Abstract

In this paper fuzzy arithmetic is used to develop a prediction model for the first site of breast cancer metastasis. Membership functions for the overly expressed protein biomarkers are developed. Three first sites of metastasis i.e., bone, liver and lung are chosen. A fuzzy relation between these metastases and the protein biomarkers is computed. A neural network is trained with these fuzzy relations to predict the first site of metastasis. The trained network is tested for different combination of proteins for the prediction of metastasis. These fuzzy relations were defuzzified using Lambda cut sets to predict the first site of metastasis. These results were compared to that obtained with the neuro-fuzzy model. The different protein biomarkers chosen were COX2, HER2, ER, PgR, CK5, CK18, GATA3 and E-cadherin.

Authors and Affiliations