How to implement digital pathology in tissue-based diagnosis (surgical pathology)?

Abstract

Background Digital pathology has proven to become a reliable tool to be applied in pathology education and experimental measurements. However, it is still waiting at the front door of routine surgical pathology or tissue – based diagnosis. Why? Definitions Tissue – based diagnoses includes all diagnostic tools and algorithms that serve or can serve for diagnostic purposes in general. Any medical diagnosis procedure maps a panel of “findings” to a set of discrete diseases that include recommendations for the most efficient treatment. In surgical pathology, the “findings” include data that are derived from a) image content information, b) clinical history, c) expertise of the pathologist, d) knowledge about the disease. These are transferred using a statistical decision algorithm (neural network, discriminate analysis, factor analysis, etc.) into a diagnosis. Image content information (ICI) Several commonly not well trained pathologists assume that microscopic images or any morphology is representative for all kinds of diagnoses. They overestimate ICI which includes the information that can be derived from an image without any external knowledge. In fact, a disease can manifest with different morphologies and vice versa. Therefore, a correct diagnosis has to include all contributions of information, especially when computerized algorithms are used. Electronic communication is a useful tool to collect all available information for diagnostic purposes. Such a system should consider the different clinical impact and the level (details) of a stated diagnosis. It ranges from crude diagnosis such as cancer, infection, etc. to predictive diagnosis or even the evaluation of a disease risk (beast cancer associated genes) prior to its manifestation. Preparations of image content information measurements Any measurement is a communication to transfer image information to interpretation. This procedure has to be standardized by pre - analysis algorithms. They include optical issues (shading, gray value distribution, image size, object size, etc.) and issues of interpretation (regions of interest, object features, structures, and texture. They can be used to control and improve the laboratory’s quality. Conclusions Digital pathology offers numerous tools that can be composed to an effective and appropriate diagnosis system. The tools should collect data from all contributing information sources, standardize the collection and interpretation algorithms and grade the obtained diseases in appropriate disease levels that direct the available treatments. The coordination and standardization of the essential factors seem to be a main constraint of digital pathology implementation into routine pathology.