Anatomic Pathology Information Laboratory Information Systems and Natural Language Processing: Early History

Abstract

Anatomic Pathology laboratories began to encode and retrieve diagnoses from paper reports using punched cards in the 1960s. When Anatomic Pathology Laboratory Information systems (APLIS) began to be used in the 1970s, pathology departments found that searching free text for diagnoses was hindered by the variability of terms for the same concept and by the amount of computer resources required. Consequently, system developers began to engineer natural language processing (NLP) algorithms to translate computationally ambiguous written language into coded concepts. At the same time, pathologists began to develop the Systemized Nomenclature of Pathologists (SNOP), which eventually developed into the internationally utilized ontology called the Systematized Nomenclature of Medicine Clinical Terms (SNOMED-CT). Systems in the 1970s used keywords based on concepts to encode reports and the development of query languages to retrieve them. By the 1980s, the user was guided through decision trees to the best matching concept code from data dictionaries that linked pathology systems with medical record systems, and linguistic analysis that converted natural language into a structured database. At that time, some advanced NLP systems had knowledge bases that indicated how experts would interpret the meaning of words in particular contexts, but most query languages were still based on the search and retrieval of encoded terms. Much has occurred since then, including the recognition of Pathology Informatics as a discipline and the use of many NLP algorithms as well as algorithms to scrub identifi ers from free text reports for research purposes.