NIRS has the following limitations [5] -[7] :

1) It detects regional cerebral oxygenation (small area under the optodes).

2) Optical path length is difficult to identify.

3) Arterial versus venous changes are not distinguished.

4) The effect of extracranial tissue on NIRS signal is unknown.

5) Absolute quantification is lacking.

6) Vulnerable to light and movement artifacts.

7) Dissolved oxygen not measured.

Several methods of simulation and modelling were used to overcome such limitations [8] -[11] . Chang et al. [9] used a simulation method for understanding photon migration dynamics in human brain based on Monte Carlo algorithm by using three-dimensional magnetic resonance imaging (MRI) data with multi-wavelength illumina- tion. MRI imaged the human brain with five layers; scalp, skull, cerebrospinal fluid (CSF), gray matter and white matter. They found that such patient-oriented simulation model can provide significant characteristics on the optimal choice of source-detector separation within 3.3 cm for individualized case. Also, significant distor- tions were observed around cerebral cortex folding. Spatial sensitivity penetrated deeper to the brain in the case of expanded CSF. Such finding suggests that the optical method may provide not only functional signal from the brain activation but also structural information of brain atrophy with expanded CSF. It appeared that the three- dimensional time-resolved brain modelling method approached the realistic human brain and provides useful information for NIRS systematic design and calibration for individualized case with prior MRI.

Others [11] described a four-layered for NIR light propagation in a human head based on Mont Carlo method. They used wave lengths of 690 nm, 800 nm and 1300 nm. The four layers are scalp and skull (one layer), CSF, grey matter and white matter. They found that the wave length of 1300 nm is more appropriate for monitoring of human brain activity. Also, grey matter absorption was more sensitive at 1300 nm, and the spatial resolution was improved at 1300 nm.

It appears that NIRS is a promising brain oxygenation monitor. It has all the theoretical advantages that could make it a suitable device for real-time brain monitoring during cardiac surgery. It is to be hoped that increasing- ly sophisticated computer modelling will permit a better understanding of its limitations and that modern equipment will make possible the real-time correction for the scattering effects of extracerebral tissue.