chromatograms.

This and other objects and advantages are achieved in accordance with the invention by a method comprising the following steps:

• • a) under an assumption that the shape of the peaks (P) in a chromatogram respectively corresponds approximately to a Gaussian function ƒ (t, σ) with standard deviation σ and the Fourier transform F (f, σ) of the Gaussian function ƒ (t, σ) describes the frequency spectrum of a peak (P), a limit frequency f_G(F_G, σ) at which the Fourier transform F (f, σ) has decreased to a predetermined limit value F_G=F (f_G, σ₀) is determined,
  • b) the height h₀, width b₀ and retention time t_R0 of an individual peak (P₀) of the chromatogram, or of a chromatogram previously recorded under the same conditions, are determined,
  • c) the constant factor K is determined with the aid of the relation σ₀/h₀=K·t_R0,
  • d) the functional dependency of the limit frequency f_G(F_G, t_R) on the retention time t_R as a variable quantity is determined with the aid of the relation σ/h=K·t_R and
  • e) the chromatogram is lowpass-filtered with the limit frequency f_G (F_G, t_R) as a function of the retention time t_R.

The invention is based on the observation that the ratio of width b to height h of a peak increases linearly with the retention time t_R, so that b/h=K′·t_R. Apart from a few exceptions, the shape of the peak becomes ever closer to a classical Gaussian distribution as the retention time t_R increases. Consequently, a peak of width b can be described with sufficient accuracy by a Gaussian function ƒ (t, σ). Depending on where the peak is measured, the peak width b is a multiple of the standard deviation σ, and is, for example, at half peak height b=2σ{square root over (2 ln 2)}=2.355σ. The aforementioned ratio of width b to height h of a peak can therefore also be described by σ/h=K·t_R.

The height h and the standard deviation σ of the Gaussian function are associated with one another by the relation h=1/(σ{square root over (2π)}). The factor K can therefore be determined with the aid of the measured height h₀, width b₀ and retention time t_R0 of a selected individual peak, where the standard deviation σ₀ is calculated from the peak width b₀.

It is now possible to express the Gaussian function ƒ (t, σ) in a functional dependency on the retention time t_R as a variable quantity: ƒ (t, t_R). The Fourier transform F (f, t_R) of the Gaussian function ƒ (t, t_R) then describes the frequency spectrum of a peak as a function of the retention time t_R. A limit value F_G is now established for this frequency spectrum F (f, t_R), with frequencies f having transforms above this limit value F_G being regarded as analytical information and frequencies f having transforms below this limit value F_G being regarded as noise. Both the limit value F_G and the limit frequency f_G associated with it are dependent on the retention time, i.e., F_G=F_G(t_R) and f_G=f_G (t_R). The chromatogram is now smoothed by lowpass filtering with the limit frequency f_G (t_R) thus determined as a function of the retention time t_R.

The individual peak that is used for determining the factor K may be selected from the chromatogram currently to be evaluated or from a chromatogram recorded earlier for the same measurement applications and under the same measurement conditions. The latter includes, for example, the option of obtaining the values of the peak from any available sources in standard measurement applications.

In accordance with the above-described method step d), the functional dependency f_G (F_G, t_R) of the limit frequency f_G on the retention time t_R as a variable quantity is determined with the aid of the relationship σ/h=K·t_R. The latter includes the option of performing the conversion of the functional dependency on the peak width b (or standard deviation σ) into the functional dependency on the retention time t_R even earlier, such as by using the Gaussian function (ƒ (t, σ)→ƒ (t, t_R)) or its Fourier transform (F (f, σ)→F (f, t_R)).

The lowpass filtering, for example, a moving average, is preferably performed by an FIR filter whose filter length is varied according to the limit frequency f_G (F_G, t_R) as a function of the retention time t_R.

Other objects and features of the present invention will become apparent from the following detailed description considered in conjunction with the accompanying drawings. It is to be understood, however, that the drawings are designed solely for purposes of illustration and not as a definition of the limits of the invention, for which reference should be made to the appended claims. It should be further understood that the drawings are not necessarily drawn to scale and that, unless otherwise indicated, they are merely intended to conceptually illustrate the structures and procedures described herein.

BRIEF DESCRIPTION OF THE DRAWINGS

The further explanations of the invention will refer to the figures of the drawing; in detail:

FIG. 1 is an exemplary graphical plot of a chromatogram comprising a multiplicity of peaks;

FIG. 2 is an exemplary graphical plot of the amplitude characteristics (frequency spectra) of three peaks with different retention times;

FIG. 3 is a graphical plot of a detail of a further chromatogram; and

FIG. 4 is a flow chart of the method in accordance with the invention.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

FIG. 1 is a graphical plot of a conventional chromatogram, which consists of a baseline BL and a multiplicity of peaks P with different peak heights h and retention times t_R. Each peak P results from the detection of a specific analyte, the area of the peak P above the baseline BL being proportional to the concentration of the analyte. With reference to FIG. 1, as can be seen, the relative width, i.e., the width of the peak P in relation to the peak height, increases with an increasing retention time t_R, the shape of the peak P becoming ever closer to a classical Gaussian distribution. With respect to their shape, the peaks P can therefore be described as follows by a Gaussian function:

$\begin{array}{cc}f\left(t,\sigma \right)=\frac{1}{\sqrt{\sigma \sqrt{2\Pi }}}{}^{-\frac{1}{2}{\left(\frac{t-t_B}{\sigma }\right)}^2},& \mathrm{Eq}.1\end{array}$

where σ denotes the standard deviation.

The frequency spectrum of an individual peak P is given by the Fourier transform F (f, σ) of the Gaussian function ƒ (t, σ) as:

$\begin{array}{cc}F\left(f,\sigma \right)={\int }_{-\infty }^{+\infty }f\left(t,\sigma \right){}^{-\mathrm{j2}\pi \mathrm{ft}}t={}^{-\frac{1}{2}{\left(2\pi f\sigma \right)}^2}.& \mathrm{Eq}.2\end{array}$

The frequency spectrum simultaneously represents the amplitude characteristic, which in signal processing technology is conventionally indicated in decibels:

$\begin{array}{cc}\begin{array}{c}\frac{F\left(f,\sigma \right)}{\mathrm{dB}}=20{\log }_{10}\left(F\left(f,\sigma \right)\right)\\ =20{\log }_{10}\left({}^{-\frac{1}{2}{\left(2\pi \mathrm{fa}\right)}^2}\right)\\ =20\left(-\frac{1}{2}{\left(2\pi f\sigma \right)}^2\right){\log }_{10}e\\ =-4.343·{\left(2\pi f\sigma \right)}^2\end{array}& \mathrm{Eq}.3\end{array}$

As previously explained, it has been observed that the ratio of width b to height h of a peak P increases linearly with the retention time t_R, so that:

b/h=K′·t_R·or σ/h=K·t_R.  Eq. 4

With Eq. 4, Eq. 3 can be rewritten as follows:

$\begin{array}{cc}\frac{F\left(f,t_R\right)}{\mathrm{dB}}=-4.343·{\left(2\pi f·h·K·t_R\right)}^2.& \mathrm{Eq}.5\end{array}$

FIG. 2 shows that, according to Eq. 5, the amplitude characteristic F (f, t_R)/dB of a peak decreases with an increasing retention time t_R, so that with an increasing retention time t_R a smaller bandwidth is required to perform signal processing in a microprocessor or other similar computing device. In order to establish the bandwidth for the signal processing, a limit frequency f_G at which the amplitude characteristic F (f, t_R)/dB has decreased to a predetermined limit value F_G is selected. If, for example, F_G=−40 dB is selected, then about 99% of the signal energy still lies above this limit value in the amplitude characteristic and the signal distortion to be expected is minimal. With F_G=−40 dB, Eq. 5 establishes the following relationship for the limit frequency:

$\begin{array}{cc}{f}_{G\left(-40\mathrm{dB}\right)}\left(t_R\right)=\frac{\sqrt{40/4.343}}{2\pi ·h·K·t_B}.& \mathrm{Eq}.6\end{array}$

Knowing the height h and the factor K of a single peak, a universal filter can be developed for all peaks P of the chromatogram, the limit frequency f_G of which is varied as a function of the retention time t_R.

As will be explained below, the factor K is determined with the aid of a suitable individual peak by using the relationship of Eq. 2.

FIG. 3 shows an enlarged detail of another chromatogram, which was recorded into a device, such as a memory, under the same conditions as that according to FIG. 1. The peak height h₀, peak width b₀ at half peak height and the retention time t_R0 of a representative individual peak P₀ are measured, the following exemplary values being obtained:

h ₀=0.021

b _o=0.8 s

t _R0=37.64 s.

With b=2σ{square root over (2 ln 2)}, this gives the following for the standard deviation σ_o of the associated Gaussian function:

σ₀=0.34 s.

By using Eq. 4, the following is obtained for the factor K:

$K=\frac{{\sigma }_0}{h_0·t_{R0}}=\frac{0.34s}{0.021·37.64s}=0.043.$

The following is therefore obtained according to Eq. 6 for the limit frequency f_{G (−40 dB)}, or the −40 dB bandwidth of the selected peak P₀:

$\begin{array}{c}{\stackrel{\_}{z}}_{G\left(-40\mathrm{dB}\right)}\left(t_{F0}\right)=\frac{\sqrt{40/4.343}}{2\pi ·h_0·K·t_{R0}}\\ =\frac{\sqrt{40/4.343}}{2\pi ·0.021·0.43·37.64\varepsilon }\\ =1.42\mathrm{Hz}.\end{array}$

All signal components with frequencies greater than 1.42 Hz can therefore be removed by suitable signal processing from the peak P₀ considered here by filtering with virtually no loss of information, the signal/noise ratio or the detection limit being increased.

The entire chromatogram of FIG. 1 can now be lowpass-filtered with the limit frequency varied in accordance with Eq. 6 as a function of the retention time t_R.

$\begin{array}{cc}{f}_{G\left(-40\mathrm{dB}\right)}\left(t_R\right)=\frac{\sqrt{40/4.343}}{2\pi ·h_0·K·t_R}=53.489·\frac{1}{t_F}.& \mathrm{Eq}.7\end{array}$

To this end, for example, an FIR filter may be used. With a moving average, the following applies for the −3 dB limit frequency f_c of the FIR filter:

$\begin{array}{cc}{f}_0\approx \frac{f_A}{2N_F},& \mathrm{Eq}.8\end{array}$

where f_A denotes the sampling frequency of the analog/digital conversion and N_F denotes the number of sampled values of the chromatogram that are used for the averaging. By using Eq. 7 and Eq. 8, with f_c=f_{G (−40 dB)} (t_R), the following filter length to be varied with the retention time t_R is obtained:

$N_r=\frac{f_A}{2f_{G\left(-40\mathrm{dB}\right)}\left(t_{\lambda }\right)}=0.009348·f_A·t_B,$

where N_F is rounded to an integer.

The signal/noise ratio, or the detection limit, is thereby increased by a factor of:

$\frac{S}{N}=\sqrt{N_F}.$

Thus, while there have shown and described and pointed out fundamental novel features of the invention as applied to a preferred embodiment thereof, it will be understood that various omissions and substitutions and changes in the form and details of the devices illustrated, and in their operation, may be made by those skilled in the art without departing from the spirit of the invention. For example, it is expressly intended that all combinations of those elements and/or method steps which perform substantially the same function in substantially the same way to achieve the same results are within the scope of the invention. Moreover, it should be recognized that structures and/or elements and/or method steps shown and/or described in connection with any disclosed form or embodiment of the invention may be incorporated in any other disclosed or described or suggested form or embodiment as a general matter of design choice. It is the intention, therefore, to be limited only as indicated by the scope of the claims appended hereto.

Claims

1.-2. (canceled)

3. A method for filtering a chromatogram, comprising: a) determining, at a processor, a limit frequency at which a Fourier transform has decreased to a predetermined limit value, based on an assumption that shapes of peaks in the chromatogram respectively correspond approximately to a Gaussian function having a standard deviation and the Fourier transform of the Gaussian function describing a frequency spectrum of a peak;b) determining, by the processor, a height, width and retention time of one of an individual peak of the chromatogram and a chromatogram previously recorded under identical conditions;c) determining, at the processor, a constant factor based on a first predetermined relationship;d) determining, at the processor, a functional dependency of the limit frequency on a retention time as a variable quantity in accordance with a second predetermined relationship; ande) low-pass filtering the chromatogram with the limit frequency as a function of the retention time as the variable quantity.

4. The method as claimed in claim 1, wherein the low-pass filtering is performed by a finite impulse response filter having a filter length varied in accordance with the limit frequency as a function of the second retention time.

6. The method of claim 3, wherein the first predetermined relationship is σ/0/h0=K·tR0, with σ0 being the standard deviation of the Gaussian function, h0 being the height of the individual peak, K being constant factor and tR0 being the retention time of the individual peak.

7. The method of claim 3, wherein the second predetermined relationship is σ/h=K·tR, with σ being the standard deviation of the Gaussian function of a respective one of the peaks, h being the height of the respective one of the peaks, K being the constant factor and tR being the retention time of the respective one of the peaks.

8. The method of claim 6, wherein the second predetermined relationship is σ/h=K·tR, with σ being the standard deviation of the Gaussian function of a respective one of the peaks, h being the height of the respective one of the peaks, K being the constant factor and tR being the retention time of the respective one of the peaks.