Intelligent Fuzzy Logic Models for Predicting Post-Operative Cardiac Complications

Cardiac complications are one of the biggest concerns after a patient has undergone cardiac surgery.It is therefore important to have predictive tools or models that can give both a reliable estimate of the patient's overall risk as well as a measure of how reliably this model predicts their individual level of risk.The purpose of this research was to develop an evidence-conserving selective fuzzy inference system (EC-SFIS) in order to predict the occurrence of serious cardiac complications postoperatively and postoperative atrial fibrillation from clinical input values including age, left ventricular ejection fraction (LVEF), serum creatinine, and duration of cardiopulmonary bypass (CPB).Unlike conventional fuzzy inference, the proposed framework decomposes rule-based information into supporting, opposing, and unresolved evidence and introduces a conflict-conservation operator that transfers contradictory evidence into an unresolved component while preserving the directional evidence balance Evidence-ratio defuzzification, probability calibration, and a selective prediction mechanism are subsequently employed to accept predictions with adequate evidence resolution and low conflict while flagging uncertain cases for additional assessment.A numerical case study demonstrates that the framework can distinguish between predictions having similar evidence resolution but different conflict levels, resulting in outcome-specific acceptance or flagging decisions.The comparative evaluation presented in the study indicates lower prediction error across risk-coverage levels and favorable probability calibration relative to the considered fuzzy and statistical comparators.The proposed EC-SFIS therefore provides an interpretable mathematical framework in which postoperative risk prediction is accompanied by explicit quantification of evidence consistency and prediction reliability.