Stain Normalization: A method of normalizing breast cancer histopathological images using an algorithm that converts RGB slides to optical density (OD) and removes transparent pixels or data with lower than threshold β. The algorithm then computes singular value decomposition (SVD) for OD tuples, creates a plane from the SVD directions corresponding to the two largest singular values (higher variance), projects the values onto the plane spanned by the eigenvectors corresponding to the two largest eigenvalues, calculates the angle of each point for the first SVD direction, determines the stain concentrations by finding robust extremes, α th and (100−α) th percentiles of the angle, converts the extreme values back to OD space, and finally deconvolves the image using the determined stain vectors.
Stain Normalization: A method of normalizing breast cancer histopathological images using an algorithm that converts RGB slides to optical density (OD) and removes transparent pixels or data with lower than threshold β. The algorithm then computes singular value decomposition (SVD) for OD tuples, creates a plane from the SVD directions corresponding to the two largest singular values (higher variance), projects the values onto the plane spanned by the eigenvectors corresponding to the two largest eigenvalues, calculates the angle of each point for the first SVD direction, determines the stain concentrations by finding robust extremes, α th and (100−α) th percentiles of the angle, converts the extreme values back to OD space, and finally deconvolves the image using the determined stain vectors.