Online Software Help Manual

BookmarkIndexPrint
Contents
Display Legacy Contents

Baseline correction

A perfect curve shape of analytical 2D data objects include a constant base level value, where no signals are observed. This base level is called the baseline of a 2D data object. Because of changes in experimental conditions during measurement, temperature influences or any other interference, the baseline sometimes drifts away from its original base level. In this case, the baseline of a 2D data object might be corrected after a measurement has been completed using the baseline correction function of the software. It might be applied, whenever consecutive operations are required like ATR correction or finding peaks.

How does baseline correction work?

Several baseline algorithms are available as described in the following:

  • Polyline algorithm

  • Horizontal algorithm

  • Peak detection algorithm

  • Linear least squares regression algorithm

  • Two point algorithm

  • Spline algorithm

  • AirPLS algorithm

To select one of the algorithms, choose the entry from the algorithm drop down box in the Mathematics tab. Each algorithm provides a set of individual parameters, which can be adjusted after selection.

Polyline algorithm

This algorithm shows a linear baseline drawn directly between first and last data point of the current visible part of the 2D data object in the data view. The end points of the line possess baseline knots (red squares) which can be moved to adjust bias and slope of the line:

Additional knots can be added optionally to get a polygonal line for correction. Please review the Baseline Correction section in the chapter "Commands" to learn how to add and remove baseline knots. After adding some baseline knots, correction can be carried out.

The polyline algorithm offers an additional autodetect option. Choosing this option will modify the polyline so that the baseline knots of the resulting corrected spectra will lie on the x-axis:

  • Autodetect selected: All spectra will be corrected so that the selected baseline knots will lie on the x-axis with an y-value of 0.

  • Autodetect unselected: The selected polyline will used "as is" to correct the spectra. The resulting baseline may not coincide exactly with the x-axis.

The corrected data object looks like this:

After correction a new baseline will be proposed automatically for subsequent correction.

Which region is considered for baseline correction?

Baseline correction is only performed on the area covered by the line! Data points outside the area remain unchanged.

Horizontal algorithm

The baseline will be corrected by using a horizontal line. In principle this algorithm works like an offset correction. The entered value will be subtracted from the entire spectrum, thus shifting it up or down for a certain amount. Therefore this algorithm is especially useful for correcting spectra with a constant offset. The correction value can either be entered by moving the horizontal line directly in the spectrum or by entering a numerical value into the parameter "absolute height". By clicking the Reset button the horizontal line and the parameter "absolute height" will automatically be set to the minimum intensity present in the spectrum. You can use this algorithm to correct positive or negative offsets.

Peak detection algorithm

A polygonal fit function will be determined automatically, which follows the slope of the graph of the 2D data object by neglecting detected peaks. Significant changes in the bias indicate the starting and ending point of a peak signal. These regions will be excluded from baseline detection automatically. The user might assist the automatic peak detection algorithm by adjusting the following parameters:

Group adjacent peaks

This parameter indicates, whether adjacent or overlapping peaks will be interpreted as a single peak or multiple peaks. If the flag is set true, such overlapping peaks will be ignored and identified as one peak. The baseline of the following spectrum excerpt is shown in the figure below (red line):

If the parameter is set false, at least one data point between the end of the first peak and the starting point of the next peak will be interpreted as a base line point. In this case, the same baseline correction detects multiple peaks within the displayed area from above. The baseline follows the red line in this case:

The resulting spectrum looks like this:

Minimum Peak Height

This parameter controls the peak detection concerning the minimum peak height. It is a threshold value relative from the imaginary base line along the graph slope, that must be overridden to identify a peak. The minimum relative peak height is given in fractions of y-axis units.

Minimum Peak Width

This parameter controls the peak detection concerning the minimum peak width. A minimum expected peak width is adjusted here. This value might alter depending on the data type. The minimum estimated peak width is given in fractions of x-axis units.

Linear least squares regression algorithm

This baseline correction algorithm facilitates the Standard Normal Variate correction. Besides the noise and background correction the overall graph slope can be detrended. A polynomial fit function is applied for detrending.

Detrend

Detrending can be included into the baseline correction or not. The following parameter settings are available:

  • None
    No detrending is performed.

  • Polynomial Fit
    A polynomial fit function is applied for detrending.

Two point algorithm

The baseline will be corrected by a line defined by two points. These endpoints can be automatically calculated from a predefined region in the spectrum. This is useful for sets of spectra which exhibit slight shifts. Several methods calculating the start and end point are available. The selection for the two point baseline correction looks like this:

The endpoint preselection is done by moving the red selection rectangles to the desired position. The width of the selection rectangle can be adjusted by the grey tracker boxes and defines the region of points from which the actual endpoint will be calculated by one of the following methods. The actual calculated endpoints are shown as red vertical lines with red boxes inside the selection rectangles.

Single

Two distinct points are directly selected as endpoints. The point are show as red vertical lines without selection areas in the spectrum. This is similar to the simple line algorithm.

Average

The endpoints are calculated as the average of the group of points that are defined by the selection rectangle.

Minimum

The endpoints are calculated as the minimum of the group of points that are defined by the selection rectangle.

Maximum

The endpoints are calculated as the maximum of the group of points that are defined by the selection rectangle.

None

No endpoint selection algorithm is used.

Alternatively the numerical values can be entered directly using the parameter sets Startx, EndX or Start Minimum, Start Maximum, End Minimum and End Maximum:

The two-point algorithm is also part of the baseline correction used in the command thickness correction. Please refer to the chapter thickness correction for a detailed description of the selection methods.

Spline Algorithm

The spline algorithm works similar to the Polyline algorithm except for the handling of the baseline knots. When using the spline algorithm, each baseline knot will have additional spline controls to help adjusting the line shape. By moving the spline controls the user is able to accurately adapt the baseline shape to the spectrum. The following picture shows an example of an intermediate baseline knot with two spline controls:


Adaptive iteratively reweighted penalized least squares (AirPLS)

The AirPLS algorithm is a baseline correction algorithm which works completely on its own and that does not require any user intervention or prior information, such as peak detection etc. Parameters provided by user are only the maximum amount of iteration and a Lambda.

  • Iterations
    The maximum amount of iterations the algorithm should do.
  • Lambda
    A detail parameter for the algorithm.

Which Lambda fits my needs?

The lower Lambda is set, the more detailed the algorithm will work but will also need more time. Little shifts in baseline will be more likely corrected with a low Lambda. In most cases the default Parameter (Lambda = 10) is sufficient for baseline correction.



Iterations: 15
Lambda: 10

For further information about the AirPLS algorithm please read the paper of Zhi-Min Zhang, Shan Chen and Yi-Zeng Liang.
Baseline correction using adaptive iteratively reweighted penalized least squares