Hybrid intelligence system for diagnosing coronary stenosis. Combining fuzzy generalized operators with decision rules generated by machine learning algorithms

The authors' approach involved taking a set of "crisp" decision rules generated by a machine learning algorithm and converting them into fuzzy rules. These fuzzy rules utilized the previously specified 30 fuzzy sets. Then, fuzzy sets were derived to represent major coronary artery stenosis, using generalized fuzzy operators. Four such fuzzy sets were obtained, to represent obstructions in the three main coronary arteries, LAD, RCA, and CCX, and to represent normal patients. The authors employed the generalized aggregation operations, which are generalized intersection, union and mean. For union and intersection, they used a special class of generalized operators defined by Dombi (1982). Each fuzzy set, representing one of the 30 regions, was defined using a logarithmic scale dividing a range of possible value of a perfusion defect (0-100) into 8 intervals.<>