Method and system for simultaneous decomposition of multiple hyperspectral datasets and signal recovery of unknown fluorophores in a biochemical system

1. A method for generating an isolated autofluorescence image of lipofuscins endogenously generated in a retinal pigment epithelium (RPE) of an eye from composite autofluorescence detected from a plurality of fluorophors in the eye, wherein the plurality of fluorophors comprises the lipofuscins, the method comprising:

obtaining, from a plurality of hyperspectral sensors, a first data matrix representative of a first electromagnetic emission spectrum from the plurality of fluorophors in response to an excitation signal;

isolating, with a processor, a second data matrix from the first data matrix, the second data matrix being representative of a second electromagnetic emission spectrum from a retinal pigment epithelium (RPE) of the eye in response to the excitation signal;

identifying, with the processor, one or more Gaussian functions, wherein each Gaussian function is an initial approximation representative of a spectral component of the second electromagnetic emission spectrum;

analyzing, with the processor, the second data matrix using non-negative matrix factorization initialized with the one or more Gaussian functions to generate one or more component matrices, wherein each component matrix is a second approximation representative of the corresponding spectral component of the second electromagnetic emission spectrum; and

generating, with the processor, an image of the lipofuscins corresponding to one of the component matrices, wherein the image approximately corresponds, in part, to a histological image of the eye;

quantifying, with the processor, an intensity of one of the spectral components represented by the corresponding component matrix; and

administering a treatment for a state of AMD disease progression corresponding to the quantified intensity.

2. The method of claim 1 , wherein a composite of the one or more Gaussian functions approximates the second data matrix.

3. The method of claim 1 , wherein the number of Gaussian functions corresponds to the number of peaks within the second electromagnetic emission spectrum.

4. A method for generating an isolated autofluorescence image of a component fluorophor from a plurality of fluorophors in an eye from composite autofluorescence detected from the fluorophors in the eye, comprising:

obtaining, from a plurality of hyperspectral sensors, a first data cube representative of first electromagnetic emission spectra from the plurality of fluorophors in response to excitation signals having two or more different wavelengths;

isolating, with a processor, a second data cube from the first data cube, the second data cube comprising a plurality of data matrices, each data matrix being representative of an individual emission spectrum from a retinal pigment epithelium (RPE) of the eye in response to the excitation signal at each of the different wavelengths;

identifying, with the processor, one or more Gaussian functions for each data matrix, wherein each Gaussian function is an initial approximation representative of a spectral component of the corresponding individual emission spectrum;

analyzing, with the processor, the second data cube using non-negative tensor factorization initialized with the one or more Gaussian functions for each data matrix to generate one or more component matrix functions, wherein the component matrix functions are concatenated across the plurality of data matrices such that each component matrix function is a common approximation representative of each corresponding spectral component shared across each individual emission spectrum; and

generating, with the processor, an image from one of the plurality of fluorophors corresponding to one of the component matrix functions, wherein the image approximately corresponds, in part, to a histological image of the eye, wherein the fluorophors comprise lipofuscins endogenously generated in the RPE;

quantifying, with the processor, an intensity of one of the spectral components represented by the corresponding component matrix; and

administering a treatment for an eye disease corresponding to the quantified intensity.

5. The method of claim 4 , wherein the excitation signals have three or more different wavelengths.

6. The method of claim 4 , wherein a composite of the one or more Gaussian functions for each data matrix approximates the corresponding data matrix.

7. The method of claim 4 , wherein the number of Gaussian functions for each data matrix corresponds to the number of peaks within each individual emission spectrum.

8. The method of claim 4 , wherein the excitation signals are in 10 nm intervals between 420 nm to 720 nm.