Breast Cancer Risk Assessment Using adaptive neuro-fuzzy inference system (ANFIS) and Subtractive Clustering Algorithm

Introduction: The adaptive neuro-fuzzy inference system (ANFIS) is a soft computing model based on neural network precision and fuzzy decision-making advantages, which can highly facilitate diagnostic modeling. In this study we used this model in breast cancer detection. Methodology: A set of 1,508 records on cancerous and non-cancerous participant’s risk factors was used.  First, the risk factors were classified into three priorities according to their importance level, were fuzzified and the subtractive clustering method was employed for inputting them with the same order. Randomly, the dataset was divided into two groups of 70 and 30 percent of the total records, and used for training and testing the new model respectively. After the training, the system was separately tested with the Wisconsin and real Clinic's data, and the results were reported. Result: The desired fuzzy functions were defined for the variables, and the model was trained with the combined dataset. The testing was then conducted first with 30 percent of that dataset, then with the real data obtained from a real Clinic (BCRC) data, while the model's precision for the above stages was 81(sensivity=85.1%, specifity=74.5%) and 84.5 percent (sensivity=89.3%, specifity=79.9%) respectively. Conclusion: A final ANFIS model was developed and tested for two standard and real datasets on breast cancer. The resulting model could be employed with high precision for the BCRC Clinic's database, as well as conducting similar studies and re-evaluating other databases.