Kavli Affiliate: Denis Wirtz
| Authors: Florin Selaru, Jude Phillip, Denis Wirtz and Pei-Hsun Wu
| Summary:
Advancements in computational approaches have enabled robust utilization of histological tissue data. A crucial step in the development of computational tools for the objective and quantitative analysis of tissue sections has been color deconvolution. Color deconvolution functions by separating the absorption of colors corresponding to stained molecular or tissue compartments. The most widely used color deconvolution method in digital pathology, linear color deconvolution as described in Ruifrok 2001 et al., decomposes color images according to the absorbance values for individual stains. However, linear color deconvolution assumes that stains are linearly decomposable, and it relies heavily upon identifying optimal color vectors of stains, which is often challenging. Furthermore, linear deconvolution methods cannot deconvolve the image with more than three stains, further limiting their broader applicability. To combat the limitations of previous methods, we developed an intuitive and robust color deconvolution method that effectively and accurately separates more than three stain signals, does not rely on predetermined color vectors, and doesn’t rely on identifying optimal stain vectors. The proposed method, NLTD 2.0, presents a robust and efficient solution to tackle color variations in histopathology images, enhancing the reliability and precision of computational pathology. Additionally, incorporating the method as an ImageJ plugin amplifies accessibility and usability, enabling researchers and pathologists to leverage its capabilities without specialized programming skills. The intuitive interface streamlines the application, fostering broader acceptance within the computational pathology community.