ST-ONCODIAG: A semantic rule-base approach to diagnosing breast cancer base on Wisconsin datasets

Abstract Breast cancer is a major terminal disease that occurs largely among females. This disease stems from abnormal mutations in the genes of normal cells, thereby resulting in development of cancerous cells. Though there have being several research breakthroughs in the field of medicine in taming this disease, however, computer aided diagnosis on the other hand has proven very supportive in the quest. Techniques such as Machine Learning (ML) and Medical Expert Systems (MES) algorithms have added impetus to the use of artificial intelligence in detecting and diagnosing breast cancer. While MES may seem promising in machine based diagnostic systems, their accuracy is often impaired by inefficient medical reasoning algorithms employed. This paper therefore seeks to address the limitation of one such reasoning algorithm known as Select and Test (ST). The approach in this paper is to first create an efficient input mechanism that enables the system to read, filter and clean input from datasets. Secondly, semantic web languages (ontologies and rule languages) were used to create a coordinated rule set and a knowledge representation framework was created to aid the reasoning algorithm. As a result, the reasoning structures of ST were modified to accommodate this enhancement. Thereafter, the input generating mechanism was used to transform instances of the databases of Breast Cancer Wisconsin Data set retrieved from UCI Learning Repository. The generated inputs were passed into the improved ST algorithm to diagnose breast cancer in patients captured in the datasets. Experiments were carried out, and result show that 26.60%, 56.17%, and 54.05% were diagnosed of breast cancer in Wisconsin Breast Cancer Database (WBCD), Wisconsin Diagnostic Breast Cancer (WDBC), and Wisconsin Prognostic Breast Cancer (WPBC) respectively.