Extracting fuzzy rules for the diagnosis of breast cancer


Abstract: About one million women are diagnosed with breast cancer every year. Breast cancer makes up one-third of all cancer diagnoses in women. Diagnosing breast cancer early is vital for successful treatment. Among the breast cancer screening methods available today, mammography is the most effective, although the low precision rate of breast biopsy caused by mammogram interpretation results in approximately 70% unnecessary biopsies with benign outcomes. The aim of this study was to extract strong diagnostic fuzzy rules for the inference engine of an expert system to be used for the diagnosis of breast cancer. These rules have been extracted through the use of artificial intelligence technologies. For this, a neuro-fuzzy classification tool called NEFCLASS was used. The learning algorithm of this tool is heuristic and it has efficient performance diagnosis and classification tasks. The rule base to be used for diagnosis consists of 9 rules using the Breast Imaging Reporting and Data System (BI-RADS), mass shape, and mass margin attributes. The positive predictive value of this rule base is 75% and the negative predictive value is 93%. When the approximately 70% rate of unnecessary biopsy in the diagnosis process is taken into consideration, an expert system that has this strong rule base with a high predictive value can be used by doctors in deciding whether to conduct biopsies.

Keywords: Neuro-fuzzy system, extracting rule, medical diagnosis, breast cancer, mammography

Full Text: PDF