Evaluating and improving the quality of time-dependent, diffuse reflectance spectroscopic signals measured from in vivo brain during craniotomy

Abstract

BackgroundOptical spectroscopy can be used to assess the pathophysiological characteristics of diseased and injured biological tissue in vivo in a non-destructive way. It is often used in conjunction with a contact optical probe for the purposes of operating and sensing in a sterile field. Since the probe is often held by the hand of an investigator during data acquisition, any hand instability can affect the quality of acquired data and, hence, degrade the accuracy of diagnosis. This study was designed to quantitatively characterize these artifacts, and then propose an effective engineering solution to remove them. MethodsTime-dependent diffuse reflectance spectra (Rd(λ,t)) were acquired from the normal cortex region of pediatric patients undergoing epilepsy surgery. They were acquired at a rate of 33Hz, and their range was 400 and 900nm. Two distinct ways of collecting data were tested: one with the fiber optical probe held by the surgeon's hand during data acquisition, and the other with the probe held by a specially designed probe holder. The probe holder was designed and constructed to minimize the variations in probe contact pressure and contact point for the full duration of any given investigation. Spectral data acquired using versus not using the probe holder were characterized and compared in the time, wavelength, and frequency domains, using both descriptive and inferential statistics. ResultsHand motion manifested as strong random variations in Rd(λ,t) which impacted temporal and frequency characteristics of Rd(λ,t). The percentage standard deviation %STD of Rd(λ,t) acquired without probe holder could be as high as 60%, and they are significantly higher than those with probe holder at all wavelengths. This difference is especially prominent between 400 and 600nm. Rd(λ,t) acquired without the probe holder also processed a higher spectral power energy in the frequency domain than those with the probe holder. The correlation analysis revealed that the hand motions induced synchronistic variations in Rd(λ,t) between 600 and 800nm, but this synchronicity is not obvious between 400 and 600nm. ConclusionThe results of this investigation demonstrate the nature and the magnitude of hand motion induced artifacts in in vivo diffuse reflectance spectra and propose one potential solution (i.e., a probe holder) to remove them. These findings allow us to improve the quality of time-dependent, diffuse reflectance signals acquired to study the dynamic characteristics of biological tissues, like brain, in vivo.