Environmentally-Controlled Near Infrared Spectroscopic Imaging of Bone Water

Abstract

We have designed an environmentally-controlled chamber for near infrared spectroscopic imaging (NIRSI) to monitor changes in cortical bone water content, an emerging biomarker related to bone quality assessment. The chamber is required to ensure repeatable spectroscopic measurements of tissues without the influence of atmospheric moisture. A calibration curve to predict gravimetric water content from human cadaveric cortical bone was created using NIRSI data obtained at six different lyophilization time points. Partial least squares (PLS) models successfully predicted bone water content that ranged from 0-10% (R = 0.96, p < 0.05, root mean square error of prediction (RMSEP) = 7.39%), as well as in the physiologic range of 4-10% of wet tissue weight (R = 0.87, p < 0.05, RMSEP = 14.5%). Similar results were obtained with univariate and bivariate regression models for prediction of water in the 0-10% range. Further, we identified two new NIR bone absorbances, at 6560 cm^-1 and 6688 cm^-1, associated with water and collagen respectively. Such data will be useful in pre-clinical studies that investigate changes in bone quality with disease, aging and with therapeutic use.

Figure 1

Small environmental chamber with input and output for airflow tubing, and humidity sensor.

Figure 2

Large environmental chamber with input and output for airflow tubing, and humidity sensor.

Figure 3

RH inside the imaging chamber during one hour of spectral data collection from wet and dry bone samples in high (a) and low (b) RH conditions. Average pixel intensity at 5184 cm⁻¹ (inverted second derivative intensity of NIRSI water absorbance) inside the imaging chamber during one hour of spectral data collection from wet and dry bone samples in high (c) and low (d) humidity.

Figure 4

Average NIRSI pixel intensities at 5184 cm⁻¹ (water content) of wet and lyophilized bone samples under 0% RH in the large chamber. Wet samples gradually lose water content over an hour, while dry samples maintain their water content. (*) Average pixel intensities at 5184 cm⁻¹ for wet sample group at 0 and 60 minutes were statistically significant at p < 0.05.

Figure 5

(a) Raw NIR spectra of wet (hydrated) and dry bone. (b) NIR second derivative of wet (hydrated) and dry bone. The water and matrix peaks are more resolved in the second derivative spectra compared to raw spectra. A reduction in the 5184 cm⁻¹ water peak can be seen in both raw and second derivative spectra of dry bone. (c) Second derivative NIR spectra of serially dehydrated bone. The absorbance of the water peaks (5184 cm⁻¹, 6560 cm⁻¹ and 7008 cm⁻¹) decreased with increasing lyophilization time.

Figure 6

(a) The lyophilization time points separate from right to left, reflective of increasing lyophilization time. (b) Factor 1, which underlies most of the data separation, is dominated by the 5184 cm⁻¹ water absorbance, while factor 2 is dominated by matrix absorbances.

Figure 7

Independent NIRSI prediction of gravimetric water in cortical bone over the (a) 0–10%, and (b) 4–10% (physiologic) water content ranges. RMSEP for the two models 0–10% range and 4–10% range are 7.39% and 14.5% respectively.

Figure 8

Schematic of the experimental setup for data collection from human cadaveric tissue samples used in creation of the water calibration curve.