FILTER Tool

Updated on 2022-08-08

The FILTER tool was developed by CAST member Leon Niezen to promote the development of improved background correction and smoothing algorithms.

Aim #

The tool was developed to rigorously compare a number of recently developed background correction methodologies (i.e. baseline-drift correction as well as noise removal) in a comprehensive and critical manner. The tool uses experimental data on backgrounds and peaks to create large sets of hybrid (part experimental, part simulated) data. It can be used to evaluate a number of different algorithms as described in https://doi.org/10.1016/j.aca.2022.339605, but it can also be used for preprocessing or curvefitting purposes.

Download & Installation #

You can download the latest version here. Note that, currently the app only works on Windows systems.

Using the Tool #
Tab A: Data Pre-Processing
#

Step 1: Load data in the app

The first step in the approach requires loading in the data.

 

Step 2: Specify Input Parameters

Select whether the data should be smoothed, drift corrected, both, or neither. Select the method and input parameters.

 

Step 3: Execute Algorithm

Click Correct Data to apply the selected algorithm to all loaded signals.

The background and/or smoothed signals will be shown on the specified section of the chromatogram.

The signal can be exported for use in other programs.

Tab B: CURVE FITTING / PEAK CHARACTERIZATION #

Step 1: Select Threshold

Select detection threshold based on peak intensity.

Step 2: Select Distribution Function

Select which distribution function or peak model to use.

Step 3: Determine Initial Settings

Automatically determine initial estimates for the fitting based on detection threshold and the selected peak model.

Step 4: Select # of Iterations

Set the number of iterations to use for the fitting.

Step 5: Fit Peaks

Click the button to fit peaks. Note, it is only unlocked after all the parameters are entered correctly.

Step 6: Export

Export fit results to Excel file for use in signal generation.

Optional: Use to determine initial estimates for all loaded measurement using the specified peak model/threshold

Tab C: SIGNAL GENERATION #

Step 1: Load Signals

Load fitted peaks (Excel file) into the app.

Load a blank background measurement.

Step 3: Select Peak Coverage Range

Set a range for peak coverage, the amount of peak overlap, the peak model, the number of outliers and the x-axis range in which to create peaks.

Step 3: Select Signals and Execute

Select the number of signal to generate and execute.

When peak overlap is 0, all peaks will be (nearly) baseline resolved.

Typically, more realistic signals are obtained when overlap is set higher.

Optional: Select a range of different noise intensity to add to the signals.

Optional: Set seed to repeatedly generate the same data.

Optional: Export the generated signals and all its components to an Excel file.

Note: when both a peak coverage range, and noise range are used the quadratic number of signals will be generated, to account for all possible combinations.

Tab D: Algorithm Performance Evaluation #

Step 1: Select Algorithms

Select the number and type of BGC and smoothing algorithms to evaluate for the generated data.

Step 2: Algorithm Comparison

Click Generate Comparison to perform algorithm comparison on generated data.

Top: Each color represents a particular algorithm. As indicated in the left figure.

Optional: Review the corrections of the signals using these dropdown menus.

Optional: Export the corrected data, the (optimal) parameters for the algorithms per signal and the RMSE values, to an excel file.

Further Reading #

Readers interested in learning more about the tool, its scientific validation and the related fundamental concepts of background correction are referred to our publication in Analytical Chimica Acta along which this tool was published. The concept of background correction and its significant role in signal processing is also further discussed in our 2021 review in Journal of Separation Sciences. Both publications have been released open-access and thus can be downloaded and shared freely.

Critical Comparison of Background Correction Algorithms
Recent Applications of Chemometrics in 1D and 2D Chromatography

In addition, another useful resource is the book Data Analysis and Signal Processing in Chromatography by Atilla Felinger.

References #

[1] L.E. Niezen, P.J. Schoenmakers & B.W.J. Pirok, Critical comparison of background correction algorithms used in chromatography, Anal. Chim. Acta, 2022, 1201, 339605, DOI: 10.1016/j.aca.2022.339605.

[2] T.S. Bos, W.C. Knol, S.R.A. Molenaar, L.E. Niezen, P.J. Schoenmakers, G.W. Somsen & B.W.J. Pirok, Recent applications of chemo metrics in one- and two-dimensional chromatography, J. Sep. Sci., 2022, 43(9-10), 1678-1727, DOI: 10.1002/jssc.202000011.

[3] A. Fellinger, Data Analysis and Signal Processing in Chromatography, 1998, Elsevier, ISBN: 9780444820662.