Automatic Evaluation of Mannitol Dehydration Treatments on Controlling Intracranial Pressure Using Electrical Impedance Tomography

Intracranial hypertension can potentially cause ischemia, herniation, and even death. Mannitol dehydration treatment is a common method for reducing increased intracranial pressure (ICP). However, the efficacy varies among patients and improper use of mannitol may cause severe side effects. This study aims to find ways to individually evaluate and provide guidance for mannitol dehydration treatments based on dynamic brain electrical impedance tomography (EIT). First, clustering methods were applied to analyze the relationship between EIT data and patients' different responses to mannitol. With the clustering results as well as empirical experiences, doctors could establish a rule for classifying the efficacy of mannitol dehydration treatments. Then, to rapidly categorize following EIT datasets, an automatic classifier was developed using the multiclass support vector machines (SVMs). Additionally, to enhance the accuracy of the classifier, a novel priori and principle components (PPCs) based index was introduced as the input data for training the classifier. The experimental results demonstrated that different clustering methods yielded similar results and the optimal number of clusters was three, which corresponded to three types of efficacy of mannitol. For the automatic classifier, its classification accuracy improved to 93.3% with the PPCs index. As such, it can be concluded that brain EIT promises to be an effective tool for individually evaluating the efficacy of the dehydration treatments and make guidance for rational use of mannitol.