Digital simulation of staining in histopathology multispectral images: enhancement and linear transformation of spectral transmittance

• Published in: Journal of Biomedical Optics Vol. 17; no. 5; p. 056013
• Main Authors: Bautista, Pinky A, Yagi, Yukako
• Format: Journal Article
• Language: English
• Published: United States 01.05.2012
• Subjects: Algorithms; Digital; Digital simulation; digital staining; Humans; hyperspectral imaging; image analysis; Image enhancement; Image Enhancement - methods; Image Interpretation, Computer-Assisted - methods; image processing; Image transmission; image understanding; Linear Models; Liver - cytology; Microscopy - methods; multispectral enhancement; Pixels; Reproducibility of Results; Sensitivity and Specificity; Signal Processing, Computer-Assisted; Spectra; Spectrum Analysis - methods; Staining; Staining and Labeling; Transformations
• ISSN: 1083-3668; 1560-2281
• DOI: 10.1117/1.JBO.17.5.056013

Hematoxylin and eosin (H&E) stain is currently the most popular for routine histopathology staining. Special and/or immuno-histochemical (IHC) staining is often requested to further corroborate the initial diagnosis on H&E stained tissue sections. Digital simulation of staining (or digital staining) can be a very valuable tool to produce the desired stained images from the H&E stained tissue sections instantaneously. We present an approach to digital staining of histopathology multispectral images by combining the effects of spectral enhancement and spectral transformation. Spectral enhancement is accomplished by shifting the -band original spectrum of the multispectral pixel with the weighted difference between the pixel's original and estimated spectrum; the spectrum is estimated using < principal component (PC) vectors. The pixel's enhanced spectrum is transformed to the spectral configuration associated to its reaction to a specific stain by utilizing an × transformation matrix, which is derived through application of least mean squares method to the enhanced and target spectral transmittance samples of the different tissue components found in the image. Results of our experiments on the digital conversion of an H&E stained multispectral image to its Masson's trichrome stained equivalent show the viability of the method.