Histopathology is the science of identifying cell types and disease states through examination of tissue at the microscale. The practice of histopathology, comprising tissue staining and recognition of morphological patterns, has remained largely unchanged for approx. 150 years. Staining is the first essential component in this process as tissue has little intrinsic contrast in brightfield microscopy (Figure 1, left). Hence, dyes are added to identify protein-rich and nucleic acid-rich regions of the tissue (Figure 1, right).
Figure 1. Brightfield optical microscopy of unstained tissue (left) is not very informative. Tissues stained with hematoxylin and eosin (right) allow a trained observer to recognize structures through sub-cellular color contrast.
The second essential ingredient in histopathologic practice is the use of a trained human observer to recognize patterns of cell. For example, a grading scheme for Prostate cancer requires a pathologist to identify morphologic patterns (figure 2a) in epithelial cell ensembles around ducts (termed lumen). Careful cataloging of various morphologies and their historical relationship with patient outcomes has led to a classification of the severity/extent of disease. For example, prostate cancer is diagnosed and graded into five classes by recognizing subtle differences in tissue morphology.
Although integral to clinical and research activities, histopathologic recognition remains a time-consuming, subjective process to which only limited statistical confidence can be assigned because of inherent variability between humans, quantization of a spectrum of disease and imperfect correlation of morphologic grading with patient outcome. By combining molecular information in spectroscopy and structural information inherent in microscopy, our goals are to provide practical methods that:
- Determine all cell types in tissue in an automated and objective manner
- Detect disease (especially cancer) accurately, early and with a defined degree of confidence
- Provide clinical input that will allow for more effective detection and treatment of human cancers
With the development of fast infrared imaging, high throughput sampling and advanced numerical processing, we have recently employed FTIR spectroscopic imaging to provide the necessary rigor in demonstrating validity in clinical measurements. [1] Vibrational spectroscopic approaches directly provide molecular descriptors, but a practical spectroscopic protocol for histopathology is lacking. The automated histologic segmentation was applied to routine archival tissue samples, incorporated well-defined tests of statistical significance and eliminated any requirement for dyes or molecular probes. The result of the developed protocol is a tissue image that is similar to those normally encountered in clinical practice. The end result of this process, however, is an image that is significantly more informative, providing both structural distribution and objective recognition (Figure 2).
Figure 2. Conventional staining (above) and objective histologic recognition (below), in which cells are identified by computer generated “stains” based on their spectral information.
[1] For further details see Nature Biotechnology 23, 469-474 (2005) . For a copy of the paper click here. (doi:10.1038/nbt1080)
