Analysis of neurosurgery data using statistical and data mining methods

Abstract

The data concerning the outcomes of surgical clipping and endovascular treatment in acute aneurysmal subarachnoid hemorrhage (SAH) patients have been analyzed to reveal relations between subjective neuropsychological assessments, measurable characteristics of the patient and the disease, and the type of treatment the patient had undergone one year before. We build upon results of previous analyses where have been found that the differences in neuropsychological assessment of the patients treated by either coiling or clipping was small and slightly in favor of surgical group. Using this data, we compare the “classical” statistical and data mining approach. While statistics offers techniques based on contingency tables, where the compared variables have to be manually selected, data mining methods like association rules, decision rules or decision trees offer the possibility to generate and evaluate a number of more complex hypotheses about the hidden relationships. We used SAS JMP to perform the statistical analysis and LISp-Miner system for the data mining experiments.