Infrared spectroscopic imaging for label-free and automated histopathology

A new paradigm in biomedical imaging is emerging in which the intrinsic chemical content of tissue is used to provide contrast in images. The approach utilizes spectroscopic methods to record the chemical information and computational methods to visualize the information. While knowledge extraction from chemical imaging is very powerful in that a single recording of data from unperturbed samples can be related to a variety of pathophysiologic states, the process of knowledge extraction and quantification of confidence in information is not straightforward. The constituent instrumentation, numerical methods, samples and statistics all play inter-related roles in the quality of information obtained. Here, we first present a case study of rapid analysis of breast biopsies to determine cancer, in which the role of sample and data quality is elucidated. We describe next strategies to ensure quality control over extracted information and provide examples of the use of numerical methods for the same. A synergistic control over quality of instrumentation, data and information extraction is shown to speed up decision-making for histopathology by almost two orders of magnitude in time.