Fuzzy medical expert system for prediction of prostate cancer

Abstract

We developed the fuzzy medical expert system (F-MES) based on fuzzy inference system (FIS) Mamdani using a different approach to prostate cancer risk (PCR) prediction. The difference in our research is that we modify the membership function on the input variable according to medical standards. We used the same input variables as the previous study, namely age, prostate-specific antigen (PSA), prostate volume (PV), and percentage (%) free PSA (%FPSA). The data on the input variable is used as input into F-MES and displays the output in the form of a percentage (%) of PCR. If the PCR is >50%, then the patient is advised to undergo a biopsy test. We conducted an analysis with the doctor to create a simple domain and rule base of 24 rules. Our number of rules is lower than previous studies of 80 and 240, but our prediction results are better the F-MES evaluation used the same 56 patients, that the F-MES we developed had an accuracy of 857%. This score is better than previous studies of 75% and 76%. Our F-MES is simple but effective and can be used as a supporting tool in decision-making in medical diagnosis.