Baseline corrections

Baseline correction. The flattening or zeroing of the regions of a frequency spectrum that are devoid of resonances. 

Phase cycling. The alternation of the phase of applied RF pulses and/ or receiver detection on successive passes through a pulse seguence without the variation of pulse lengths or delays. Data collected when only RF pulse phases are varied Is often added to data collected with different RF pulse phases to cancel unwanted components of the detected signal, or to cancel artifacts either inherent in single executions of a specific pulse sequence or inherent to unavoidable instrumental limitations. 

Rotamer. Syn. rotational isomer. An isomer generated by rotation (usually 120°) about a chemical bond. 

Rotational isomerism. Interconversion of rotational isomers or rotamers. 

Another problem arising from the correction of baseline distortion stems from the presence of minor components, including the satellite peaks observed in NMR spectra caused by spins. Whereas some baseline correction algorithms work automatically, others require the user to define what is baseline and what is not. When defining baseline, we take care not to include any peaks, however weak, in the baseline region. Failure to exclude small peaks from 

FIGURE 3.7 An example of an infrared spectrum whose baseline is offset by 0.2 absorbance units. This is the spectrum of an aspirin tablet obtained with the KBr pellet sampling method. The baseline offset is caused by the pellet being too thick. 

FIGURE 3.8 An example of a spectrum with a sloped baseline. This is the infrared spectrum of a KBr pellet where the KBr and sample were not ground enough. The resultant scattering of the light beam increases with wavenumber, causing the baseline distortion. 

Note that these three types of baseline problems (offset, slope, and curvature) have known causes, which means they have known solutions. You should do everything in your power to cure the experimental causes of baseline problems before applying any correction algorithms. However, sometimes, despite our best efforts, baseline problems cannot be fixed experimentally. For example, if there are particles embedded 

FIGURE 3.9 An example of a spectrum with a severely curved baseline. 2011 by Taylor Francis Group, LLC 

FIGURE 3.13 Bottom A spectrum in need of baseline correction, and a very poorly drawn parallel function. Top The result of using such a parallel function. 

This function performs a baseline correction on the selected spectra. First load the spectrum that you would like to correct on the first page of the Baseline Correction dialog box shown in Fig. 10.2. The baseline correction can be performed simultaneously for several spectra. 

In order to automatically correct the baseline of the spectrum, switch to the Select Method page (Fig. 10.5) after loading the spectra. Here, you have to specify the correction method and the number of baseline points to be used. As an option you can exclude the spectral region that contain CO2 bands. In this case, data points between 2400 and 2275 cm and between 680 and 660 cm are not taken into account in the calculation. This feature is available for the Automatic option only. 

Select Files Select Method File( ) to Correct----------------- 

The number of baseline points can be varied by clicking on the Number of Baseline Points input field and entering a value between 10 and 200 (unless the function range selected was too small, a minimum of 10 and a maximum of 200 baseline points will be used). The default value is 64. Choose a correction method and start the computation by clicking on Correct. 

For construction of the baseline, the spectrum is divided into n ranges (n being the number of baseline points) of equal size. In the case of absorbance spectra the minimum y-value of each range is determined. Connecting the minima with straight lines creates the baseline. Starting from below , a rubber band is stretched over this curve. The rubber band is the baseline. The baseline points that do not lie on the rubber band are discarded. 

A preliminary first step before applying most methods in this chapter is often baseline correction, especially when using older instruments. The reason for this is that most chemometric approaches look at variation above a baseline, so if baseline correction is not done artefacts can be introduced. 

Dividing data into regions prior to baseline correction 

The simplest form of baseline error is an offset. For example, the sample may attenuate the spectrometer beam by the same amount at all wavelengths and consequently raise the absorbance of the sample with respect to the background. As all baseline corrections should be done with the spectrum linear in absorbance, so that the offset manifests itself as a baseline error above or below zero absorbance. To correct this offset, a constant (typically, the minimum value in the absorbance spectrum) is subtracted from all spectral data points. This results in a removal of the offset so that the baseUne is reset to zero absorbance. Peak absorbance may then be read directly. This procedure is often known as one-point baseline correction. 

A sloping baseUne, one that has a Unear slope from one end of the spectrum of the other, can simply be corrected by subtraction of a ramp function from the absorbance spectrum (two-point baseline correction). Most spectral software incorporates this correction, and the user simply picks a point on the baseUne at one end of the spectrum and a second point at the other end. A linear interpolation is made through the two points, and resulting Une is subtracted from the spectrum. 

Fourier Trartrform Infrared Spectrometry, Second Edition, by Peter R. Griffiths and James A. de Haseth Copyright 2007 John Wiley Sons, Inc. 

Unfortunately, residual baseline anomalies are not often simple offsets or linear slopes, but are more complex functions. 

An extension of two-point baseline correction is to correct a nonlinear baseline with a series of linear interpolations. If the slope of the basehne changes slowly, this method can give acceptable results. Several points can be picked along the baseline, and a linear interpolation is made between each pair of points. Each interpolated section is subtracted from the spectmm. If the baseline does not have a slope that changes slowly, a curved or undulating baseline may result. In addition, basehne discontinuities or abrupt changes in relative slope may result. Nonetheless, this is a useful correction if the original baseline only varies slowly. 

This is a commonly used preprocessing in image analysis that works better if it is performed after image denoising. In this way, the baseline is better defined and can, as a consequence, be better subtracted. Again, a typical basehne correction. 

Figu re 2.2 (a) Raw spectra of a Raman emulsion layer image (b) Spectra after de-noising by principal component analysis (PCA) (c) Spectra after de-noising and baseline correction by asymmetric least squares. 

However, Raman and infrared images are often affected by intense and irregularly shaped baselines. This complex behavior is often attributed to fluorescence contributions in Raman hyperspectral images or to other phenomena, such as the Mie scattering [15], commonly found in biological IR images. For these situations, more sophisticated baseline correction methods that can handle complex baselines changing from pixel to pixel have appeared. 

variations of the classical MSC method, such as extended multiplicative scatter correction (EMSC), have appeared to handle nonlinear baselines that may be better expressed by polynomial functions [16]. Recently, modifications of the EMSC algorithm, especially adapted to handle the Mie scattering effects caused by biological images, have also been implemented [17]. 

3 Detection and Suppression of Anomalous Pixels or Anomalous Spectral Readings 

After processing in the time domain, Fourier transformation, phasing and basic processing (calibration, peak picking, integration) ahs been performed, additional processing steps to improve spectral quality are at your disposal. This includes operations common to both ID and 2D spectra e.g. baseline correction in the frequency domain, as well as operations specific to these different type.s of data sets. 

Non ideal baselines, deviating from horizontal flat lines, are introduced either by 

After selecting the appropriate correction mode, edit fields to set the corresponding correction parameters become active and the functions in the buttons panel are adapted accordingly. Use the Help tool for more information. 

Offset Correction simply translates either the complete spectrum or a defined interval along the y - dimension. 

By defined Points lets you mark certain data points as being on the baseline. From these points a baseline is then calculated according to the specified baseline function, which may (and should) be inspected before the correction is applied. 

One last cautionary note the first order phase can be increased beyond +/- 360° - but shouldn t be If this happens, you will end up with a distorted, wavy baseline. A sine wave is in effect superimposed on the spectrum, so if you see a wavy baseline, check that you haven t wrapped the phase too far. Spectrum 4.4 shows what happens when you go a bit mad with first order phase If you end up in this position, do not attempt any kind of baseline correction as this will add to your problems. Just set both your phase parameters back to zero and start again... 

The spectrum was recorded with the use of an ATR sampling accessory and has been ratioed against a single-beam spectrum of water. 

---

Artifact removal and/or linearization. A common form of artifact removal is baseline correction of a spectrum or chromatogram. Common linearizations are the conversion of spectral transmittance into spectral absorbance and the multiplicative scatter correction for diffuse reflectance spectra. We must be very careful when attempting to remove artifacts. If we do not remove them correctly, we can actually introduce other artifacts that are worse than the ones we are trying to remove. But, for every artifact that we can correctly remove from the data, we make available additional degrees-of-freedom that the model can use to fit the relationship between the concentrations and the absorbances. This translates into greater precision and robustness of the calibration. Thus, if we can do it properly, it is always better to remove an artifact than to rely on the calibration to fit it. Similar reasoning applies to data linearization. 

Discrete area segments were conq)uted over four second time intervals during polymer elution and baseline corrections were made. The raw data were transferred to a Hewlett Packard HP9830 computer for evaluation and plotting. The following conditions were used to analyze the trimer samples and low MW fractions Sample concentration - lyg/yl Injection volume - 50yl Mobile phase - THF Flow rate - 2ml/min... 

Figures 8 and 9 show the first order kinetic plots for the isomerization and crosslinking reactions, respectively. In the data analysis the area of the isoimide peak was measured between consistent limits chosen to exclude any contribution from the 1775 cm imide band. These data were generated by measuring the area of the appropriate peak in a baseline corrected spectrum and ratioing this area to that of a reference peak (which was invarient during the experiment) in the same spectrum. This concentration indicative number was then ratioed to the concentration ratio observed on the initial scan. Plots of the log of the ratio of the concentration of the functionality at time "t" to the concentration of the functionality at t = 0 were then constructed. In order to insure that the trends in the data were not artifacts of this procedure or of the baseline correction routine, we also plotted the data in terms of peak intensity in absorbance units and observed the same trends but with more scatter in the data.

The IR spectra are finally analyzed to determine the effluent concentration from each reactor channel. The quantification of species concentration is performed using either univariate or multivariate calibration methods. For non-overlapping peaks, like CO, C02, and N20, we can use univariate calibration. This is simply performed by baseline correction, the peak areas and/or peak heights and then converting these values... 

Absolute quantification of metabolites complicated (relaxation effects, baseline correction, etc.)... 

Some extensions are not essential for CC, but greatly improve its capabilities. Interfaces to a data storage device and to a hard copy unit are valuable. Some facility for data processing afterwards (baseline correction and peak area determination) is desirable. 

Different baseline correction methods vary with respect to the both the properties of the baseline component d and the means of determining the constant k. One of the simpler options, baseline ojfset correction, nses a flat-line baseline component (d = vector of Is), where k can be simply assigned to a single intensity of the spectrum x at a specific variable, or the mean of several intensities in the spectrum. More elaborate baseline correction schemes allow for more complex baseline components, such as linear, quadratic or user-defined functions. These schemes can also utilize different methods for determining k, such as least-squares regression. 

Step 2 Use raw spectra. (Do not use baseline-corrected, derivatized or pretreated spectra.)... 

Figure 7. Baseline corrected normalized retention volume data over the region to be analyzed.

Selected entries from Methods in Enzymology [vol, page(s)j Boundary analysis [baseline correction, 240, 479, 485-486, 492, 501 second moment, 240, 482-483 time derivative, 240, 479, 485-486, 492, 501 transport method, 240, 483-486] computation of sedimentation coefficient distribution functions, 240, 492-497 diffusion effects, correction [differential distribution functions, 240, 500-501 integral distribution functions, 240, 501] weight average sedimentation coefficient estimation, 240, 497, 499-500. 

Note that raw data (e.g., spreadsheet calculations, uncalibrated measurements, original spectra before they have been baseline corrected) should not be included in a poster. It is important that you present results only after they have been carefully analyzed and prepared for public view. 

---