Keith Cheng

Annotations

1. Computational 3D histological phenotyping of whole zebrafish by X-ray histotomography

   To our knowledge, no existing method has the combination of throughput, resolution, field-of-view, and soft-tissue contrast necessary for whole-organism 3D phenotypic screens that are inclusive of histopathological evaluation.

   The origin and uniqueness of histotomography were persistent questions during its development.

   What we did not include in the paper is the multiple motivations underlying this work. A primary motivation was provided by medicine, which is, at its root, phenotype and causation-oriented, and since the 1800s, has been oriented towards complete phenotyping - which has not been a strength of reductionist science. Other motivations came from bioinformatics, genomics, and systems biology, which are strikingly molecule oriented, and weak on phenotype. Bioinformatic assignment of gene function based on incomplete phenotyping has been extremely misleading. Solving this problem requires us to bring the power of histopathology to bear, but in a way that allows objectivity and quantitative rigor. This would require a new ability that did not exist: volumetric and cell population measurements. A 3-dimensional histology was needed, but did not exist.

   Early work with commercial sources led to a realization that I needed to turn to some form of synchrotron microCT. This paper is a summation of about a decade of collaborative work since that time. My broader hope has been for this new technology to enable a phenotype-oriented, "computational phenomics" that, through the development of more phenotype-oriented bioinformatic resources, can serve as a foundation for its stated goals.

   We have been frequently asked, in various ways: "Other microCT studies have have nominally achieved sufficient resolution. Why wasn't this form of imaging developed earlier?"

   Answering these questions requires a consideration of what characteristics of histology were required to make possible the scientific resolution we call the cell theory: 1) sub-micron resolution, 2) centimeter scale field-of-view, and 3) soft tissue contrast sufficient to visually distinguish cell types and extracellular materials from each other. These requirements were all met by development of the compound microscope.

   The same combination of requirements is required for 3-dimensional histology, which had not been achieved in prior microCT studies. A practical answer, however, has to do with the test applied throughout the development of x-ray histotomography: Can the histological features of normal and abnormal tissues be discerned from the images? Until all the modifications reported here were applied to microCT, the answer has been, "No." Thus, histopathology played a key role in the development of x-ray histotomography.

   Additional features needed to make 3-dimensional histology practical included mechanisms of 3D visualization and demonstration of 3D computational histological phenotyping. Fulfilling these additional requirements has required the emergence of only recently achieved degrees of computational power.

2. Computational 3D histological phenotyping of whole zebrafish by X-ray histotomography

   The images demonstrate the near histological resolution of X-ray histotomography

   Note that using multi-slice "slabs" close to 5 microns can facilitate the transition from using histology sections to histotomography. These slabs do not, however, take advantage of the superior z-axis (slice) resolution of X-ray histotomography.

3. Computational 3D histological phenotyping of whole zebrafish by X-ray histotomography

   f

   Typo. The sentence should read, "Plastic embedding of specimens, while not essential for synchrotron micro-CT, adds sample stability (Lin et al., 2018), to allow image re-acquisition with improved histotomography implementations."

4. Computational 3D histological phenotyping of whole zebrafish by X-ray histotomography

   For bandwidth considerations, only every fourth slice is currently shown in the viewer by default.

   We anticipate that every slice in orthogonal slice sets will become practical with with implementation of 5G and faster networks.

5. Computational 3D histological phenotyping of whole zebrafish by X-ray histotomography

   The gill structure is visualized in a whole juvenile (33 dpf) zebrafish in order to demonstrate the ability of soft tissue synchrotron micro-CT to resolve the complex structure of 3D tissues in detail

   This movie was especially striking to the authors. The "bumps" in the surface of the epithelial cells of the gills and gill cavity used by their nuclei were striking, but may be dependent on cytoplasmic shrinkage associated with fixation.

6. Computational 3D histological phenotyping of whole zebrafish by X-ray histotomography

   This work was inspired by both the power and limitations of pathology - a foundation of modern medicine. In pathology, microscopy is routinely used to distinguish between disease mechanisms affecting animals, including humans, based on its ability to tell us about every cell type and extracellular material comprising a tissue. Notably, micron-scale changes historiography was designed for, are found in virtually every human disease. Histology is unlike any imaging method (such as fluorescence) that depends on staining of a single molecule. Driven by the potential power of large-scale phenotypic analysis of the effects of genetic and chemical manipulations of animal models ("phenome" projects), our vision was to convert pathology's 2-dimensional, qualitative, and largely subjective analysis to a 3-dimensional one that is potentially inclusive all cellular and extracellular elements, and just as critically, quantitative, reproducible, and accessible to computation. Otherwise stated, we wanted to open the power of pathology to the masses by turning morphology into math.

   The desire to "slice" the tissue in any direction without distortion requires 3-dimensional pixels, or voxels, to have equal dimensions in all directions ("isotropic"). Isotropy makes undistorted visualizations of histotomographic data, for medicine, science, education, and fun, possible.

   The derivation of "histotomography" comes from "histo-" for tissue, and "tomo-" for "cut". "Cutting" in histotomography is purely digital, because the tissue remains intact even after imaging is complete. Histotomography includes "toto", which emphasizes the detection of every cell type and structure in entire (small) specimens.

   Finally, it is well-worth taking pause to note that this work was never before possible in human history, due to its dependence upon 1) infrastructures only possible when governments support scientific education, 2) the availability of merit-based use of a U.S. Department of Energy synchrotron, 3) NIH funding of interdisciplinary projects dedicated to resource development, and 4) the massive, recent increase in computational power, including GPU processing and rapid internet-based communication.