The microsampling approach to spectroscopy applied here utilizes the local confinement of the signal to reduce the need for complex light transport modeling, which is often needed in deeper tissue signals. The confocal optical system confines the signal to approximately 100 μm laterally and to approximately a few hundred microns in the depth coordinate. The fitting used in equation 3
, while empirical, has been found to be robust and a useful way to quantify scatter power from the remitted signal. The scatter power is independent of most coupling errors and uniquely associated with features related to the effective scatterer sizes 36-38
. A key feature in the design of this instrument has been to have a standard reference and acquire the spectrum in a referenced manner, so that measurements between samples could be compared directly for average value differences and for examination of the variation within a ‘homogeneous’ tissue type.
Despite this approach, it is believed that the remitted irradiance is perhaps a less reliable parameter than the scatter power, because this latter parameter is independent of coupling errors in the imaging system. Though all the spectral images acquired by the scatter imaging system were referenced to Spectralon to minimize referencing artifacts, the average scattered irradiance and scatter amplitude images were not entirely free from such artifacts due to small sample positioning and other instrumental issues. The increased “spread” in the average scattered irradiance compared to the scattering power as observed in is consistent with this behavior. As a result of this, high variance was observed in the average scattered irradiance box plots shown in Figures & . In addition to referencing issues, the scatter amplitude images also exhibited some coupling artifacts and hence were considered less reliable. However, the scatter power images were found to provide consistent trends across different tumor samples. This was due to the fact that scatter power relates to the slope of the wavelength dependent scatter function and hence is relatively free from referencing artifacts39
. Still, interpreting the variation in scatter observed within a single tissue sample clearly indicates that the integrated scatter irradiance can vary up to 50% in most tumor samples, which is a level of heterogeneity which would be significant in optical therapies such as photothermal or photodynamic therapy. These imaging results can be used to better appreciate that there is a clear need for in situ
light dosimetry in photodynamic therapy, since this high heterogeneity level will lead to significant variations in the tissue being treated. Further analysis of the effect that this variation has upon the transmitted light irradiance is clearly warranted.
However, for diagnostic purposes, one additional conclusion is that the scatter power may be a more reliable metric to compare between different samples, as it eliminates errors related to the coupling of the tissue to the microscope imager. The scattering power distributions of various tissue types plotted in Figures and were compared. The median scatter power of proliferative epithelium was found to be about 26% less than the same for mature epithelium in both pancreas and prostate tumor samples. It is well known that a significant portion of light scattered from tissue at higher scattering angles (>40°) comes from nuclear features and intracellular organelles in cytoplasm such as mitochondria 40-44
, although recent work also shows that the stroma between the cells can influence the scattering signature significantly45
. Histopathology observations on the measured tissue types in this study revealed that LPI tumor cells exhibited higher cytoplasmic content and lower cell density, as compared to the HPI tumor cells which contained bigger nuclei and were more tightly packed. The higher organelle density per unit volume seen in mature tumor cells, due to increased cytoplasmic content, would clearly lead to higher scatter power as the number of small scattering features increases.
Exudative necrosis exhibited lowest scattering power response among all tissue types considered, with its median scatter power about 53% less than that of LPI tumor cells in pancreas tumor samples and 59% less in prostate tumor samples. This decrease has a physical explanation since the density of cells is significantly decreased and inflammatory infiltrate (edema, extracellular water or protein solute) occupies much of the space where tumor cells used to be. In comparison to exudative necrosis, regions of focal necrosis primarily contain mature tumor cells with spots of dead cells interspersed, and the scattering from these latter regions was similar to the mature tumor cell regions. The median scatter power of these two regions differed only by about 6%. Thus, the presence of the focal necrotic regions did not seem to have a significant reduction in the overall scatter. However based upon the reduction seen with exudative necrosis, it might be hypothesized that the scatter signal reduces in proportion to the fraction of cellular content within the region.
The prostate tumor samples only contained one region exhibiting significant mature fibrosis. The median scatter power in that region was about 13% lower than that of similar regions found in the pancreas tumor samples. Within the pancreas tumor samples, no significant difference in scatter power was observed between mature and early fibrosis regions. The intermediate fibrosis region was found to scatter more, with about 16% increase in the median scatter power. It is more challenging to interpret what these numbers mean, if anything, about how the morphology relates to the light scattering response. Since the changes are so subtle, it is likely difficult to quantify this further.
Finally, in all the measurements discussed above, the acceptance angle defined by the detection optical train limits the angular region sampled in the back-scatter geometry. So it should be noted that the scatter signal measured here is actually a complex mixture of both the scattering spectrum and the scatter phase function spectrum. Changes in the phase distribution of scattering could have significant influence on the discussed measurements. A comprehensive study on the effects of phase in these measurements would require discrete sampling of a significant portion of the back-scatter hemisphere and is beyond the scope of this study, although is being studied in set ups which are more conducive to this problem44, 46-48
. Unfortunately the problem of measuring angularly resolved data from bulk tissue is complex, and so sampling of scatter from a narrow angle of collection is likely the only practical way to sample the scatter spectrum in thick tissues.