Noninvasive intracranial hypertension diagnosis using ensemble sparse classifiers

  • Authors:

  • Sunghan Kim;Fabien Scalzo;Xiao Hu

  • Affiliations:

  • University of California at Los Angeles, Los Angeles, CA, USA;University of California at Los Angeles, Los Angeles, CA, USA;University of California at Los Angeles, Los Angeles, CA, USA

  • Venue:
  • Proceedings of the 2011 workshop on Data mining for medicine and healthcare
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

The level of intracranial pressure (ICP), which is the pressure inside the skull, is a valuable indicator for the diagnosis and management of various neurological disorders such as traumatic brain injury (TBI). The current gold standard of ICP measurement involves an invasive procedure for insertion of a catheter to the intracranial compartment. Due to its risks and cost, however, various noninvasive ICP assessment techniques have been proposed. So far, none of them has achieved sufficient accuracy to be adopted into routine clinical practice. We propose a novel classification technique to address a simpler problem, which is the diagnosis of intracranial hypertension (IH). Our goal is to diagnose IH of head-injured patients based on a full set of morphological features extracted from cerebral blood flow velocity (CBFV) waveforms. Our result demonstrates that the proposed technique is a better diagnostic tool for IH than a fine-tuned conventional classification technique such as Support Vector Machines (SVM).