The medical conditions of humans are usually an important indicator which are monitored by human doctors. Since the boom of expert systems in the 1970s some attempts were made to use computer software for diagnosing health. A normal expert system is working with rules which contains of a bayesian component. Such a system is working precise and is able to support the normal doctor. But some expert systems are realized with Fuzzy logic in mind. For example, for detecting Coronary Artery Disease. The problem with these systems is, that the number of input variables is very high, and as a result the amount of needed rules growths exponential. In the example, 10 input variables were used which results into 10x10 rules. It's obvious, that such amount of rules can't be created from scratch and as a result the system makes a lot of trouble. It's an example for using computer technology the wrong way.
But let us take a look into the paper itself. On page 5 there is the table with the input variables. If the patient is typing in his real age which is 52 the expert system doesn't know which category he is. He will become two of them which is old and very old. This kind of wrong assumption results into a low quality rule inference. If the software isn't sure in which age category the person is located, how can the resulting output become correct? The figure on the same page from the Matlab toolbox isn't improving the confidence into the system very much, because on the screen a large amount of graphics is given and nobody knows what the meaning is.
A similar technique was implemented for detecting Diabetes Mellitus in a web-based fuzzy expert system. The user has to enter first a lot of features (10 exactly) and then a single output is calculated. The rule base in the paper is hard to read and it's not possible to reproduce the results. The system isn't made online available and perhaps this is an advantage, because the system accuracy is low.
- Kasbe, Tanmay, and Ravi Singh Pippal. "Enhancement in Diagnosis of Coronary Artery Disease using Fuzzy Expert System." (2018).
- Mujawar, I. K., B. T. Jadhav, and Kapil Patil. "Web-based Fuzzy Expert System for Symptomatic Risk Assessment of Diabetes Mellitus." International Journal of Computer Applications 975 (2018): 8887.