Order models for optical microscopy of biological tissues

Tissue order and architecture are important aspects of determining different types, stages, and malignancy of pathologies. Edema and inflammation, among others, can change the architecture of tissues by changing ordering of the cells, their shape, size, and their density. Determination and quantization of these differences is an important aspect of pathology, that is partially addressed by digital histology. The latter usually relies on geometrical properties of cells and different statistical measures of order, which offer no deeper insight into the changes. In physics, the problem of sample order and architecture is a well-known topic, since ordering withing the samples dictate their physical properties. In solid state physics, crystal latices are used to describe the ordering of atoms, while orientation of long molecules within liquid crystals is described using director fields in soft matter physics. Additionally, transformation from real space into the Fourier space can offer deeper insight into the structure and is a commonly used approach across wide range of physics fields. In this project we propose the development of analogous concepts in the field of digital histology. We propose the study of tissue architecture and order through Fourier space analysis, where orientations and characteristic length scales can easily be determined. Furthermore, we propose the development of algorithms that can calculate constructs analogous to the crystal lattice and director fields. Such constructs will be then used to determine properties of tissue such as its regularity, density, and orientation. Additionally, presence of abnormal inclusion in the tissue can manifest themselves as localized deformations of the tissue lattice or director field thus presenting a specific type of defects withing the normal lattice or director field of healthy tissue. In the first stage of the project, polarimetric measurements will be integrated into hyperspectral imaging microscope to enable acquisition of multi-modal images. Next, images of different pathologies exhibiting different order and architecture will be studied using classical white light microscopy, hyperspectral and polarization sensitive microscopy. Images will be pre-processed, properties from hyperspectral images extracted using physical models and principal component analysis, coregistrated to same areas, annotated with types of tissue and pathologies and assembled into a database for further processing. Further, algorithm for evaluation of tissue lattice and director field will be developed and tested. Based on the tissue lattice and director field, metrices of order and architecture will be proposed, developed, and tested. Additionally, images will be studied in the Fourier space to determine metrices such as spectral entropy and to evaluate tissue orientation and characteristic length scales. Finally, all the proposed methods and metrices will be compared and evaluated for their statistical power as well as interpreted in terms of observable changes in the architecture of the samples. In summary, this project will strive to translate physics-based approaches into biomedical sciences. In the process, new algorithms for calculation and evaluation of order will be developed and tested. New methods will follow guidelines for quantitative imaging biomarkers thus enabling translation into clinical practice.