Process Trace Classification for Stroke Management Quality Assessment

Abstract

Stroke is a medical condition where poor blood flow to the brain may result in cell damage, possibly leading to patient’s death or disability. Acute stroke care is best performed in dedicated and well-organized centers. Medical process trace classification can support stroke management quality assessment, since it allows to verify whether better-equipped Stroke Centers actually implement more complete processes, suitable to manage complex patients as well. In our previous work, we developed a semantic similarity metric able to compare process traces. In this paper, we adopt such a metric to perform k-Nearest Neighbour (k-NN) classification in the field of stroke management; moreover, we present an alternative classification approach based on deep learning techniques. Experimental results have shown the feasibility of deep learning classification for stroke management quality assessment, which performed better than the application of the semantic similarity metric. Improvements and future research in this direction will therefore be considered. Difficulties in classifying patients treated in less-equipped hospitals also suggest to identify and manage possible organizational problems.