Resolving Water, Proteins, and Lipids from In Vivo Confocal Raman Spectra of Stratum Corneum through a Chemometric Approach

Direct measurement of water, proteins, and lipids with depth resolution in human subjects is very important for skin-related diseases, for the characterization of skincare product performance. This method, along with the subsequent analysis, leverages chemometrics to extract chemical information. The main advantage of this technique is that it allows for the clinical Raman data set to be collected by trained instrument operators that lack technical expertise to identify, exclude, and remediate all sources of spectroscopic artifacts.

The resulting data set can then be processed to identify outliers that need to be excluded from the date prior to analysis. During the data analysis, a key challenge is the removal of the outliers, and identification of the number of the key components in the data set. The approach show in this video leverages of prior knowledge of the clinical data set and chemometrics approach to successfully extract the water, protein and lipid with depth resolution.

Demonstrating the procedure will be Li Yang, technician from our C&T lab. To begin, have the subject place a marked lesion body site or control site in close contact with the imaging window of the in vivo confocal Raman instrument. Be sure that they cover the whole window to avoid the impact of room light on imaging.

Then, open the software and move the focus until a spectrum similar to the one shown here, is seen. Afterwards, move the focus 10 microns away from skin surface. Collect data for 26 steps with a two-micron step size in the frequency region shown here, using an exposure time of one second.

Measure eight replicates for each area, lasting up to 15 minutes in total. First, use the command window and MATLAB to change the file extension of the collected data from ric to mat. Then load the mat file to the MATLAB software platform, as shown here.

Correct the data set's baseline using the automatic weighted least squares method, by going to the PLS_Workspace window and right-clicking the imported data set, scrolling to Analyze, selecting Other Tools and clicking on Preprocessing. In the window that pops up, click on Show. Then, scroll down the Available Methods toolbar to the Automatic Weighted Leased Squares Baseline filtering, and select Add.

Next, click on Ok to set the options and to apply preprocessing to the data. Save this as Spectra_baseline. Next, return to the command window and replace the data using the baseline corrected result.

Now, go to the Text Editor and run the program as shown here. This will sum up the values between 2910 and 2965 inverse centimeters to obtain the intensity values under each Raman spectrum from the 26 consecutive steps measurement, and store them in an Excel file. In MATLAB, go to the Workspace and set the path for the data in Depth_save, as shown.

Next, use the process described here to interpolate the instrument offset value from 26 to 260 using the linspace function in MATLAB. This process will interpolate the intensity value from 26 to 260 using the spline method, leveraging the newly generated 260 position values. Additionally, it will use the 260 position and intensity values as x and y inputs for the polyfit function respectively, setting the degree value to 20.

Then, it will use the output coefficients and the 260 extended position values as the input for polyval, to obtain the final 260 intensity values. Next, it will calculate the mean intensity and find the point in the curve which is closest to the mean intensity. It will also change the depth value according to the skin surface, in the known two-micron step size.

Now, run the program. Load the Raman spectra data set after removal of the outer skin spectra into the PLS_Toolbox software, under the MATLAB platform and right-click the data set to choose Analyze and then select PCA. Next, click on Choose Preprocessing.

Select Normalize as the preprocessing approach. Then, choose none for the cross-validation. Then, build the model using the three components for the PCA decomposition analysis.

Now, remove the cover on the in vivo Ramen instrument collection window, and collect the room light spectra in the high frequency region using the same parameters used for the reference material's data collection. Identify the room light effect factor through comparison with the room light background. Now, review the scores and remove the spectra with the significantly higher corresponding score value than normal.

This means removing score values of more than 99.8%of the whole data set, which is 0.16 in this study. Save the resulting calibration X-block data. Finally, go to the PLS_Workspace Browser and edit the new file.

Select the Row Labels and go to Hard Delete Excluded to permanently delete the excluded data before re-saving the file. Begin by correcting the Raman spectra baseline using the same just shown. Next, perform the PCA analysis on the preprocessed data set.

Plot the eigenvalues in logarithmic scale along with the number of components by clicking on the Choose Component button, and select log(eigenvalues)as the y value. To perform multivariate curve resolution analysis, first use the data selection button to load the data set into the MCR_main software. Manually choose the number of components and set the component number to between three and eight.

Then, under the Initial Estimation tab, click on the Pure button. Next, select Concentration and click the Do button. Once the screen refreshes, click the Ok button, and then on Continue to move to the next page.

Now, click Continue, and under Implementation, apply fnnls. Then, select six from the drop-down menu for the number of species with non-negativity profiles, and click Continue. On the next page, choose the same parameters and click Continue.

To determine the location at the skin surface, the area under the protein's Raman peak was integrated to obtain the depth profile of the protein signal. The skin surface was defined as the location where the intensity value from the interpolated depth profile was closest to the mean intensity. The exact location of the skin surface does not need to coincide with an experimental data point.

A total of 30, 862 Raman spectra were collected with the data collection protocol described in this video. This large spectral data set contains 20%spectral outliers. Proper identification and removal of the outlier spectra is important to achieve a proper data set.

Here, one can see the contribution from room lights, superimposed on a reference spectra of Roomlight. Principle component analysis was performed on the preprocessed confocal Raman data set, and the eigenvalue, along with the number of factors used, are plotted here. A significant decrease in the eigenvalue was observed for factor nine.

This observation suggests investigating models with the number of principle components varying between three and eight factors for inclusion in the multivariate curve resolution model. When attempting this procedure, it is critical to start a depth profile above the skin surface, to accurately determine the location of the skin surface. This methodology enables the investigation into the impact of skincare products on key skin components, including water, proteins and lipids.