Optical phantoms play an important role in calibrating and evaluating the performance optical spectroscopy and spectral imaging systems. Many different models of liquid-based, gelatin-based, and solid optical phantoms have been developed for this purpose [1]. In our previous works, different materials were selected to fabricate liquid-based phantoms for study of diffuse reflectance spectroscopy and fluorescence spectroscopy [2, 3]. In general, desirable characteristics of optical phantoms are well-controlled optical properties, and the ability to mimic the spectral shape of tissue scattering and absorption. While hemoglobin (Hb) absorption represents a key source of crosstalk in optical properties extraction and distortion in collected signal [3, 4], its optical signatures over a wide spectral range are the keys to non-invasive assessment of tissue oxygenation with multi-spectral imaging (MSI) systems [5, 6]. Although powdered Hb (i.e., H0267, Sigma-Aldrich) has been widely used to fabricate phantoms for calibration of fiber-optics-based techniques [2–4, 7], it is not an ideal material for varying and accurately controlling of oxygen saturation. This is due to the autoxidation of the present heme group leading to the formation of the ferric derivative (methemoglobin) over time in addition to the already existed small amount of methemoglobin in the powder. Furthermore, powdered Hb is expensive whereas wide field near-infrared imaging technique requires large-volume phantoms to meet semi-infinite condition for photon propagation. Thus, the cost of phantom production significantly increases in such cases.

Meanwhile, although real human blood can be a good alternative to study oxygenation, it requires complicated sample preparation and storage. For example, whole blood needs to be in a pH and temperature controlled medium, and requires oxygenators and blood analyzers to change and test oxygen levels [8, 9]. Another alternative is to use non-biological dye to mimic the absorption of whole blood in its deoxygenated state [10]. However, such approach is limited to a few near infrared wavelengths.

The purpose of this study is to develop a simple gelatin-based phantom model for quick calibration of spectral imaging systems at a broad-band spectrum. These phantoms should exhibit the spectral absorption characteristics of Hb and can be used to assess oxygen information. Beer Lambert law and Mie theory were used to guide the design of the phantom and the resulting optical properties and StO2 were validated experimentally. Measurements of the phantoms are also performed on a FDA-approved multi-spectral imaging (MSI) system.

Figure 2 shows μ_a spectra of four control phantoms over period of 2 weeks whereas (Figure 3) compares the normalized μ_a spectra of four phantoms in the wavelength region of 450–1,000 nm. In general, the characteristic absorption peaks at 540 and 580 nm were observed in all phantoms (Figures 2, 3A), and no significant change in μ_a spectra was observed in the period of 2 weeks in all phantoms (Figure 2). Figure 2E shows the reduced scattering. In addition, a noticeable μ_a peak at the 750 nm region and a shift upward in μ_a spectrum in the 650 nm region were observed in phantom with high yeast concentrations of 15% (Figure 3A). These are well-known features when oxy-Hb changes into deoxy- Hb [18, 19]. Figure 3B shows StO2 values in four control phantoms over 2 weeks. In each phantom, the percentage difference between StO2 value in day 1 and week 2 was within 6%. On average, the calculated StO2 value was 94.6% ± 2.7%, 92.2% ± 3.2%, 81.93 ± 5.2%, 48.2 ± 5.9% for phantom Ctrl.HbY0, CTrl.HbY5, Ctrl.HbY10, and Ctrl.HbY15, respectively. Error bars in Figure 3B represent the standard deviation for measurements of three similar phantoms that were fabricated separately but with the same yeast concentration (and everything else). Such averaging accounts for errors in fabrication procedure. Comparing week 1 and week 2, StO2 increases 5% in Ctrl.HbY10, and decreases 3% in Ctrl. HbY5, and 7% in Ctrl. HbY15.

All the gel phantoms were prepared at 65°C in 30 s. Therefore, to determine the effect of heating duration on StO2, a set of Ctrl.HbY0 phantoms were made, then heated at 65°C in different time periods up to 1 min. In general, prolonging heating time will increase absorption due to the loss of water through evaporation (Figure 4A). More specifically, μ_a increases by 2.5 and 25.4% compared to the reference μ_a if heating period is 30 s and 1 min, respectively. However, μ_a spectral shape was not affected by heating (Figure 4B). Because StO2 is determined by of the relative μ_a values on the spectrum [4, 20], the stability of μ_a spectral shape ensures stability of the StO2 value. In these phantoms, the average StO2 value was 96 ± 3%.

The MSI system was used to capture the experimental phantoms so that 2D spatial images of StO2 were generated for each phantom over course of 1 week to account for day to day variations and aging. Each phantom was imaged at the date of manufacture, day 3, and day 7. For each phantom, averaging over three regions of interests (ROIs) was performed to account for spatial variation. (Figures 5A–C). As shown in Figure 5D, a good agreement between spectrophotometer measurements and MSI measurement is obtained with an averaged percentage difference of 8%.

In this paper, the phantoms are selected so that the reflectance signal is within the detection capability of the commercial multi-spectral imaging system. Currently, this system can extract useful information from highly turbid media with minimal absorption (where diffusion theory is valid and where signal to noise ratio is high enough). A further increase in yeast concentration and/or hemoglobin concentration can increase the total absorption coefficient, reducing the accuracy of the Presygen system, and hence is avoided. Meanwhile, vaporization is significant with prolonged aging, causing an increase in absorber concentration, thus increasing total absorption of the phantoms. Therefore, longer aging is not considered when using the Presygen system. Nevertheless, the selected ranges of optical properties and oxygen saturation in the current phantom model are within literature for tissues and optical phantoms [7, 21, 22].

In this study, we have demonstrated a gel-based optical phantom model with controlled oxygenation that can be used for quick calibration of spectral imaging devices without the use of whole blood or powdered hemoglobin, and that yeast can be used to decrease the oxygen saturation of a commercial blood product that is fully oxygenated and is more cost effective than powdered hemoglobin. In addition, although increasing yeast concentration decreases oxygen saturation of the phantom, the phenomenon does not occur linearly. Also, a short period of heating can vary the concentration of hemoglobin but does not have major effect on the phantom oxygen saturation. In other words, heating changes the magnitude of absorption but does not change its spectral shape. Specifically, μ_a increases about 2.5% due to heating duration of 30 s at 65°C whereas the absorption spectral shape of the selected hemoglobin was not affected. In addition, the current gel model is able to produce the consistent results in the period of 2 weeks with a relative uncertainty of <6%. The phantom is low cost and simple to make with excellent repeatability, which is critical for its practical application. Therefore, the current gel phantoms have desirable characteristics for calibrating multi-spectral imaging systems, including cost effectiveness, well-controlled optical properties and oxygen saturation, and the ability to retain stability over an adequate period of time.

The datasets generated for this study are available on request to the corresponding author.

VL supervised the production of all optical phantoms, applied theories and developed algorithms to control the optical properties, and oxygen saturation in the phantom. MM assisted with data measurement and phantom fabrication. SG and BW measured the phantoms with the spectral imaging system. JH and QF advised the progress of the study.