Intelligence demonstrated by machines is defined as artificial intelligence (AI). Expert systems are an application of AI. They interpret the meaning of information stored in data bases, by adopting problem solving and decision making strategies that would typically be undertaken by humans with expertise in the same field. The challenge of using artificial intelligence in medicine is that traditional computer modelling is based on numerical constructs, whereas medical diagnosis and management is generally a qualitative process of reasoning and observation using “if-then” logic.
Possibly the first expert system ever developed was a computer program written in a language called “LISP”, popular for its ease of coding. The system called “DENDRAL”, (combining “dendritic” and “algorithm”) was written under the direction of Edward Feigenbaum at Stanford University in the mid 1960’s. It’s purpose was to test whether a set of “if-then” rules obtained from extensive interviews with industry experts (now called “knowledge engineering”) and then written into a rules engine, could be used to test scientific hypotheses. This expert system would be deemed a success if its accuracy was at least comparable to that obtained by using human experts for the same purpose. The researchers asked a group of organic chemists to test the expert system. They concluded that the program could successfully identify unknown chemical structures by analysing their mass spectrometry data, but that the outcomes were only as good as the input rules. MYCIN, was a second expert system written by the same research group. It was used to identify blood borne infections, and assign appropriate antibiosis. A second objective was to diagnose blood clotting disease. It too was a success. However, it was only ever used as a research tool.
There are now many applications of AI. Google startups and other tech giants are making significant in-roads into AI and health. Machine learning systems use algorithms to learn, without the need for additional instructions. Brain IQ used neural networks and machine learning to develop treatment plans for non-ambulant patients. Cytovale has developed early detection tools for sepsis, whilst Augmedix and Google Glass combine to interpret doctors’ clinical notes during a consultation, then transcribing them to reduce consultation times.
Medical literature is being produced at a phenomenal rate. In cardiology research alone, thousands of new publications are published each year. The potential for this wealth of knowledge to be added to expert systems is tantalizing. However, the evolution of technology also raises ethical concerns. Possibly the most important of these is whether a machine or a human should ultimately be responsible for clinical decisions in medicine. A second issue is whether machines are capable of incorporating human empathy in their decision criteria. It is a certainty that AI in medicine will continue to evolve and will increasingly find its way into medical practice. However, ethical considerations may dictate regulations in relation to some aspects of its use.
Incarta’s Altegix is artificial intelligence software which helps to identify hospitalized patients at risk of life threatening illness before it occurs. Importantly, it does not make any decisions on behalf of the doctor, but acts as an early alert system to a potential problem. Even so, the software has the ability to save lives. Altegix captures and interprets key clinical pathology data, analyses it, and creates alerts based on its findings.