2.1.

Hybrid Ultrasound and Optoacoustic Tomography

Figure 1 presents a schematic of the hybrid imaging platform with main components. The principal module shared between both systems is a 512-element concave ultrasound detector array (DA) [Imasonics SaS, Voray, France; array type: spherical concave array covering 270 deg; number of channels: 512; mechanical focalization: toroidal focusing; radius of curvature SR $40\mathrm{mm}\pm 1\mathrm{mm}$; radius of the elements (elevation): $37\mathrm{mm}+1/-2\mathrm{mm}$; elementary pitch ($p$): $\sim 0.735\mathrm{mm}/\sim 0.37\mathrm{mm}$; interelement spacing ($e$): 0.1 mm; width of the elements ($h$): 15 mm (chord); center frequency ($-6\mathrm{dB}$): $5\mathrm{MHz}\pm 10\%$; bandwidth ($-6\mathrm{dB}$): $\ge 55\%$ in transmit/receive mode] coupled to laser beam and controlling multiplexer (MUX). The concave multielement detector is connected to two data acquisition (DAQ) systems, each corresponding to one of the acquisition modes. Despite a common mechanism of US and OA signal detection, a second DAQ has to be employed due to (1) conventional OA DAQ electronics not supporting the transmission of US pulses and (2) differences in amplification and input impedance needed to acquire readouts from the two modalities. The data are stored and processed on a personal computer (PC). In the multispectral tomography mode [MSOT, Fig. 1(a)], short laser pulses pass through a diffuser, to ensure even illumination of the specimen. The ultrasound waves produced due to the thermoelastic expansion of the specimen are detected and processed to form images with proprietary software (iThera Medical GmbH, Munich, Germany). In the UST acquisition mode [Fig. 1(b)], the specimen is interrogated by US pulses through an array and reflected signals are sampled with acoustic sensors as specified elsewhere.³⁶

Fig. 1

Hybrid MSOT–UST imaging system: a schematic. (a) MSOT: a laser pulse triggers the DAQ system and illuminates the mouse (M) generating the ultrasonic waves due to light absorption, which are sampled by transducers on a detector ring and reconstructed into an image by PC. (b) UST: the mouse is interrogated by US waves that are generated by the transducers on the DA; the US waves reflected from within the sample due to the acoustic impedance mismatch of the different tissues are subsequently detected and processed into an image. MUX is been used for switching between various acquisitions modes.

A holder consisting of a semiflexible light- and sound-permeable polyethylene membrane is used to horizontally mount the animal into the center of the concave DA. For optimal sound coupling, the detector and the animal holder are immersed in water. The plastic membrane separates the animal from being in contact with the water, while offering $\sim 360\mathrm{deg}$ of water coupling and $\sim 90\mathrm{deg}$ of open access to the animal from the top. In both MSOT and US acquisition modes, one two-dimensional (2-D) cross-sectional plane ($\sim 1\text{-}\mathrm{mm}$ thick) is scanned at a time. Plane sectioning is performed by horizontal translation of the animal holder with respect to the detection ring with a minimum step of 0.1 mm. Due to the unconventional geometry of the detector used, specific reconstruction algorithms have to be employed to form a US image.³⁷ To switch from one acquisition mode into another, a custom made multiplex programmable switchboard was used.³⁶ For MSOT, 9-ns laser pulses of different wavelengths have been used ($\lambda =715$, 730, 775, 800, 850, and 900 nm), which were produced with an Nd:YAG-pumped optico-parametric modulator (InnoLas Laser GmbH, Krailling, Germany) at 10 Hz and per-pulse energy of 120 mJ, with the intensity kept constant throughout the imaging sessions (laser pulse fluence on the surface of the imaged objects under $20\mathrm{mJ}/{\mathrm{cm}}^2$ for similar illumination³⁸). Affirmatory imaging of used specimen at specified above wavelengths was also performed using a commercial MSOT scanner (MSOT256-TF, iThera Medical GmbH, Munich, Germany), with OA signals collected with a 256-element detector ring.³⁴ MSOT images have been acquired at 10 Hz and averaged 10 frames per wavelength, to reduce the variability due to the breathing motion and improve the signal-to-noise ratio.

2.2.

Animals and Experimental Procedures

Young adult mice (Hsd:Athymic Nude-Foxn1nu/nu) were used for these experiments, which were housed in the animal housing facility ($21°\mathrm{C}\pm 2°\mathrm{C}$, humidity $36\%\pm 2\%$ at 12/12-h light/dark cycle) at the Institute of Biomedical Imaging, Helmholtz Zentrum München, with food and water provided ad libitum. All procedures involving animal experimentations were conducted according to the institutional guidelines and the government of Upper Bavaria and complied with German Federal and EU law. Efforts were made to reduce the animal usage and suffering.

MSOT and UST of the head and abdominal organs were conducted and related to anatomical references.³⁹ Experiments were carried out under general anesthesia (1.8% isoflurane in 100% ${\mathrm{O}}_2$ at $0.81\mathrm{ml}/\min$). Baseline measurements at normal breathing were followed by a gas challenge. In total, four breathing states were cycled with supplied gas composition altered as follows: (1) medical air, (2) a mixture of medical air and 10% ${\mathrm{CO}}_2$, (3) 100% oxygen, and (4) medical air. Every breathing condition lasted for $\sim 2\min$ resultant in a total of $\sim 8\min$ per imaging session.

2.3.

Data Analysis and Presentation

Raw data have been analyzed off-line in MATLAB^®. Signals were filtered with a bandpass filter (400 kHz to 8 MHz) with multispectral images reconstructed using a model-based algorithm⁴⁰ resulting in 2-D images of the measured planes. MSOT images of the head region taken at 900 nm with MSOT512 and MSOT256 were automatically coregistered in MATLAB^®. For resolution comparison, the intensity of selected high-contrast regions of interest (ROIs) have been profiled; data were extracted from corresponding locations of coregistered images using MATLAB^® and plotted in arbitrary units (a.u.). Intensity profile graphs were fitted using Gaussian function with full width at half maximum (FWHM) taken as an indication of the resolving power and compared between two (MSOT256 and MSOT512) systems. For correlation of MSOT and UST readouts, semiautomatic data registration was performed in MATLAB^® by matching three pairs of control points selected manually at the corresponding anatomical landmarks defined on both MSOT and US images. Anatomical MSOT images were corrected for light fluence computed with a finite element solution to diffusion approximation assuming homogeneous absorption of $0.3{\mathrm{cm}}^{-1}$ and reduced scattering of $10{\mathrm{cm}}^{-1}$. For studies of blood oxygenation dynamics, measurements were made with MSOT while inhaled gas content was altered as specified above. Prior to data unmixing, the MSOT readouts were calibrated to account for the absorption by a 4-cm-thick layer of water. The ROIs of functional recordings, which included blood vessels and parenchymal tissue distal from major blood supply routes, were defined for tracking the oxygenation saturation (${\mathrm{sO}}_2$) changes overtime. Calibrated multispectral data were subsequently unmixed based on absorption spectra of oxy- and deoxy-hemoglobin using a linear curve-fitting algorithm (MATLAB^®). No optical fluence correction was used prior to unmixing. The values for ${\mathrm{sO}}_2$ were averaged over respective ROIs, with oxygenation levels computed relative to the baseline (denoted as “relative ${\mathrm{sO}}_2$” in tables and figures) defined as the average oxygenation level in the kidney area prior to the breathing challenge. The resulting temporal profiles from the ROIs in the abdominal area were processed with median filter of order 7 to mitigate the noise. To characterize the dynamic response of tissue to gas challenge, data corresponding to the timeframes of 220 to 290 s for brain ROIs and 220 to 355 s for abdominal ROIs were extracted from the selected profiles, with values normalized to the range from 0 to 1 for individual profiles. Sigmoid functions of $1/(1+e^{\tau (x-b)})$ were fitted to the extracted data, where $x$ denotes time, $b$ characterizes the center of the curve, and $\tau$ governs steepness of ascent (higher values mean faster rise). Obtained values of $\tau$ are summarized in Table 1. Graphs and color maps have been generated in MATLAB^® or in EXCEL; the final figures were prepared using IgorPro (6.1) and Adobe Illustrator (CS6).

Table 1

Summary of sO2 dynamics and corresponding values. Anatomical references are numerated in accordance to the MSOT images in Figs. 2 and 4.

3.1.

Hybrid Imaging of the Brain

Figure 2 shows a cryosection of the mouse head ex vivo annotated for major anatomical references of the brain (a), with corresponding UST and MSOT cross sections obtained in vivo (b and c, respectively). The MSOT image is labeled to highlight the regions chosen for the analysis of hemodynamics. As evident, high-contrast structures with distinct contours can be distinguished in both UST and MSOT images, although with notable differences. In MSOT cross sections, the heterogeneous optical absorption at various depths unveils discrete outlines of major vascular elements such as the dorsal sagittal sinus, medial cerebral, maxillary and internal carotid arteries along with the contours of the parietal cortex, hippocampal formation, and optic chiasm.³⁹ On corresponding UST images, strong contrasts due to ultrasound impedance mismatch highlight the skin to skull interface, with most of the signal provided by thick cranial bones causing strong reflection of the US waves while the brain itself appears obscured. As evident from the overlaid UST and MSOT tomograms [Fig. 2(d)], the combination of two modalities results in coregistered image acquired with the shared ultrasonic DA.

Fig. 2

Hybrid UST and MSOT imaging of the brain. (a) Annotated cross section of a mouse head (brain, coronal plane) with corresponding UST, MSOT, and overlaid images (b, c, and d, respectively). Scale bars: 2 mm. Annotations: Cx, cortex; DH, dorsal hippocampus; Th, thalamus; AC, amygdalar complex; and ON, optic nerve. The regions of interest (c, 1 to 6) for analysis of the ${\mathrm{sO}}_2$ dynamics are highlighted on MSOT image with arrows (with measurements summarized in Table 1). 1, sagittal sinus; 2 and 6, maxillary arteries; 3 and 4, parenchymal tissue; and 5, medial cerebral artery. (e and f) Color-coded relative ${\mathrm{sO}}_2$ maps obtained at indicated breathing phases overlaid with anatomical MSOT images. (g and h) Time courses of changes from respective regions of interests (enboxed, 1 and 4) corresponding to sagittal sinus (g) and parietal cortex (h). Blue circles and error bars denote mean relative ${\mathrm{sO}}_2$ and standard deviations of measurements within respective ROIs, correspondingly. Experimental conditions are indicated within individual graphs. Scale bars: 1.8 mm.

Due to its spectral dimension,¹⁸ MSOT can obtain readouts of the brain hemodynamics and ${\mathrm{sO}}_2$ in conjunction with anatomical data in vivo. Figures 2(e) and 2(f) show images of the dorsal compartments of mouse brain overlaid with color-coded oxygenation maps corresponding to different breathing conditions (air and 100% ${\mathrm{O}}_2$, respectively) with supply of 100% ${\mathrm{O}}_2$ increasing tissue oxygenation levels. The overall dynamics of the ${\mathrm{sO}}_2$ in defined ROIs [Figs. 2(e), 2(f); enboxed 1 and 4] of the dorsal cerebrum with their changes related to the breathing conditions are summarized in Figs. 2(g) and 2(h), respectively (blue circles and error bars represent correspondingly mean relative oxygenation and standard deviation within respective ROIs). After a stable baseline recording, supply of 10% ${\mathrm{CO}}_2$ causes deoxygenation of the brain tissue and sagittal sinus while supply of 100% ${\mathrm{O}}_2$ leads to a steep increase in ${\mathrm{sO}}_2$ in both areas.

To evaluate the kinetics of the hemoglobin gradient changes, we analyzed the rates of the relative ${\mathrm{sO}}_2$ changes upon ${\mathrm{O}}_2$ challenge. Figure 3 presents the graphs of the kinetics of ${\mathrm{sO}}_2$ in brain tissue (a) and sagittal sinus (b). Black circles denote normalized data points pulled from regions marked by arrows 3 (a) and 1 (b) in Fig. 2(c), while red lines correspond to the fitted sigmoid functions. Faster ${\mathrm{sO}}_2$ rates within the brain tissue are evident as compared to that of sinus, consistent with measurements of the ${\mathrm{sO}}_2$ dynamics and comparison between several brain areas and major blood vessels (Table 1). These findings suggest a tighter coupling of the oxygenation of the neural tissue with the breathing state. It is important to note that changes in ${\mathrm{sO}}_2$ induced by the gas challenge were reversible, confirming that the effects of experimental interventions were transient and remain within physiological limits, and were not associated with irreversible impairments of the brain hemodynamics.

Fig. 3

Graphical representation of the kinetics of the ${\mathrm{sO}}_2$ within large blood vessels and parenchyma of the brain measured with MSOT. (a and b) Representative ${\mathrm{sO}}_2$ dynamics overtime pulled from ROIs [corresponding to arrows: 3 and 1, in Fig. 2(c), respectively]. Individual data points overtime are fitted with sigmoid functions with values (tau) defining the slope summarized in Table 1.

3.2.

Comparison of 256 Versus 512 Multispectral Optoacoustic Tomography

To investigate if an increase in the number of ultrasonic detectors in hybrid MSOT–UST platform affects the image quality and resolution of the system, we compared the intensity profiles of defined ROIs on head cross-sectional images acquired with hybrid MSOT512 with those obtained by stand-alone MSOT256.

Figures 4(a) and 4(b) present an annotated histological cross section of a mouse head ex vivo (coronal section) taken at the plane that corresponds to $-5.3\text{-}\mathrm{mm}$ Bregma of the mouse brain atlas.⁴¹ Figures 4(c)–4(f) show tomographic cross sections of the same anatomical plane obtained with the MSOT512 (c and d) and MSOT256 (e and f) systems at 715-nm (c and e) and 900-nm (d and f) wavelengths. As optical absorption of brain tissue is wavelength dependent, the 715- and 900-nm images reveal spectral differences corresponding to different molecular constituents.¹⁶ While hemoglobin provides the major contrast at 715 nm, at 900 nm, water and adipose tissue are expected to contribute more to contrast generation.

Fig. 4

Intravital imaging of a mouse brain with MSOT256 and MSOT512: a comparative analysis. (a) Annotated histological cross section (HIST) of a mouse brain with (b) schematic representation of the corresponding anatomical cross section of the brain: Cx, cortex; DH, dorsal hippocampus; Th, thalamus; AC, amygdalar complex; and ON, optic nerve. Scale bars: 2 mm. (c–f) Tomographic images of the mouse brain obtained with MSOT512 (c and d) and MSOT256 (e and f) at 715 nm (c and e) and 900 nm (d and f). Arrows (I–III) mark approximate locations of intensity profiles extractions. Scale bars: 2.5 mm. (g) Summary of intensity profiles extracted from the coregistered images (d and f). Intensity profiles from MSOT256 (blue, dashed) and MSOT512 (black, solid) are matched at their maxima (left column), fitted with the Gaussian functions (middle and right columns). The FWHM values are presented in Table 2.

Comparison of the intensity profiles obtained from three defined anatomical references revealed subtle but interesting differences. Figure 4(g) (I to III) present the intensity profiles extracted from locations denoted in Figs. 4(d) and 4(f) (I, II, and III), respectively. The normalized graphs (g, left column) correspond to MSOT512 (black) and MSOT256 (blue, dashed). For comparison of the resolving power of MSOT512 versus MSOT256, we fitted Gaussian curves to the normalized intensity profiles (I to III) and computed their relative FWHM. Figure 4(g) (middle and right columns) shows Gaussian fits for the MSOT256 and MSOT512 measurements, respectively. The computed FWHM values are summarized in Table 2 and advocate improved imaging resolution with MSOT512.

Table 2

Comparative FWHM fits of the intensity profiles of sagittal sinus, maxillary artery, and subcutaneous artery using MSOT256 and MSOT512 (units in mm).

3.3.

Hybrid Imaging of Abdominal Organs

Figure 5 presents a cryosection of the mouse abdomen ex vivo annotated for major anatomical references (a), with corresponding UST and MSOT cross sections obtained in vivo (b and c, respectively) and their coregistered overlay (d). High absorbance of major blood supply routes, such as spinal branch of lumbar arteries and vertebral column with adjacent caudal vena cava along with large parenchymal organs such as left and right kidneys, spleen, the head of the pancreas, and epigastria vessels, is visible on MSOT images. UST images, on the other hand, offer better contrast and visualization of organs at greater depths, revealing kidneys, liver, pancreas, vertebral column with the spinal cord and other structures. Importantly, blood reach formations, such as medium and small vascular elements, and parts of liver hardly distinguishable on UST images are revealed better by MSOT, while collagen and connective tissue rich structures and bones reveal higher contrast at UST. Thus, UST of the abdominal organs unveils rich anatomical data on deeper structures, while MSOT provides complementary features based on optical contrast. The latter is especially important for functional imaging, as it affords intravital assessment of local ${\mathrm{sO}}_2$ and hemoglobin concentration. Figures 5(e) and 5(f) present abdominal cross sections merged with color-coded oxygenation maps computed from MSOT data corresponding to two different breathing phases [air (e) and 100% ${\mathrm{O}}_2$ (f)]. Similar to brain measurements, assessment of the ${\mathrm{sO}}_2$ kinetics of abdominal organs under different breathing conditions in two ROIs (enboxed 1 and 4) revealed clear correspondence of the hemodynamics with the composition of supplied gas, with dynamics of the ${\mathrm{sO}}_2$ changes related to the gas challenges readily visible [Figs. 5(g) and 5(h)]. After baseline recording while breathing air, supply of ${\mathrm{CO}}_2$ causes gradual deoxygenation of the abdominal structures while supply of pure ${\mathrm{O}}_2$ results in rapid increase in tissue and blood oxygenation. Similar to the brain, these processes remain within physiological limits and reversed upon supply of normal breathing air.

Fig. 5

Hybrid UST and MSOT imaging of abdominal organs. (a) Annotated cross section of a mouse abdomen with UST, MSOT, and overlaid images (b, c, and d, respectively). Annotations: SC, spinal cord; Ki, kidney; Sp, spleen; VC, vena cava; PV, portal vein; Pa, pancreas; Lv, liver; and In, intestine. Scale bars: 3 mm. The regions of interest (1 to 6) for analysis of the ${\mathrm{sO}}_2$ dynamics are highlighted on (c) MSOT image with arrows (summarized in Table 1). 1 and 5, kidneys; 2, 3, and 6, superficial vessels; and 4, spinal branch of lumbar artery. (e and f) Color-coded relative ${\mathrm{sO}}_2$ maps obtained at indicated breathing phases overlaid with anatomical images. (g and h) Time courses of changes from respective regions of interests (enboxed, 1 and 4). Blue circles and error bars denote mean relative ${\mathrm{sO}}_2$ and standard deviations of measurements within respective ROIs, correspondingly. Experimental conditions are indicated within individual graphs. Scale bars: 2 mm.

To evaluate the kinetics of the hemoglobin gradient changes in abdominal organs, we analyzed the rates of the relative ${\mathrm{sO}}_2$ changes upon ${\mathrm{O}}_2$ challenge. Figure 6 presents the analysis of the ${\mathrm{sO}}_2$ kinetics of kidney tissue (a) and an abdominal blood vessel (b), with data obtained from regions marked by arrows 1 and 3 in Fig. 5(c), respectively. Black circles correspond to the normalized data points with red lines showing the fitted sigmoid functions. Parenchymal organs showed slower ${\mathrm{sO}}_2$ rates compared to those of the brain (Table 1), supporting tighter coupling of the brain ${\mathrm{sO}}_2$ with breathing state and overall greater oxygen consumption of neural tissue and its lower tolerance to hypoxia.^42,43

Fig. 6

Graphical representation of the kinetics of the ${\mathrm{sO}}_2$ within large blood vessels and parenchyma of abdominal organs measured with MSOT. (a and b) Representative ${\mathrm{sO}}_2$ dynamics overtime pulled from ROIs [corresponding to arrows: 1 and 3, in Fig. 5(c), respectively]. Individual data points overtime are fitted with sigmoid functions with values (tau) defining the slope summarized in Table 1.