The non-invasive assessment of brain oxygenation in patients is an important quest in different clinical settings like traumatic brain injuries, stroke, heart surgery, pre-term neonates, rehabilitation. Near-Infrared Spectroscopy (NIRS) can help in probing the brain non-invasively by detecting changes in photon random path caused by oxygen-related changes in the absorption spectrum of hemoglobin. Based on NIRS, different medical grade commercial instruments are widely distributed in hospitals to monitor the level of brain oxygenation and identify critical thresholds requiring interventions. Also, great impetus is set on research to devise new approaches (e.g. frequency-domain or time-domain systems, tomographic schemes) which could help in increasing the accuracy and reliability of brain oxygenation readings.
In our work, we have studied the alterations caused by the inter-subject variability on the optical properties of the head and brain as well as the effect of layered geometry on the diffuse optics estimate of brain oxygenation. The starting point is a multi-laboratory study where we investigate the in-vivo optical properties of the brain in the range 600-1100 nm in 10 adult subjects using both continuous-wave (CW) and time-domain (TD) instruments. This study was performed in collaboration with different research groups and the different techniques were applied to the same set of subjects.
This set of subject specific optical properties was then used to reproduce 10 different models of the brain on which simulations were performed using a solution of the Radiative Transfer Equation under the Diffusion Approximation for a 2-layer geometry. On this basis, we studied the error in the retrieval of brain oxygenation caused by different approaches and experimental set-ups.
The results demonstrated high inaccuracy (up to 15-20%) in the retrieval of the true brain oxygenation using multi-distance CW techniques as a consequence of inter-subject variability in scattering and absorption properties in the head, and also in the masking effect caused by the skull, which leads to a strong reduction in the sensitivity to brain oxygenation changes. The potential improvements in disentangling superficial from deep effects using TD approaches based on picosecond laser pulses and time-correlated single-photon counting detection will also be addressed.