Longitudinal Sparse Regression for Neuroimage Based Consciousness Assessing and Tracking of Hydrocephalus Patients

Hydrocephalus is a condition that causes ventricle enlargement by pathological accumulation of cerebrospinal fluid (CSF) and ends with different levels of disorders of consciousness (DOCs). Assessment of the consciousness level will help for the planning of the treatment of the patients. In this paper, a longitudinal sparse regression model featured by the temporal constraint is proposed to assess the levels of consciousness for hydrocephalus patients and to track their temporal alterations based on magnetic resonance (MR) images. Specifically, for the time points before and after neurosurgeries, we extract features from the corresponding MR scans and then regress out the clinical scores that reflect the respective consciousness levels. The longitudinal regression model can thus be applied to automatically track and evaluate the consciousness level change for individual patients, while the reading of the regression can act as an important indicator for the planning of subsequent treatment in clinical practice.