Medical Software Technology: “Expert Systems”

http://rhapody2000.free.fr/images

I work in the medical software industry, and quite often, people will ask me what is involved with that. In my mind, medical software involves many things, including image scanning, electronic health records, bioinformatics, and much to my surprise artificial intelligence technologies such as “expert systems”! Today, I would like to talk about this hot topic in the medical software industry – “expert systems”.

So what is an expert system? Wikipedia gives the definition that it is “software that attempts to reproduce the performance of one or more human experts.” And also, “an expert system uses a knowledge base (or rulebase) and an inference engine to simulate the reasoning process that a human expert uses to analyze a problem and arrive at a conclusion.” This might be accomplished using “confidences” or certainty factors that are meant to imitate the confidences humans use in reasoning, rather than use the strict probability rules of mathematics.

The architecture behind expert systems is not too complicated. Basically, an expert system just consists of a very large knowledge base (or rulebase), usually consisting of “IF / THEN” type statements, and an inference engine that might operate by using forward or backward chaining logic.

The end-user will usually be required to answer a series of questions, and the large knowledge base will then be queried in order to spit out some sort of conclusion. For example, the conclusion could be a disease diagnosis based on a number of symptoms that the patient has. Or, an expert system could be designed to alert a pharmacist of potentially harmful drug interactions when entering a prescription order.

I wouldn’t be surprised if the idea of these expert systems was met with quite a bit of skepticism. People might wonder how successful these systems really are in practice. In my mind, these systems could be very helpful to doctors and nurses in the decision making process, but should not be used to try and replace these experts.

Here is a list of some common pitfalls:

The systems are prone to making errors that humans would easily spot (i.e. lack of common sense)
The knowledge base has to be constantly updated and maintained to keep it up-to-date
Too many alerts and reminders could overwhelm doctors and nurses, causing the alerts to be ignored altogether
Workflow integration difficulties – will the system slow the physician down?

But here are some benefits: