Skip to main content

Wavelength optimized real-time feedback of brain activation by principal component analysis on near-infrared spectroscopy data

Wednesday, October 26, 2011 — Poster Session III

10:00 a.m. – Noon

Natcher Conference Center




  • J Kainerstorfer
  • I Styles
  • H Dehghani
  • L Najafizadeh
  • F Amyot
  • J Riley
  • P Smith
  • A Medvedev
  • A Gandjbakhche


Rehabilitation of traumatic brain injury (TBI) can be difficult to assess at the bedside, since real-time feedback of activation is desired. Near-Infrared Spectroscopy (NIRS) can measure brain functions in means of total (HbT), oxygenated (HbO), and deoxygenated (Hb) hemoglobin, but calculations of those values are model based, hence time consuming. We are proposing Principal Component Analysis (PCA) for extracting these values quantitatively, model independent and in real time. We have previously shown in skin data that eigenvector 1 corresponds to HbT and eigenvector 2 to HbO. Our data also suggest that there is an underlying mechanism, which allows extraction of HbT and HbO specific eigenvectors, which are wavelength dependent, but tissue independent. We also performed wavelength optimization and were able to find a finite set, which reliably separate HbT and HbO in eigenvector 1 and 2. In-vivo data was acquired using NIRS with the imaging probe placed over the frontal cortex. Applying PCA, we were able to obtain HbT and HbO values, demonstrating that PCA can be used for obtaining hemodynamic responses in real-time. Thus, PCA on NIRS data becomes a powerful tool for real-time feedback of brain responses, which will have great impact on assessing rehabilitation of TBI patients.

back to top