Non-Invasive Hemoglobin Assessment with NIR Imaging of Blood Vessels in Transmittance Geometry: Monte Carlo and Experimental Evaluation

Abstract

Non-invasive methods for determining blood hemoglobin (Hb) concentration are urgently needed to avoid the painful and time-consuming process of invasive venous blood sampling. Many such methods rely on assessing the average attenuation of light over a tissue area where hemoglobin is the dominant chromophore, without separating those areas corresponding to vessels and bloodless tissue. In this study, we investigate whether it is possible to determine hemoglobin levels in the blood by assessing the changes in light intensity when passing through large vessels in comparison to adjacent tissues, using this as a Hb level predictor. Using Monte Carlo light transport modeling, we evaluate the accuracy of determining hemoglobin levels via light intensity contrast and vessel widths estimated in the transmittance illumination geometry and estimate the influence of physiologically significant parameters such as vessel depth, dermis vascularization, and melanin content in the epidermis on the blood Hb prediction error. The results show that physiological variations in tissue parameters limit the mean absolute error of this method to ~15 g/L for blood Hb levels varying in the 60–160 g/L range, which finding is also supported by experimental data obtained for volunteers with different total blood Hb levels that have been determined invasively. We believe the application of new approaches to the non-invasive assessment of Hb levels will lead to the creation of reliable and accurate devices that are applicable in point-of-care and clinical practice.

Keywords

• non-invasive hemoglobin assessment
• diffuse imaging
• Monte Carlo light transport
• machine learning approaches