AI Revolution in Histopathology and Cancer Research

The history of pathology starts from 1850 when German doctor Rudolf Virchow first recognized that a cell is a basic entity of tissues and organs. Since then, our view on the role of cells and the methods we use to study them changed drastically, with automated technology being widely used in research workflows.

Over the past 10 years we moved from digital pathology to next-gen digital pathology, where not only imaging of tissue slides was largely automated but also various image analysis software solutions were introduced and adopted into laboratory operations. This transition calls for robust and reliable technology for improving data quality and diagnostic precision. Although automation in image acquisition and image analysis will not mitigate the lack of pathologists and restricted access to healthcare systems in resource deprived areas, it will surely be a substantial help for pathology specialists, researchers and patients.

Particularly, as we find ourselves in the rise of an artificial intelligence (AI) era, the possibilities of AI use in biological and medical research seem to be endless. AI can be used for various tasks in cell and tissue analysis, such as identifying patterns and morphological characteristics of nuclei, allowing for precise segmentation even in dense and heterogeneous tissue samples. Beyond cell segmentation, AI can assist in tissue detection, identifying and classifying different tissue types within a sample. These and other modalities are often included in digital microscopy systems, offering automated workflows for pathology and research labs. Regulatory organs such as the US Food and Drug Administration (FDA) already opened new ways for AI-based decision support systems for diagnosis of cancer-related diseases.

With cancer being a complex multifactorial disease, tumor biology and cancer research could potentially benefit most from what AI can contribute to improved diagnosis, treatment and monitoring of the disease. Immunohistochemistry staining is used as one of the main methods to study tumor cells in their native tissue microenvironment. Combined with whole slide imaging, this technique delivers valuable data on marker distribution in human tissues. A large number of studies employed innovative machine learning-based segmentation and data-mining functionality to align the data with clinicopathological parameters.