| 1 |
:octicon:`book;1em;sd-text-info` Processing a spectrum |
| 2 |
====================================================== |
| 3 |
|
| 4 |
.. meta:: |
| 5 |
:description: Tutorial on how to process a spectrum with DataLab, the open-source scientific data analysis and visualization platform |
| 6 |
:keywords: spectrum, signal processing, data analysis, scientific data, visualization, open-source, software, DataLab, tutorial |
| 7 |
|
| 8 |
This example shows how to process a spectrum with DataLab: |
| 9 |
|
| 10 |
- Read the spectrum from a file |
| 11 |
- Apply a filter to the spectrum |
| 12 |
- Extract a region of interest |
| 13 |
- Fit a model to the spectrum |
| 14 |
- Save the workspace to a file |
| 15 |
|
| 16 |
DataLab menus change depending on the context. Since we are going to work with a |
| 17 |
spectrum, a 1D signal, we need to select the Signal Panel before proceeding. |
| 18 |
|
| 19 |
In a typical workflow, we would open DataLab and read the spectrum from a file using |
| 20 |
"File > Open...", the |fileopen_sig| button in the toolbar, or by dragging and |
| 21 |
dropping the file into the panel on the right. However, for this tutorial, |
| 22 |
we will use the "Plugins > Test data > Load spectrum of paracetamol" feature to |
| 23 |
generate a test spectrum, which is convenient for demonstration purposes. |
| 24 |
|
| 25 |
.. |fileopen_sig| image:: ../../../datalab/data/icons/io/fileopen_sig.svg |
| 26 |
:width: 24px |
| 27 |
:height: 24px |
| 28 |
:class: dark-light no-scaled-link |
| 29 |
|
| 30 |
.. figure:: ../../images/tutorials/spectrum/01.png |
| 31 |
|
| 32 |
The "File > Open..." menu to open a spectrum file. |
| 33 |
|
| 34 |
.. figure:: ../../images/tutorials/spectrum/02.png |
| 35 |
|
| 36 |
The "Plugins > Test data > Load spectrum of paracetamol" plugin to generate the |
| 37 |
test spectrum for this tutorial. |
| 38 |
|
| 39 |
Once opened, the spectrum is displayed in the main window. It is a 1D signal, so it is |
| 40 |
displayed as a curve. The horizontal axis is the energy axis, and the vertical axis |
| 41 |
is the intensity axis. |
| 42 |
|
| 43 |
.. figure:: ../../images/tutorials/spectrum/03.png |
| 44 |
|
| 45 |
The spectrum displayed on the "Signal View" panel. |
| 46 |
|
| 47 |
The signal is quite clean. However, to demonstrate DataLab's filtering capabilities, |
| 48 |
we will apply a Wiener filter to reduce any residual noise while |
| 49 |
preserving the spectral features. This is available under |
| 50 |
"Processing > Noise Reduction > Wiener filter". |
| 51 |
|
| 52 |
.. figure:: ../../images/tutorials/spectrum/04.png |
| 53 |
|
| 54 |
The "Processing > Noise Reduction > Wiener filter" menu option. |
| 55 |
|
| 56 |
.. figure:: ../../images/tutorials/spectrum/05.png |
| 57 |
|
| 58 |
The result displayed in the main window. |
| 59 |
|
| 60 |
Let's focus our analysis on one of the peaks of interest. To do that, |
| 61 |
we define a region of interest (ROI) around the feature we want to analyze |
| 62 |
and use the "ROI > Extract..." menu to extract it. |
| 63 |
The "Regions of interest" dialog box will be displayed. Select an area and click "OK". |
| 64 |
A confirmation window will appear—click "Yes" to extract the region of interest. |
| 65 |
A new signal containing the ROI will be created and displayed in the main window. |
| 66 |
|
| 67 |
.. figure:: ../../images/tutorials/spectrum/06.png |
| 68 |
|
| 69 |
The "Operations > ROI extraction" menu. |
| 70 |
|
| 71 |
.. figure:: ../../images/tutorials/spectrum/07.png |
| 72 |
|
| 73 |
The "Regions of interest" dialog box displayed. |
| 74 |
|
| 75 |
.. figure:: ../../images/tutorials/spectrum/08.png |
| 76 |
|
| 77 |
The region of interest displayed in the main window. |
| 78 |
|
| 79 |
.. figure:: ../../images/tutorials/spectrum/09.png |
| 80 |
|
| 81 |
Open the model fitting window with "Processing > Fitting > Gaussian fit". |
| 82 |
|
| 83 |
|
| 84 |
.. figure:: ../../images/tutorials/spectrum/11.png |
| 85 |
|
| 86 |
The result of the fit displayed in the main window. |
| 87 |
|
| 88 |
.. figure:: ../../images/tutorials/spectrum/12.png |
| 89 |
|
| 90 |
Both the full spectrum and the fit are selected in the "Signals" panel, so that |
| 91 |
both are displayed in the visualization panel on the left. This allows for easy |
| 92 |
visual comparison if needed for the analysis. |
| 93 |
|
| 94 |
Linear detrending |
| 95 |
----------------- |
| 96 |
|
| 97 |
After fitting the main peak, we may want to remove any baseline drift |
| 98 |
present in the entire spectrum. |
| 99 |
The detrending function in DataLab performs a linear fit across the entire signal, |
| 100 |
including the peaks. In our signal, the peaks occupy a large portion of the data, |
| 101 |
which is acceptable for signals where peaks are symmetrically distributed around the |
| 102 |
center with similar amplitudes. However, this is not the case here, so we cannot |
| 103 |
expect this function to work well. Nevertheless, this example illustrates |
| 104 |
how DataLab functions can be combined to perform more advanced analyses. |
| 105 |
|
| 106 |
To visualize the limitation mentioned above, we will apply the detrending |
| 107 |
function directly to the filtered signal. It's important to remember that we previously set |
| 108 |
a ROI on the signal to focus on the main peak. We need to remove |
| 109 |
this ROI constraint to perform detrending on the full signal. |
| 110 |
To execute the detrending, we use "Processing > Detrending", choose |
| 111 |
the linear detrending method, and click "OK". The result will be |
| 112 |
displayed in the main window. |
| 113 |
|
| 114 |
.. figure:: ../../images/tutorials/spectrum/13.png |
| 115 |
|
| 116 |
The "Processing > Detrending" feature. |
| 117 |
|
| 118 |
.. figure:: ../../images/tutorials/spectrum/14.png |
| 119 |
|
| 120 |
We choose a linear detrending method, and we click on "OK". |
| 121 |
|
| 122 |
.. figure:: ../../images/tutorials/spectrum/15.png |
| 123 |
|
| 124 |
The result of the detrending displayed in the main window. |
| 125 |
|
| 126 |
Comparing the filtered and detrended signals shows, as expected, |
| 127 |
that the detrending function does not work well on this signal. |
| 128 |
As explained earlier, this is because the algorithm performs a linear fit |
| 129 |
across the entire signal, including the peaks. The effect is clearly visible in the plot: |
| 130 |
the higher peaks on the left start at an intensity value lower than those on the right |
| 131 |
after detrending, and all peaks have a baseline below zero. |
| 132 |
|
| 133 |
Improved detrending with peak exclusion |
| 134 |
----------------------------------------- |
| 135 |
|
| 136 |
To overcome the detrending function's limitation, we can draw inspiration from the |
| 137 |
behavior of the detrended signal. We've already identified the problem: the |
| 138 |
linear fit includes both the baseline and the peaks. |
| 139 |
|
| 140 |
For better detrending, we can first exclude the peaks and then perform |
| 141 |
a linear fit only on the non-peak regions. We can reasonably expect this approach to |
| 142 |
provide a more accurate baseline estimation and a better detrended signal. |
| 143 |
|
| 144 |
This is illustrated in the following steps: |
| 145 |
|
| 146 |
.. figure:: ../../images/tutorials/spectrum/15b.png |
| 147 |
|
| 148 |
We select the regions corresponding to the regions without peaks. |
| 149 |
|
| 150 |
.. figure:: ../../images/tutorials/spectrum/15c.png |
| 151 |
|
| 152 |
A linear fit is performed only on the selected regions. |
| 153 |
|
| 154 |
.. figure:: ../../images/tutorials/spectrum/15d.png |
| 155 |
|
| 156 |
The linear baseline obtained from the fit is shown. |
| 157 |
|
| 158 |
.. figure:: ../../images/tutorials/spectrum/15e.png |
| 159 |
|
| 160 |
We delete the rois to apply the detrending on the whole signal. |
| 161 |
|
| 162 |
.. figure:: ../../images/tutorials/spectrum/15f.png |
| 163 |
|
| 164 |
We use the "difference" operation to subtract the baseline from the original signal. |
| 165 |
|
| 166 |
.. figure:: ../../images/tutorials/spectrum/15g.png |
| 167 |
|
| 168 |
We select the linear baseline to perform the subtraction. |
| 169 |
|
| 170 |
.. figure:: ../../images/tutorials/spectrum/15h.png |
| 171 |
|
| 172 |
The resulting detrended signal is shown, now with a correct baseline. |
| 173 |
|
| 174 |
|
| 175 |
Automatic peak detection |
| 176 |
------------------------- |
| 177 |
|
| 178 |
We can use DataLab's "Multi-Gaussian fit" function to automatically identify |
| 179 |
and fit multiple peaks in the spectrum by selecting |
| 180 |
"Processing > Fitting > Multi-Gaussian fit" from the menu. |
| 181 |
|
| 182 |
.. figure:: ../../images/tutorials/spectrum/16.png |
| 183 |
|
| 184 |
First, the "Signal peak detection" dialog box appears. You can adjust |
| 185 |
the vertical cursor position to set the threshold for peak detection |
| 186 |
and specify the minimum distance between peaks. Then click "OK". |
| 187 |
|
| 188 |
.. figure:: ../../images/tutorials/spectrum/17.png |
| 189 |
|
| 190 |
The "Multi-Gaussian fit" dialog box appears. An automatic fit is performed |
| 191 |
by default. Click "OK" (or optionally adjust the parameters manually using |
| 192 |
the sliders, or modify the automatic fitting parameters). |
| 193 |
|
| 194 |
.. figure:: ../../images/tutorials/spectrum/18.png |
| 195 |
|
| 196 |
The result of the fit is displayed in the main window. Here we selected both the |
| 197 |
spectrum and the fit in the "Signals" panel on the right, so both are displayed |
| 198 |
in the visualization panel on the left. |
| 199 |
|
| 200 |
Alternatively, we could use the "Peak detection" feature from the "Analysis" menu to |
| 201 |
detect peaks in the spectrum. This is the first step of the "Multi-Gaussian fit" |
| 202 |
function and can be used independently to detect peaks without performing a fit, |
| 203 |
creating a signal with a delta function at each detected peak position. |
| 204 |
|
| 205 |
.. figure:: ../../images/tutorials/spectrum/19.png |
| 206 |
|
| 207 |
Open the "Peak detection" window with "Analysis > Peak detection". |
| 208 |
|
| 209 |
.. figure:: ../../images/tutorials/spectrum/21.png |
| 210 |
|
| 211 |
After adjusting the peak detection parameters (using the same dialog as |
| 212 |
for the multi-Gaussian fit), click "OK". Then select both the |
| 213 |
"peak_detection" result and the original spectrum in the "Signals" panel |
| 214 |
to display them together in the visualization panel on the left. |
| 215 |
|
| 216 |
Saving the workspace |
| 217 |
-------------------- |
| 218 |
|
| 219 |
Finally, we can save the workspace to a file. The workspace contains all signals |
| 220 |
loaded or created in DataLab, along with processing results and |
| 221 |
visualization settings (such as curve colors). |
| 222 |
|
| 223 |
.. figure:: ../../images/tutorials/spectrum/21.png |
| 224 |
|
| 225 |
Save the workspace to a file with "File > Save to HDF5 file...", |
| 226 |
or the |filesave_h5| button in the toolbar. |
| 227 |
|
| 228 |
.. |filesave_h5| image:: ../../../datalab/data/icons/io/filesave_h5.svg |
| 229 |
:width: 24px |
| 230 |
:height: 24px |
| 231 |
:class: dark-light no-scaled-link |
| 232 |
|
| 233 |
If you want to load the workspace again, you can use the "File > Open HDF5 file..." |
| 234 |
(or the |fileopen_h5| button in the toolbar) to load the whole workspace, or the |
| 235 |
"File > Browse HDF5 file..." (or the |h5browser| button in the toolbar) to load |
| 236 |
only a selection of data sets from the workspace. |
| 237 |
|
| 238 |
.. |fileopen_h5| image:: ../../../datalab/data/icons/io/fileopen_h5.svg |
| 239 |
:width: 24px |
| 240 |
:height: 24px |
| 241 |
:class: dark-light no-scaled-link |
| 242 |
|
| 243 |
.. |h5browser| image:: ../../../datalab/data/icons/h5/h5browser.svg |
| 244 |
:width: 24px |
| 245 |
:height: 24px |
| 246 |
:class: dark-light no-scaled-link |