Method for computer-assisted analysis of one or more tissue sections of the human or animal body

1. A computer-implemented method for analysis of one or more tissue sections (TI) of a human or animal body for preparing a digital stain test, wherein in the digital stain test tissue zones that have a predetermined clinical picture are determined in a tested tissue section (TI′), the method comprising:

a) obtaining spatially resolved hyperspectral data from one or more predetermined tissue sections (TI) that are to be analysed and contain tissue zones having the predetermined clinical picture, wherein the hyperspectral data is mass spectrometry data and contains respective spectra (S ij ) for a plurality of measurement positions (x) in the one or more predetermined tissue sections (TI), wherein a respective spectrum (S ij ) represents a composition of the one or more predetermined tissue sections (TI) at a measurement position (x) in the form of a vector consisting of intensity values (I) for a plurality of channels (m/z), wherein an intensity value (I) is a physical parameter detected for a corresponding channel (m/z);

b) obtaining a mask (M ij ) of the one or more predetermined tissue sections (TI) containing respective scale values for the plurality of measurement positions (x) wherein zones of the one or more predetermined tissue sections (TI) having scale values that differ from the remaining area of the one or more tissue sections (TI) are tissue zones with the predetermined clinical picture;

c) determining multiple base spectra (BS 1 , BS 2 , . . . , BS 5 ) in a vector space of the hyperspectral data or in a space of derived vectors that are calculated from the hyperspectral data for each measurement position (x) using a method for calculating base vectors;

d) generating a plurality of correlation values (Pijk) for respective base spectra (BS 1 , BS 2 , . . . , BS 5 ) for each measurement position (x) by correlating the base spectra (BS 1 , BS 2 , . . . , BS 5 ) with the vectors of the hyperspectral data or with the derived vectors at each measurement position (x); and

e) generating a regression model (RM) describing the mask (M ij ) based on the plurality of correlation values (P ijk ) for the respective measurement positions (x) using a regression method, wherein the regression model (RM) and the base spectra (BS 1 , BS 2 , . . . , BS 5 ) are stored digitally.

2. The method according to claim 1 , wherein the mass spectrometry data is Matrix-Assisted Laser Desorption/Ionization (MALDI) mass spectrometry data, in which the channels (m/z) represent mass-to-charge ratios and the physical parameter detected for the corresponding channel (m/z) represents the number of molecules detected for the corresponding mass-to-charge ratio.

3. The method according to claim 1 , wherein in step c) the base spectra (BS 1 , BS 2 , . . . , BS 5 ) are determined in the space of the derived vectors, wherein a derived vector is calculated for each measurement position (x) in such manner that

for the predetermined clinical picture, hyperspectral data is prespecified for a plurality of molecular compositions, wherein the hyperspectral data for a molecular composition contains a vector in the form of a spectrum characteristic of the molecular composition consisting of intensity values (I) for a predetermined number of channels (m/z) from the plurality of channels (m/z);

the spectra characteristic of the molecular compositions are correlated with the spectra (S ij ) of the read in hyperspectral data at the respective measurement position (x), with the result that a vector consisting of multiple correlation coefficients is calculated as the derived vector for each measurement position (x).

4. The method according to claim 1 , wherein the one or more tissue sections to be analysed are each tissue sections that have undergone enzymatic digestion.

5. The method according to claim 1 , wherein the method used in step c) for determining base vectors is a non-negative matrix factorization or a PCA analysis.

6. The method according to claim 5 , wherein the non-negative matrix factorization includes the side condition of sparse population of the matrices determined using non-negative matrix factorization.

7. The method according to claim 1 , wherein the regression method used in step e) is a linear regression method or an SVM regression method.

8. The method according to claim 1 , wherein the mask (M ij ) in step b) is determined from a histological, particularly an immunohistochemical stain test of the one or more predetermined tissue sections.

9. The method according to claim 1 , wherein the predetermined clinical picture represents the presence of predetermined tumour cells or the presence of insulin islands in the one or more predetermined tissue sections (TI).

10. The method according to claim 1 , further comprising:

i) obtaining spatially resolved hyperspectral data from a preset tissue section (TI′) to be tested, wherein the hyperspectral data contains respective spectra (S ij ) for a plurality of measurement positions (y) in the preset tissue section (TI′), wherein a respective spectrum (S ij ) represents the composition of the preset tissue section (TI′) at a measurement position (y) in the form or a vector consisting of intensity values (I) for channels (m/z) from a plurality of channels (m/z), wherein an intensity value (I) is a physical parameter detected for the corresponding channel (m/z),

ii) generating a plurality of correlation values (Nijk) for respective base spectra (BS1, BS2, . . . , BS5) for each measurement position (y) by correlating the vectors of the hyperspectral data or derived vectors calculated from the hyperspectral data at each measurement position (x) with the base spectra (BS 1 , BS 2 , . . . , BS 5 );

iii) obtaining a mask (MA ij ) that contains respective scale values for the plurality of measurement positions (y), wherein zones of the preset tissue section (TI′) with scalar values that differ from the rest of the tissue section (TI′) are tissue zones with the predetermined clinical picture, by applying the regression model (RM) to the plurality of correlation values (Nijk).

11. The method according to claim 10 , further comprising:

generating a visualization on a display device comprising an image of pixels representing the measurement positions (y), wherein each pixel encodes the associated scale value of the mask (MA ij ) determined in step iii) for the measurement position (y) represented by the respective pixel (BP).

12. The method according to claim 10 , wherein the vectors derived in step ii) are correlated with the base spectra (BS 1 , BS 2 , . . . , BS 5 ), and the derived vectors are calculated in such manner that

for the predetermined clinical picture, hyperspectral data is prespecified for a plurality of molecular compositions, wherein the hyperspectral data for a molecular composition contains a vector in the form of a spectrum of intensity values (I) characteristic of the molecular composition for a predetermined number of channels (m/z) from the plurality of channels (m/z); and

the spectra characteristic of the molecular compositions are correlated with the spectra (S ij ) of the hyperspectral data obtained in step i) at the respective measurement positions (y), with the result that a vector consisting of multiple correlation coefficients is determined as a derived vector for each measurement position (y).

13. An apparatus for computer-assisted analysis of one or more tissue sections (TI) of the human or animal body for preparing a digital stain test, wherein, during the digital stain test, tissue zones in a tested tissue section (TI′) that have a predetermined clinical picture are detected with computer assistance, the apparatus comprising:

a) means for obtaining spatially resolved hyperspectral data from one or more predetermined tissue sections (TI) that are to be analysed and contain tissue zones having the predetermined clinical picture, wherein the hyperspectral data is mass spectrometry data and contains respective spectra (S ij ) for a plurality of measurement positions (x) in the one or more predetermined tissue sections (TI), wherein a respective spectrum (S ij ) represents a composition of the one or more predetermined tissue sections (TI) at a measurement position (x) in the form of a vector consisting of intensity values (I) for a plurality of channels (m/z), wherein an intensity value (I) is a physical parameter detected for a corresponding channel (m/z);

b) means for obtaining a mask (M ij ) of the one or more predetermined tissue sections (TI) containing respective scale values for the plurality of measurement positions (x), wherein zones of the one or more predetermined tissue sections (TI) having scale values that differ from the remaining area of the one or more tissue sections (TI) are tissue zones with the predetermined clinical picture;

c) means for determining multiple base spectra (BS 1 , BS 2 , . . . , BS 5 ) in the vector space of the hyperspectral data or in a space of derived vectors that are calculated from the hyperspectral data for each measurement position (x) using a method for calculating base vectors;

d) means for generating a plurality of correlation values (Pijk) for respective base spectra (BS 1 , BS 2 , . . . , BS 5 ) for each measurement position (x) by correlating the base spectra (BS 1 , BS 2 , . . . , BS 5 ) with the vectors of the hyperspectral data or with the derived vectors at each measurement position (x); and

e) means for generating a regression model (RM) describing the mask (M ij ) based on the plurality of correlation values (P ijk ) for the respective measurement positions (x) using a regression method, wherein the regression model (RM) and the base spectra (BS 1 , BS 2 , . . . , BS 5 ) are stored digitally.

14. A computer program product having a program code stored on a machine-readable medium, wherein the program code when executed on a computer causes the computer to perform the steps of:

a) obtaining spatially resolved hyperspectral data from one or more predetermined tissue sections (TI) that are to be analysed and contain tissue zones having the predetermined clinical picture, wherein the hyperspectral data is mass spectrometry data and contains respective spectra (S ij ) for a plurality of measurement positions (x) in the one or more predetermined tissue sections (TI), wherein a respective spectrum (S ij ) represents a composition of the one or more predetermined tissue sections (TI) at a measurement position (x) in the form of a vector consisting of intensity values (I) for a plurality of channels (m/z), wherein an intensity value (I) is a physical parameter detected for a corresponding channel (m/z);

b) obtaining a mask (M ij ) of the one or more predetermined tissue sections (TI) containing respective scale values for the plurality of measurement positions (x), wherein zones of the one or more predetermined tissue sections (TI) having scale values that differ from the remaining area of the one or more tissue sections (TI) are tissue zones with the predetermined clinical picture;

c) determining multiple base spectra (BS 1 , BS 2 , . . . , BS 5 ) in the vector space of the hyperspectral data or in a space of derived vectors that are calculated from the hyperspectral data for each measurement position (x) using a method for calculating base vectors;

d) generating a plurality of correlation values (Pijk) for respective base spectra (BS 1 , BS 2 , . . . , BS 5 ) for each measurement position (x) by correlating the base spectra (BS 1 , BS 2 , . . . , BS 5 ) with the vectors of the hyperspectral data or with the derived vectors at each measurement position (x); and

e) generating a regression model (RM) describing the mask (M ij ) based on the plurality of correlation values (P ijk ) for the respective measurement positions (x) with the aid of a regression method, wherein the regression model (RM) and the base spectra (BS 1 , BS 2 , . . . , BS 5 ) are stored digitally.