Risk profiling patients with left atrial appendage closure using the K-Modes algorithm

Abstract

The high incidence of stroke and bleeding in the elderly is increasingly recognised, and LAAC is one of the best treatment options to prevent atrial fibrillation. A growing number of cardiac research centers have established pre-, intra-, and postoperative follow-up databases of patients undergoing LAAC procedures. For patients, a simple division into bleeding and non-bleeding groups and stroke and non-stroke groups is performed to assess postoperative effectiveness.

To obtain valid and accurate results for data analysis, this thesis is based on data from Turku University Hospitals. The data of 153 patients were clustered and analysed using Python language and the k-modes clustering algorithm. The results were displayed graphically. In addition, the number of clusters was chosen by evaluating the clustering effect by the Silhouette score.

The final result is a more accurate clustering analysis by the k-modes clustering algorithm to evaluate risk groups from a complexity perspective.