ConceptualEyes builds on shoulders of giants that brought artificial intelligence to medicine (Part 1: 1950s – 1980s)

August 1, 2017 at 4:14 am  | 

The fascination of applying AI to medicine dates back to the late 1950s [1] followed by several academic publications documenting collaborations of doctors and computer scientists [2]. The philosophical foundation with both disciplines being the process of collecting data and applying inference rules to make a predictive diagnosis of a disease. Ledley and Lusted [1] pointed out that medical reasoning was not magic but instead contained well-recognized inference strategies: Boolean logic, symbolic inference, and Bayesian probability. Several tools such as PROMIS [3], CASNET [4], MYCIN [5], QMR [6], INTERNIST [7], DXPLAIN [8] and ILIAD [9] have become available since then. The performance of these programs is evaluated, validated and compared by running them on some challenging case reports (called clinic-pathological cases, or CPCs) such as those that appear each week in the New England Journal of Medicine. The performance analyses of these tools routinely outperformed medical students in training to be physicians. However, due to the lack of digital interoperability and standards in representing health records and the exploding nature of medical research, the computer-based expert systems were unable to sustain the momentum. Also, expert systems that demonstrated the ability to get better at diagnosing typical/common cases were unable to handle rare or mysterious illnesses. The effort to normalize and archive medical knowledge with interoperable standard terminologies led to the Unified Medical Language System (UMLS), a project at the National Library of Medicine with the goal of integrating a number of existing medical vocabularies using a common semantic structure [10] and Semantic Medline, a semantic database of biomedical research [11].

Part 2 of this blog series will begin with UMLS and MEDLINE as the foundation.

< Back to Blog