A Data-Driven Approach To Predict Cerebral Aneurysm Ruptures Using Discretized Data

Cerebral aneurysm are deformations of the cerebral vessels characterized by a bulge of the vessel wall. It poses a major clinical threat and upon diagnosis, it is complex, lengthy and costly. There are two methodologies for evaluating and treating patients with cerebral aneurysms. They are computed tomography angiography (CTA) and digital subtraction angiography (DSA) for evaluation of cerebral aneurysms. These methods evaluate the rupture status, and other imaging characteristics that guide surgeon to make appropriate and timely treatment recommendations. The main goal of this capstone current research is the creation of a data mining classifier that supports the use of Computed Tomography Angiography (CTA) to identify a cerebral aneurysm and predict Subarachnoid Hemorrhage (SAH). This study attempted to use 101 patient data with 40 features each. The features include basic demographics, clinical information, and morphological characteristics. These data were collected from the Aneurisk team at Emory University. They have shared the collected data to improve the understanding and the therapy of cerebral aneurysms. Therefore, to discover the data mining classifier, the data is first preprocessed and is then discretized using the R’s discretize function. In this study, decision tree algorithm is used as it is a good fit for this scenario; it selects the best features and provides clear rules. To access the suitability of this technique, the first unprocessed and original continuous data was fed into R’s rpart function from the caret library. To overcome the continuous barrier of the decision tree, a novel approach of using multiple decision trees for feature selection along with Apriori algorithm is used for the best association rules. To access this R’s arules library is used