Fourier space analysis of tissue order

Tissue architecture is one of the paramount aspects of diagnostic pathology and is, as such, of great importance in development of digital pathology. Inflammation can result in tissue damage, changing the tissue organization. To evaluate the disease progression and severity, the extent and intensity of these changes must be determined.

We present an approach to determine the overall order of tissue architecture in histology slides of murine abdominal wall samples with induced peritonitis. Hyperspectral microscopy between 450 nm and 750 nm is used to acquire an image of a standard H&E slide. Using the law of exponential attenuation, the image is decomposed into maps of both stains. A Fourier transformation is applied to the eosin image of cells, obtaining a distribution of spatial frequencies at different directions in the tissue. Where tissue architecture was impaired, the distribution was independent of the angle, whereas in healthy, well-organized tissues, a dominant spatial frequency direction was observed. The method was used for classification of subjects into healthy and diseased, achieving a p-value of 0.02.

The presented approach could serve as a basis for future development of quantitative metrices in digital pathology, aiding in automated, computer-assisted assessment, facilitating larger screening and longitudinal studies.