Since the inception of the microscope, researchers and clinicians have characterized tissues qualitatively through the process of embedding, sectioning, staining and evaluation with a microscope. While this practice has been the gold standard for diagnosis and addressing many research questions, this human-driven approach is highly limited in throughput, qualitative, and subjective. Many studies have shown significant inter- and intra-pathologist variability using a subjective characterization approach.
To help address the limitations of human-based slide characterization, the field of digital pathology has evolved wherein histology slides are analyzed by algorithms that can quantitatively assess staining, morphological features, and cell counts. This computational approach provides high throughput, excellent reproducibility, and quantitative metrics (e.g. total cell counts) for evaluation of histological sections. Furthermore, utilizing machine learning techniques, histological sections can be classified based on morphological or staining characteristics, with the use of training data (supervised machine learning) or by unsupervised machine learning to classify tissue sections into categories based on similarities and differences.