Anthropomorphic tissue-mimicking phantoms for oximetry validation in multispectral optical imaging

Abstract

Significance: Optical imaging of blood oxygenation ( ${\mathrm{sO}}_2$ ) can be achieved based on the differential absorption spectra of oxy- and deoxyhemoglobin. A key challenge in realizing clinical validation of the ${\mathrm{sO}}_2$ biomarkers is the absence of reliable ${\mathrm{sO}}_2$ reference standards, including test objects.

Aim: To enable quantitative testing of multispectral imaging methods for assessment of ${\mathrm{sO}}_2$ by introducing anthropomorphic phantoms with appropriate tissue-mimicking optical properties.

Approach: We used the stable copolymer-in-oil base material to create physical anthropomorphic structures and optimized dyes to mimic the optical absorption of blood across a wide spectral range. Using 3D-printed phantom molds generated from a magnetic resonance image of a human forearm, we molded the material into an anthropomorphic shape. Using both reflectance hyperspectral imaging (HSI) and photoacoustic tomography (PAT), we acquired images of the forearm phantoms and evaluated the performance of linear spectral unmixing (LSU).

Results: Based on 10 fabricated forearm phantoms with vessel-like structures featuring five distinct ${\mathrm{sO}}_2$ levels (between 0 and 100%), we showed that the measured absorption spectra of the material correlated well with HSI and PAT data with a Pearson correlation coefficient consistently above 0.8. Further, the application of LSU enabled a quantification of the mean absolute error in ${\mathrm{sO}}_2$ assessment with HSI and PAT.

Conclusions: Our anthropomorphic tissue-mimicking phantoms hold potential to provide a robust tool for developing, standardising, and validating optical imaging of ${\mathrm{sO}}_2$ .

Keywords: anthropomorphic phantoms; hyperspectral imaging; optical imaging; oximetry; photoacoustic imaging.

Fig. 1

Workflow and key contributions of this study. (Top) First, 26 dyes were investigated to mimic the absorption spectra of oxyhemoglobin (${\mathrm{HbO}}_2$) and deoxyhemoglobin (Hb) in the wavelength ($\lambda$) range of 700 to 850 nm. After a selection and optimization process, two proxy dyes that could be mixed to five levels of oxygen saturation (${\mathrm{sO}}_2$) (0%, 30%, 50%, 70%, and 100%) were found. Second, for realistic tissue morphology, a 3D-printable mold was created based on an open-source magnetic resonance (MR) image of a human forearm. (bottom) Third, ten forearm phantoms were fabricated and finally, photoacoustic (PA) and hyperspectral (HS) images were acquired and we show that example images of these can be used to validate oximetry methods.

Fig. 2

3D-printable models were constructed from an open-source magnetic resonance (MR) image for a structured phantom fabrication and annotation process. From top left to top right: The MR image was manually segmented for bones, muscles, and fat. A physician performed a quality check ensuring morphological correctness. Based on this segmentation, a 3D-printable model of the outer hull of the forearm has been designed and printed including the two bones, radius and ulna, that will stay in the mold as positives. From bottom right to bottom left: Each half of the mold is filled successively with background material and vessels are placed, which are annotated in a cross-sectional sketch of the mold. Finally, after acquiring photoacoustic (PA) images from multiple angles, each vessel in the PA images can be assigned to its corresponding oxygen saturation during manual annotation.

Fig. 3

IR-1061 and Spectrasense-765 can mimic the absorption characteristics of the blood. Dashed lines indicate the absorption coefficient (${\mu }_a$) spectra of oxyhemoglobin (${\mathrm{HbO}}_2$) and deoxyhemoglobin (Hb). Solid lines indicate the measured absorption spectra of the proxy dyes. Bands around the measured spectra indicate the standard deviation across the 12 measurement points for each optical sample slab. Mean absolute errors between targets and proxy dyes are 0.78 and $0.64{\mathrm{cm}}^{-1}$ for ${\mathrm{HbO}}_2$ and Hb, respectively.

Fig. 4

Five oxygen saturation (${\mathrm{sO}}_2$) levels were used for forearm phantom fabrication. Based on IR-1061 and Spectrasense-765, five ${\mathrm{sO}}_2$ levels (in %) were derived with the respective mixture ratios of 100:0, 97:3, 95:5, 93:7, and 90:10. Panels (a) and (b) represent the absorption coefficient (${\mu }_a$) and scattering coefficient (${\mu }_s$), respectively. Solid lines are the spectra that are used as endmembers (0% ${\mathrm{sO}}_2$ and 100% ${\mathrm{sO}}_2$) for linear spectral unmixing (LSU). Dashed lines represent the intermediate levels, and the corresponding percentages in the legend are the LSU results when using the solid lines as endmembers. Bands around the spectra indicate the standard deviation across the 12 measurement points for each optical sample slab.

Fig. 5

Measurement setup for hyperspectral imaging (HSI) (a), corresponding image (b), and signal correlation of the 50% oxygen saturation (${\mathrm{sO}}_2$) superficial vessel (c). The phantoms were mounted on a rotational stage, and the camera was adjusted for each phantom such that the middle of the phantoms was in focus (red dot). Images from eight angles in steps of 45 deg were acquired per phantom. Panel (b) shows an example of an RGB-reconstructed image using the spectral range of 530 to 725 nm including a region of interest (ROI) indicated by yellow margins in the middle of a superficial vessel. The black solid lines in panel (c) represent the resulting linear regression function with the corresponding Pearson correlation coefficient ($R$ value). The inset plot shows the measured absorption (blue) and estimated absorption (green, using the correlation function) as qualitative confirmation.

Fig. 6

Measurement setup for photoacoustic tomography (PAT) (a), corresponding image (b), and signal correlation of the 50% oxygen saturation (${\mathrm{sO}}_2$) superficial vessel (c). The phantoms were mounted on a rotational stage in a water bath, and images were acquired from eight angles in steps of 45 deg in three distinct locations along the phantoms. For each measurement location, the PAT probe, which was attached to a mechanical arm to minimize motion during image acquisitions, was adjusted such that it was approximately in the middle of the phantoms and 2 mm above the uppermost point of the phantom. Panel (b) shows an example PA image at 800 nm including a region of interest (ROI) indicated by yellow margins in the middle of a superficial vessel. The black solid lines in panel (c) represent the resulting linear regression function with the corresponding Pearson correlation coefficient ($R$ value). The inset plot shows the measured absorption (blue) and estimated absorption (green, using the correlation function) as qualitative confirmation.

Fig. 7

Error analysis of three oximetry methods for photoacoustic tomography as a function of depth. Mean absolute error (MAE) in percentage points (p.p.) of linear spectral unmixing (LSU) without (blue) and with calibration (green) using a superficial vessel, and with fluence compensation (pink) is plotted against the depth of the evaluation pixel for both the whole phantom (top) and for the vessels only (bottom). The bands around the solid lines indicate the standard deviation.