A semantic web rule and ontologies based architecture for diagnosing breast cancer using select and test algorithm

Background and Objective: Breast cancer is widely known as the most lethal and chronic disease among women. Computational techniques have been applied to support and aid early detection of the disease. Methods such as image analysis, knowledgebase systems, machine learning, reasoning algorithm and natural language processing techniques. Approaches aimed at leveraging the availability of patient electronic records have been investigated to support formalisation of medical reasoning thereby achieving early detection of breast cancer. However, computational performance of these approaches is often limited largely due to formalism used for representing the records. We consider that a body of information silo in patient record systems, when accurately formalized, could be efficiently used for improving the performance of computational diagnostic systems which detects breast cancer. The objective of this study is the use of an accurate method for representation and formalisation of text-based patient records and domain knowledge to provide support to medical reasoning algorithm aimed at diagnosing breast cancer. Methods: We propose an architecture which supports formalism of knowledgebase that is applied to reasoning algorithm used for diagnosing breast cancer. The method applied allow for the use of ontologies for the formalisation of both patient records and domain knowledge. Diagnostic procedure and guidelines in the domain were represented using rules based on semantic web rule language (SWRL). Furthermore, we applied the formalized ontologies and rules to Select and Test (ST) medical reasoning algorithm. Results: Experimentation was done using records of ten (10) patient collected from University Teaching Hospital. Result obtained showed that our proposed system achieved accuracy gain of 23.5% and AUC of (0.49, 1.0). Conclusion: The impressive performance of the proposed architecture demonstrates the effectiveness of using rules and ontologies for knowledge representation. In addition, we found the interesting performance of the applied ST algorithm as a pointer that it is a potential algorithm in modeling computer aided diagnostic systems (CADs) for detection of breast cancer.