The subject of the invention is the method for determining the malignancy grades of human brain tumors in optical detection by Raman spectroscopy, for use in oncological clinical diagnostics and in the determination of the margin between the normal and the tumorous tissue.

Currently, the grades of malignancy of human brain tumors are determined based on routine histopathological examination. The histological assessment of the malignancy of the tumors is based on the histopathological features and the clinical properties of the tumors including the ability to infiltrate, to destroy surrounding tissues, and to spread to distant locations by metastasis. The determination of the grades of malignancy is based on microscopic examination of tissue stained with hematoxylin and eosin, sometimes ex tended to other methods, for example immunohistochemistry. The architecture of tumor tissue and the cytological features of tumor cells are studied. Cytological features take into account: the degree of anaplasia, the size and shape of the cell nuclei, the volume of the cytoplasm, the relative percentage of dividing cells (mitotic index), and from the mor phological features the histological organization of the tissue and a safety margin for the removal of the tumor are assessed.

Usually, the four-stage malignancy scale is used: from the grade I (Gl) for tumors with the lowest grade to grade IV (G4) for tumors with the highest grade of malignancy); GO describes the non tumorous tissue (normal tissue).

The patent applications P. 424257 and P. 427376 demonstrate the usefulness of Raman biomarkers: l293o/2844and Iis84/i440 in differentiating normal tissue from the tumor ous without taking into account the grades of malignancy for brain tumor tissue.

From the patent description P. 424257 the method for detecting brain tumors is known for use in the intraoperative determination of the margin between normal and tumorous tissue based on the ratio R of the intensity of Raman bands 2930 cm ¹ and 2844 cm ¹ , according to the-formula: intensity of the band 2930 cm ^-1 intensity of the band 2844 cm ¹ For human brain tumors, the intensity ratio I2930/I2844 for the non tumorous tissue is in the range 1.60 ± 0.63, for the tumorous tissue the range is: 2.56 ± 0.35.

The patent application P. 427376 describes the method for human brain tumors detection based on the ratio of Raman band intensities 1584 and 1440 cm ¹ described by the formula: intensity of the band 1584 cm ^-1 intensity of the band 1440 cm ^-1

After determining the value of the ratio, the method of statistical analysis of the data is analogous to that in the patent application P. 424257. Ultimately, for human brain tumors, the intensity ratio I1584/I1440 for non tumorous tissue is in the range 0.32 ± 0.18, for tumorous tissue the range is: 1.92 ± 0.64.

The method according to the invention solves the problem of determining the grades of malignancy of human brain tumors based on the calibration curves of Raman biomarkers and thus shortening the time of tumor diagnosis, which will be used in onco logical diagnostics and during routine "Raman histopathology" in diagnostic laboratories.

The method of determining the grades of malignancy of human brain tumors in optical detection using Raman spectroscopy according to the invention comprises 4 stages: sample preparation, measurements of Raman spectra, statistical analysis of data including determination of biomarkers sensitivity and specificity, determination of malig nancy grade calibration curves for human brain tumors. The first two steps are carried out analogously to the methods of the above-mentioned patent applications.

The method of determining the malignancy grades of human brain tumors in op tical detection using Raman spectroscopy compriseing the steps of: preparation of human brain tissue samples from the tumor mass and from the safety margin in the form of slices with a minimum thickness of 16 Elm, placing the tissue samples on a slide made of material with the transmittance in a wide spectral range of 200- 4000 cm ¹ , without use of any adhesive substances, with the shortest possible time from collection to the sample prep aration for measurements using the Raman spectroscopy, transportation of tissue sam ples to the room in which the measurements by using Raman spectroscopy will be per formed in no more than a few minutes in a way that minimizes sample destruction, cali- bration of a Raman microscope coupled with a confocal microscope or a Raman spectrom eter coupled with a fiber optical probe, based on the measurement of the Raman spec trum of the reference sample in the form of silicon - reference band 520,7 cm- ¹ or equiv alent, recording the Stokes component of the Raman spectra for samples at room tem perature using Raman spectrometer coupled with confocal microscope, connected to a computer equipped with software for recording and preprocessing of Raman spectra, or by bringing of the optical probe closer to the examined brain tissues using a Raman spec trometer coupled with a fiber optical probe, also connected to a computer equipped with a software for recording Raman spectra, in the spectral range 200 - 4000 cm ¹ using a CCD camera, wherein, when using a Raman spectrometer coupled to a confocal microscope, the samples are illuminated by laser light transmitted via a single-mode optical fibre, with a minimum beam power of 10 mW, spatial resolution minimum 1 pm, with integration time at least 0.5 s, using objectives from 20x to lOOx, using a monochromator with a dif fraction grating with a minimum number of grooves 1200/mm, or in the case of using a Raman spectrometer coupled with a fiber optical probe by illuminating the samples with laser light with a beam power of at least 25 mW, using holographic grid with volume phase, subjecting the recorded spectra to pre-treatment by eliminating cosmic rays, cut ting off the baseline using the computer software coupled with the Raman spectrometer or equivalent program, smoothing the spectrum, determining of analyzed biomarkers from the spectra after processing, according to the invention is characterized in that after preprocessing, the spectra are divided into three classes: protein-rich, fat-rich and mixed protein-fat profiles, of which as protein-rich and fat-rich spectra are selected average spectra from clusters created by using K-Means Cluster Analysis - number of clusters 2, wherein to the class of protein-rich spectra were assigned average spectra from cluster analysis , in which the ratio of the intensity of bands 2854 and 2930 cm ¹ was close to zero (0.1 +/- 0.01), for the fat-rich spectra were assigned spectra for which the ratio of the intensity of bands 2854 and 2930 cm ¹ was greater than 0.1, mixed protein-fat profile were assigned the average spectra from the whole area of the Raman imaging, whrein in a case of using the Raman spectrometer coupled with the optical probe, as spectra belonging to the mixed protein-fat spectrum are selected spectra, which are automatically generated, then, the intensity of the bands from the selected spectra in a program that allows to work with the recorded Raman spectra:2975, 2935, band with maximum intensity in the range of 2920-2940, bands 2885, 2854, 1660, 1586, 1444 cm ¹ are read out, further, the ratios of band intensities: 2975/2885, 2935/2885, 2854 /max (max - band with the maxi mum intensity in the range 2920-2940), 1586/1660, 1586/1444 which are Raman bi omarkers are calculated, after that the statistical analysis of Raman biomarkers is per formed, consisting in determining whether the statistical distribution of biomarker values is characterized by the normal distribution or not, for the case of the normal distribution, the Anova parametric tests are used in further analysis, and if the distribution is not nor mal, the Anova-Kruskal-Wallis test is used to determine whether the biomarker values for different grades of brain tumors belong to statistically different groups at a given level of statistically significance p<0.05, moreover, the sensitivity and specificity of Raman bi omarkers are determined using the PLS-DA (Partial Least Squares Discriminant Analysis) method - and cross-validation. Then, the calibration curves are drawn to determine the intervals characteristic of each grade G1-G4 of tumor and for normal tissue marked GO, whereby the calibration curve for the Raman biomarker 11586/1444 is described by equations: for the tumor grades G0-G2 y = 0,37+ 0,005x, for the tumor grades G2-G3 y = -3,19+l,78x, for the tumor grades G3-G4 y = -2,6+0,15x, in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker Iis86/i444, the calibration curve for the Raman biomarker I1586/1660 is described by equations: for the tumor grades G0-G2 y = 0,36 + 0,028x, for the tumor grades G2-G3 y = -6,84 + 3,63x, for the tumor grades G3-G4 y = 11,1 - 2,35x, in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker I1586/1660, the calibration curve for the Raman biomarker 1 ₂₈₅₄ / _max is described by equations: for the tumor grades G0-G1 y = 0,8 - 0,4x, for the tumor grades G1-G4 y = 0,42 + 0,001x, in which x is the grade of tumorf(for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker 1 ₂₈₅₄ / _max , the calibration curve for the Raman biomarker I2975/2885 is described by equations: for the tumor grades G0-G1 y = 0,38 + 0,26x, for the tumor grades G1-G4 y = 0,65 + 0,0054x, in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker I2975/2885, the calibration curve for the Raman biomarker I2935/2885 is described by equations: for the tumor grades G0-G1 y = 0,99 + 0,47x, for the tumor grades G1-G4 y = 1,45 + 0,0036x, in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker I2935/2885, and for a sample of human brain tissue with an unknown G grade, the Raman biomarker value plotted on the calibration curve enables reading out of the G value of this sample, whereby for biomarkers 2975/2885 i 1586/1444 in order to obtain the most accurate di agnosis, an additional 2D plot of the value of the biomarker 2975/2885 as a function of value of the biomarker 1586/1444 is drawn.

Preferably a sample of human brain tissue is put to the CaF ₂ slide.

Equally preferably, the placement of human brain tissue samples on the slide takes in a period of minutes.

Preferably, dry ice is used to transport tissue samples to the room where the measurement using Raman spectroscopy is performed.

It is also good if the sample is illuminated using 532 nm laser line.

Preferably, the base line of the Raman spectra is cut off when processing using ORIGIN software.

The Raman spectrum is smoothed using the Savitzky-Golay method using the number of points 5 and the order of polynomial 2.

From the selected spectra the intensities of the bands: 2975, 2935, bands with a maximum intensity in the range 2920-2940, bands 2885, 2854, 1660, 1586, 1444 cm ¹ are read out, using the software for confocal microscope or the ORIGIN software.

Preferably, Statistical analysis of calculated Raman biomarkers is performed using ORIGIN software.

The method according to the invention allows in a very short time of the order of minutes, based on the obtained Raman biomarker calibration curves to determine une quivocal the grades of malignancy of human brain tumors for each type of tumor, which will be used in oncological diagnostics and during routine "Raman histopathology" in di agnostic laboratories.

To date, methodologies requiring several hours of tissue preprocessing are used, so the proposed solution in the invention is a response to the needs of surgeons for quick and unambiguous diagnostics. The solution helps to reduce the waiting time for the result and eliminates the need for multistage tissue staining with the use of exogenous dyes, which is typical for histopathological analysis. The method of the invention is a fast, ob jective and simple tool for assessing the grade of malignancy in real-time in clinical and diagnostic practice based on Raman vibrational spectra of human brain tissue.

The invention will now be described by way of examples and with reference to the accompanying drawings in which Fig. 1-6 show the calibration curves obtained in the example of Raman biomarkers dependence on the tumor grades G1-G4, GO corresponds to non tumorous tissue.

Example

For the preparation of Raman biomarkers calibration curves, human brain tumor ous tissue samples from patients diagnosed with cancer were used, with the prior in formed consent of each patient and with the consent of the Bioethics Committee con firmed by an appropriate document.

The analysis was performed for the human brain samples of: medulloblastoma, embryonic tumor PNS, ependynoma anaplastic, ependymoma, astrcytoma fibrous, astro cytoma, ganglioma, astrocytoma pilocytic, subependymoma, hemangioblastoma, cranio pharyngioma, dysembryoplastic neuroepithelial tumor, papillary glioneuronal tumor with different grades of malignancy. The samples of normal human brain tissue and rat were also used.

Tissue samples were cut using a microtome into 16 0m thick slices and trans ported within a few minutes to the room where the Raman spectrometer was located. For the preparation of Raman spectra, a confocal Raman spectrometer connected to a computer equipped with software for eliminating cosmic rays, subtracting the baseline and smoothing the spectra, calibrated by performing the Raman spectrum of the silicon sample was used. The samples were illuminated with laser light of wavelength 532 nm and a power of 25 mW transmitted through the single-mode fiber with a diameter of 50 pm. The spectra were recorded at room temperature using a grid with the number of grooves 1200 grooves/mm in the spectral range 200 - 4000 cm ¹ using a CCD camera op erating at -60 ^Q C.

Raman spectra of the Stokes component of Raman scattering were recorded. Ob tained spectra were preprocessed using the ORIGIN software and the software of the computer coupled with Raman spectrometer. The analysis involving the elimination of cosmic rays and subtracting the baseline so that the baseline was at zero position. Each spectrum was smoothed using the Savitzky-Golay method for the parameters: number of points 5, order of polynomial 2.

After preprocessing, the spectra are divided into protein-rich, fat-rich and mixed protein-fat profiles. Protein-rich and fat-rich spectra were obtained as a result of registra tion either an average spectrum of clusters generated by the Cluster Analysis -K-Means Cluster Analysis (number of clusters 2), wherein to the class of protein-rich spectra were assigned the average spectra from cluster analysis, in which the ratio of the intensity of bands 2854 and 2930 cm ¹ was close to zero (0.1 +/- 0.01), to fat-rich spectra were as signed spectra for which the ratio of the intensity of bands 2854 and 2930 cm ¹ was greater than 0.1. The protein-fat mixed spectra were generated automatically by using a Raman spectrometer coupled with fiber optical probes or as averaged spectra from the whole area of the Raman imaging for a confocal Raman microscope. Then, in the ORIGIN software enabling work with the registered Raman spectra, the intensity of bands in [cm ¹ ]: 2975, 2935, band with the maximum intensity in the range of 2920-2940, 2854, 1660, 1586, 1444 was read out for appropriately classified Raman spectra. The following band intensity ratios were determined: 2975/2885, 2935/2885, 2854 / maximum in the range 2920-2940, 1586/1660, 1586/1444, which were Raman biomarkers. Subsequently, the statistical analysis for the calculated values of these biomarkers was performed. First, us ing the Shapiro-Wilk test, it was determined that non-parametric tests should be used for all biomarkers, because the Shapiro-Wilk test had showed lack of normal distribution, and the ANOVA (Kruskal - Wallis'a) test was used in further proceedings. Statistical significance was set at p<0.05. The Kruskal-Wallis test confirmed that Raman biomarker values belong to groups statistically significantly different. The sensitivity and the specificity were ob tained directly using PLS-DA (Partial Least Squares Discriminant Analysis) and cross-vali dation. For Raman biomarker I1586/1444 (for a ratio of band intensities 1586/1444) in spectra with a mixed fat-protein profile from the entire imaging area of human brain tumor tissue at 532 nm excitation and normal rat brain tissue (type Wistar) at 785 nm excitation, the Kruskal- Wallis test at the confidence level p < 0.05 has confirmed that the values of this biomarker for particular grades of brain tumors belong to statistically different groups.

The established Raman biomarker values for individual tumor malignancy grades are shown in Table 1 with SD - standard deviation.

Table 1

Based on the average values obtained for the Raman biomarker Iis86/i444, a cali bration curve of the dependence of this biomarker's values on the malignancy tumor grade was drawn up. The calibration curve is shown in Fig. 1 of the figure. Figure 1 shows the ratio of the intensity of 1586 to 1444 cm ¹ bands as a function of brain tumor malig- nancy grade based on the average spectra of the whole area of the imaging- mixed pro tein-fat profile, wherein ■ concerns human brain tissue, at excitation 532 nm, · concerns healthy rat brain tissue (Wistar type), at excitation 785 nm.

The calibration curve for the Raman biomarker Iis86/i444 is described by equations: for the tumor grades G0-G2 y = 0.37+ 0.005x, for the tumor grades G2-G3 y = -3.19+1.78x, for the tumor grades G3-G4 y = -2.6+0.15x, in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker I1586/1444.

Because tumor development is a continuous process, the boundary between G2 and G3 grades was determined from the equation y = -3.19 + 1.78x for x = 2.5, then for the above-mentioned biomarker, values below 1.26 should be assigned to the malignancy grade of G0-G2, values above 1.26 to the grades G3 / G4.

The sensitivity and specificity for this biomarker were obtained directly from PLS- DA and cross-validation and were equal to: sensitivity: 100% for G0-G2 and 100% for G3-G4 for calibration specificity: 100% for G0-G2 and 100% for G3-G4 for calibration sensitivity average value: 100%, specificity average value: 100% for calibration sensitivity: 88.9% for G0-G2 and 100% for G3-G4 for cross-validation specificity: 100% for G0-G2 and 88,9% for G3-G4 for cross-validation sensitivity average value: 94.45%, specificity average value: 94.45% for cross-validation.

For a brain tissue sample of unknown G, the determined value of the Raman bi omarker Ii /i was plotted on the prepared calibration curve and the readout value corresponded to the GO value, which meant that the brain tissue was non tumorous tis sue. For Raman biomarker Iis ₈₆ /i ₆₆ o(for a ratio of band intensities 1586/1660) in spec tra with a protein-rich profile from the cluster analysis (K-Means Cluster Analysis - num ber of clusters 2) for brain tumor tissue at 532 nm excitation and normal rat brain tissue (type Wistar) at 785 nm excitation, the Kruskal-Wallis test at the confidence level p < 0.05 has confirmed that the values of this biomarker for particular grades of brain tumors be- long to statistically different groups.

The established Raman biomarker values for individual tumor malignancy grades are shown in Table 2 with SD - standard deviation.

Table 2 Based on the average values obtained for the Raman biomarker I1586/1660, a cali bration curve of the dependence of this biomarker's values on the malignancy tumor grade was drawn up. The calibration curve is shown in Fig. 2 of the figure. Figure 2 shows the ratio of the intensity of 1586 to 1660 cm ¹ bands as a function of brain tumor malig nancy grade based on the protein-rich spectra from the cluster analysis (K-Means Cluster Analysis-number of clusters 2), wherein ■ concerns human brain tissue, at excitation 532 nm, · concerns healthy rat brain tissue (Wistar type), at excitation 785 nm.

The calibration curve for the Raman biomarker I1586/1660 is described by equations: for the tumor grades G0-G2 y = 0.36 + 0.028x, for the tumor grades G2-G3 y = -6.84 + 3.63x, for the tumor grades G3-G4 y = 11.1 - 2.35x, in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker I1586/1660.

Because tumor development is a continuous process, the boundary between G2 and G3 grades was determined from the equation y = -6.84 + 3.63x for x = 2.5, then for the above-mentioned biomarker, values below 2.23 should be assigned to the malignancy grades G0-G2, values above 2.23 to the grades G3 / G4.

The sensitivity and specificity for this biomarker were obtained directly from PLSDA and cross-validation and are equal to: sensitivity: 100% for G0-G2 and 100% for G3-G4 for calibration specificity: 100% for G0-G2 and 100% for G3-G4 for calibration sensitivity average value: 100%, specificity average value: 100% for calibration sensitivity: 88.9% for G0-G2 and 100% for G3-G4 for cross-validation specificity: 100% for G0-G2 and 88.9% for G3-G4 for cross-validation sensitivity average value: 94.45%, specificity average value: 94.45% for cross-validation.

For a brain tissue sample of unknown G, the determined value of the Raman bi omarker I1586/1660 was plotted on the prepared calibration curve and the readout value corresponded to the GO value, which meant that the brain tissue was non tumorous tis sue.

For Raman biomarker 1 ₂₈₅₄ / _max (for a ratio of band intensities 2854/band of the maximum intensity from the range 2920-2940 cm ¹ ) in spectra with a mixed fat-protein profile from the entire imaging area of human brain tumor tissue at 532 nm excitation and normal rat brain tissue (type Wistar) at 671 nm excitation, the Kruskal-Wallis test at the confidence level p < 0.05 has confirmed that the values of this biomarker for particular grades of brain tumors belong to statistically different groups.

The established Raman biomarker values for individual tumor malignancy grades are shown in Table 3 with SD - standard deviation.

Table 3

Based on the average values obtained for the Raman biomarker 1 ₂₈₅₄ / _max , a calibra tion curve of the dependence of this biomarker's values on the malignancy tumor grade was drawn up. The calibration curve is shown in Fig. 3 of the figure. Figure 3 shows the ratio of the intensity of 2854 cm to max (maximum band from the range 2920-2940 cm ¹ ) as a function of brain tumor malignancy grade based on the average spectra of the whole area of the imaging- mixed protein-fat profile, wherein ■ concerns human brain tissue, at excitation 532 nm, · concerns healthy rat brain tissue (Wistar type), at excitation 671 nm.

The calibration curve for the Raman biomarker 1 ₂₈₅₄ / _max was described by equa tions: for the tumor grades G0-G1 y = 0.8 - 0.4x, for the tumor grades G1-G4 y = 0.42 + O.OOlx. in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker 1 ₂₈₅₄ / _max -

Because tumor development is a continuous process, the boundary between GO and G1 grades was determined from the equation y = 0.8 - 0,4x for x = 0.5, then for the above-mentioned biomarker, values above 0.6 should be assigned to the grade GO, values below 0.6 to malignancy tumor grades G1-G4. The sensitivity and specificity for this biomarker were obtained directly from PLSDA and cross-validation and are equal to: sensitivity: 100% for GO and 100% for G1-G4 for calibration specificity: 100% for GO and 100% for G1-G4 for calibration sensitivity average value: 100%, specificity average value: 100% for calibration sensitivity: 100% for GO and 100% for G1-G4 for cross-validation specificity: 100% for GO and 100% for G1-G4 for cross-validation sensitivity average value: 100%, specificity average value: 100% for cross-validation.

For a brain tissue sample of unknown G, the determined value of the Raman biomarker 12854/ _max was plotted on the prepared calibration curve and the readout value corresponded to the GO value, which meant that the brain tissue was non tumorous tissue.

For Raman biomarker I2975/2885 (for a ratio of band intensities 2975/2885) in spec tra with a protein-rich profile from the cluster analysis (K-Means Cluster Analysis - num- ber of clusters 2) for brain tumor tissue at 532 nm excitation and normal rat brain tissue (type Wistar) at 671 nm excitation, the Kruskal-Wallis test at the confidence level p < 0.05 has confirmed that the values of this biomarker for particular grades of brain tumors be long to statistically different groups.

The established Raman biomarker values for individual tumor malignancy grades are shown in Table 4 with SD - standard deviation.

Table 4

Based on the average values obtained for the Raman biomarker I ₂₉₇₅ / ₂₈₈₅ , a cali bration curve of the dependence of this biomarker's values on the malignancy tumor grade was drawn up. The calibration curve is shown in Fig. 4 of the figure. Figure 4 shows the ratio of the intensity of 2975 to 2885 cm-1 bands as a function of brain tumor malig nancy grade - based on the protein-rich spectra from the Cluster Analysis (K-Means Clus ter Analysis - number of clusters 2), wherein ■ concerns human brain tissue, at excitation 532 nm, · concerns healthy rat brain tissue (Wistar type), at excitation 671 nm,

The calibration curve for the Raman biomarker I2975/2885 was described by equa tions: for the tumor grades G0-G1 y = 0.38 - 0.26x, for the tumor grades G1-G4 y = 0.65 + 0.0054x. in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker I2975/2885.

Because tumor development is a continuous process, the boundary between GO and G1-G4 grades was determined from the equation y = 0.38 - 0,26x for x = 0.5, then for the above-mentioned biomarker, values below 0.51 should be assigned to the grade GO, values above 0.51 to malignancy grades G1-G4.

The sensitivity and specificity for this biomarker were obtained directly from PLS-DA and cross-validation and are equal to: sensitivity: 100% for GO and 91.7% for G1-G4 for calibration, specificity: 91.7% for GO and 100% for G1-G4 for calibration, sensitivity average value: 95.8%, specificity average value: 95.8% for calibration, sensitivity: 100% for GO and 91.7% for G1-G4 for cross-validation, specificity: 91.7% for GO and 100% for G1-G4 for cross-validation, sensitivity average value: 95.8%, specificity average value: 95.8% for cross-validation.

For a brain tissue sample of unknown G, the determined value of the Raman bi omarker 12875/2885 was plotted on the prepared calibration curve and the readout value corresponded to the GO value, which meant that the brain tissue was non tumorous tis sue.

For Raman biomarker I2935/2885 (for a ratio of band intensities 2935/2885) in spec tra with a protein-rich profile from the cluster analysis (K-Means Cluster Analysis - num ber of clusters 2) for brain tumor tissue at 532 nm excitation and normal rat brain tissue (type Wistar) at 671 nm excitation, the Kruskal-Wallis test at the confidence level p < 0.05 has confirmed that the values of this biomarker for particular grades of brain tumors be long to statistically different groups. The established Raman biomarker values for individual tumor malignancy grades are shown in Table 5 with SD - standard deviation.

Table 5

Based on the average values obtained for the Raman biomarker I2935/2885, a cali bration curve of the dependence of this biomarker's values on the malignancy tumor grade was drawn up. The calibration curve is shown in Fig. 5 of the figure. Figure 5 shows the ratio of the intensity of 2935 to 2885 cm ¹ bands as a function of brain tumor malig nancy grade - based on the protein-rich spectra from the Cluster Analysis (K-Means Clus ter Analysis - number of clusters 2), wherein ■ concerns human brain tissue, at excitation 532 nm, · concerns healthy rat brain tissue (Wistar type), at excitation 671 nm.

The calibration curve for the Raman biomarker I2935/2885 was described by equa tions: for the tumor grades G0-G1 y = 0.99 + 0.47x, for the tumor grades G1-G4 y = 1.45 + 0.0036x. in which x is the grade of tumor (for GO x = 0, for G1 x = 1, for G2 x = 2, for G3 x = 3, for G4 x = 4), y is the value of Raman biomarker I2935/2885.

Because tumor development is a continuous process, the boundary between GO and G1-G4 grades was determined from the equation y = 0.99 + 0.47x for x = 0.5, then for the above-mentioned biomarker, values below 1.2 should be assigned to the grade GO, values above 1.2 to malignancy grades G1-G4.

The sensitivity and specificity for this biomarker were obtained directly from PLS-DA and cross-validation and are equal to: sensitivity: 100% for GO and 100%for G1-G4 for calibration specificity: 100% for GO and 100% for G1-G4 for calibration sensitivity average value: 100%, specificity average value: 100% for calibration sensitivity: 100% for GO and 100%for G1-G4 for cross-validation specificity: 100%for GO and 100% for G1-G4 for cross-validation sensitivity average value: 100%, specificity average value: 100% for cross-validation.

For a brain tissue sample of unknown G, the determined value of the Raman bi omarker 12835/2885 was plotted on the prepared calibration curve and the readout value corresponded to the GO value, which meant that the brain tissue was non tumorous tis sue.

The calibration curve representing a 2D graph of the values of the biomarker 1 ₂₉₇₅ / ₂₈₈₅ value as a function of the values of the biomarker Il ₅ 86/i444was also prepared - Fig. 6 of the Figure. Fig. 6 shows 2D graph of values for biomarker 2975/2885 as a function of values for biomarker 1586/1444.

Within the framework of the 2D chart, these biomarkers differentiated brain tu mors GO from G1-G4 tumor grades and G1-G2 from G3-G4 tumor grades. The biomarker 1 ₂₉₇₅ / ₂₈₈₅ individually differentiates GO from G1-G4 grades, the biomarker Iis86/i444 individ ually differentiates G0-G2 from G3-G4 grades.

The calibration curves of biomarkers 2975/2885 and 1586/1444 were considered individually and as 2D plot to obtain the most accurate diagnosis.

All malignancy calibration curves for brain tumors described above based on Ra man biomarkers were used to determine the grades of human brain tumors on the Gl- G4 scale and to distinguish tumors from normal tissue described as GO based on ratios 2975/2885, 2935/2885, 2854 / max (max is the band with the highest intensity in the range 2920-2940 cm-1), 1586/1660, 1586/1444 of Raman bands intensities. Calibration curves of the grades of malignancy of brain tumors based on Raman biomarkers values were used to determine the safety margin and during "Raman histopathology" in diag nostic laboratories. For a sample with an unknown G, Raman biomarker values were plot ted on the corresponding calibration curves and G values were readout. Sensitivity and specificity for each Raman biomarker are given in the example. The result was unambig uous when the same G grade score was obtained from all calibration curves.