2.1.

Magneto-Motive Optical Doppler Tomography System Description

A typical spectrometer-based MM-ODT system is shown in Fig. 1 along with the experimental setup. The broadband light source in this study was a super-luminescent diode (SLD-35-HP, Superlum, Ltd.) with a center wavelength of 870 nm and a 3-dB bandwidth of 65 nm. The incident light beam from the laser source was split evenly by using a $50:50$ optical fiber coupler into a reference path and a sample arm. A galvanometer-based optical scanner was integrated for transverse scanning of the sample arm and it was driven at a frequency of 56 Hz. To obtain optimal results and to minimize the saturated power owing to the backscattered beams, the imaging axis was not maintained coaxially with the magnetic field gradient. Further, Fig. 1 shows the schematic of a solenoid with a core that generated the magnetic field in the study and the resulting mechanical strain fields, which were applied to the sample for moving the nanoparticles in the sample.

Fig. 1

Schematic of the MM-ODT experimental setup. BLS, broadband laser source; C, collimator; FC, fiber coupler; FG, frame grabber; GS, galvano-scanner; OL, objective lens; and PC, polarization controller.

Kasai’s autocorrelation velocity estimation method was used for determining the optical Doppler velocity.³⁴ According to the Kasai autocorrelation velocity estimation method, when a broadband light beam is incident on a flowing fluid, reflection can occur due to the presence of floating solid particles or air bubbles with velocity that is approximately similar to the fluid velocity.³⁵ The Doppler frequency shift induced by a moving particle can be expressed as $F_D=\mathrm{\Delta }\stackrel{\rightharpoonup }{\varphi }/2\pi t$, where $t$ is the time between successive A-scans and $\mathrm{\Delta }\stackrel{\rightharpoonup }{\varphi }$ is the mean phase change. The Doppler frequency shift is continuously observed during the magnetic activation. The Doppler flow velocity $\stackrel{\rightharpoonup }{V}$ can be expressed as $\stackrel{\rightharpoonup }{V}={\lambda }_0{f}_D/2n\cos (\theta )$, where ${\lambda }_0$ is the center wavelength, $n$ is the refractive index, and $\theta$ is the angle between the flow and probe beam direction (the Doppler angle).

Moreover, an iron core solenoid coil (470 turns, $2.6\mathrm{\Omega }$ impedance) with a cone-shaped ferrite core was placed below the sample to provide a magnetic flux toward the entire area of the sample during MM-ODT imaging, for moving the nanoparticles in the sample. A waveform generator (Agilent, 33220A) was used for applying the magnetic field. The magnetic field intensity was dramatically increased by using the high-power combination; the magnetic field intensity was $B_{\max }=1\mathrm{T}$ and $\mathrm{\Delta }{|B|}^2=220{\mathrm{T}}^2/\mathrm{m}$ at the tip of the core toward the targeted sample. Note that in this experiment, the maximal magnetic flux density (0.5596 T) was measured when the solenoid core was at a distance of 5 mm from the sample. However, due to the limitations of the experimental setup, the sample had to be kept at a distance of 10 mm from the solenoid core and the measured flux density at this stage was $\sim 0.33077\mathrm{T}$.

Instead of using a lock-in method-based sinusoidal magnetic force, a continuously modulating 5-Hz square wave magnetic actuation frequency with a maximal voltage of 90 $V_{p-p}$ was applied for generating the magnetic force to achieve distinguishable Doppler signals. For the maximal distinguishable detection capability of SPIO nanoparticles, the magnetic actuation frequency was set to 5 Hz to avoid reducing the magnetic actuation duty cycle below the rising time of the iron core solenoid, which would barricade the magnetic flux intensity from reaching its maximal value, according to the slew rate theorem.

Unlike the other conventional magneto-motive systems, the lateral components of the magnetic field are negligible and a concentrated magnetic field, which is dominant in axial direction applied toward the imaging region that minimizes the negative effects of the nonuniform spatial distribution of magnetic fields. Thus, the displacement of the nanoparticles in this experiment was primarily in the axial direction.

2.2.

Preparation of the Superparamagnetic Iron Oxide Nanoparticles Injected into Ex Vivo and In Vivo Samples

Commercially available SPIO magnetic nanoparticles were used as the contrast agents in this experiment. To observe the Doppler shifts in the tissue samples, we prepared a solution containing nanoparticles. As an initial preparation, SPIO Ferumoxide nanoparticles (AMI-25, Guerbet Berlex Laboratories, France) with diameters in the 80- to 150-nm range were used in this experiment. The nanoparticles solution consisted of 50 ml of 5% dextrose solution and 1 ml pure Ferumoxide nanoparticles with a concentration of $0.67\times 1012$ iron particles $\mu {\mathrm{l}}^{-1}$ ($1.12\mu \mathrm{g}$ iron $\mu {\mathrm{l}}^{-1}$). The prepared nanoparticles were dissolved in the glucose solution, and the solution was injected directly into a pork belly sample and an in vivo melanoma sample, for MM-ODT image acquisition. The experiment was repeated for different SPIO concentrations and solenoid voltage values in order to validate the system performance. Subsequently, in vivo melanoma skin cancer tissue was imaged to determine the performance of the MM-ODT system. The used animal model was a mouse (42- to 48-days old, 23 to 26 g) with an immune deficiency. The animal experiments were performed in accordance with the guidelines of the Institutional Animal Care and Use Committee of Kyungpook National University. Melanoma cells were transplanted into the hind leg skin of the mouse, which caused an externally visible solid tumor (5 to 9 mm). The transplanted melanoma cells were grown rapidly, in a short cancer cell growth cycle, as a result of the immune deficiency of the mouse.

2.3.

Implementation of Signal Processing for Magneto-Motive Optical Doppler Tomography

A compute unified device architecture (CUDA)-activated digital signal processing technique was implemented for enhancing the acquisition speed in MM-ODT measurements. Figure 2 shows the flowchart, demonstrating the data path, the thread events, and the buffering for the CUDA-activated MM-ODT system. First, the data acquisition thread stored incoming two-dimensional (2-D) raw signals in the first buffer allocated in the host memory and called the signal processing thread.

Fig. 2

CUDA-activated signal processing.

Then, the self-iterated acquisition thread continually transferred the incoming signals into the second buffer without any temporal delay between adjacent acquisition events. The signal processing thread copied the frame data stored in the buffers of the host memory. $k$-domain linearization was completed by using full-range $k$-domain linearization.³⁶ The reconstructed MM-ODT images were transferred back to the host memory for displaying. Therefore, the frame rate of the utilized spectral domain optical coherence tomography system was increased by employing this rapid image processing procedure.

3.1.

Ex Vivo Magneto-Motive Optical Doppler Tomography Imaging of Pork Belly Sample

Prior to performing the in vivo validation, we investigated the efficacy of the proposed system with respect to an ex vivo pork belly sample implanted with nanoparticles. The motion of the implanted nanoparticles was barely observable in the corresponding B-mode images of MM-OCT and MM-ODT [Figs. 3(a) and 3(b)]. The magenta-colored MM-ODT signals were overlaid on the blue-colored MM-ODT signals, making it difficult to precisely determine the nanoparticles motion. One of the reasons for the color variation was the frequency mismatch between the galvano-scanner frequency and the magnetic flux frequency. Consequently, the galvano-scanned part of the sample failed to experience the maximal magnetic force in the consistent and timely manner. Therefore, for a better identification, M-mode images consisting of $780\times 500\text{pixels}$ axially and temporally were acquired and are shown in Figs. 3(c) and 3(d). The nanoparticles motion was clearly and continuously observed in both the MM-OCT and MM-ODT M-mode images. However, the absence of the galvano-scanner motion yielded a stronger enhancement of the periodically varying MM-ODT M-Mode signal, compared with the MM-OCT M-mode signal.

Fig. 3

MM-OCT and MM-ODT images of SPIO nanoparticles injected into the ex vivo pork belly sample. (a) MM-OCT image, (b) MM-ODT image with Doppler signals, (c) MM-OCT M-mode image, and (d) MM-ODT M-mode image with identifiable Doppler signals. White vertical bar: $500\mu \mathrm{m}$ and white horizontal bar: 20 ms.

Figure 4 shows the quantitative evaluation of the average Doppler velocities as a function of the SPIO nanoparticles concentration and the applied solenoid voltage. Closed circles and dashed lines denote the measured values and the linear fitting results, respectively. Note that the experiment was repeated five consecutive times. The error bars illustrate the fluctuation range between the obtained minimum and the maximum Doppler velocity values for each corresponding concentration and induced voltage values. The linear regression represents the general tendency of linearly increasing averaged Doppler velocity values. The initially applied concentrations were 0.5, 1.0, 1.5, and $2.0\mathrm{mg}/\mathrm{ml}$ and the injected volume was $100\mu \mathrm{l}$. Second, the solenoid voltage was varied between 0 and 90 $V_{p-p}$ with an SPIO concentration of $2.0\mathrm{mg}/\mathrm{ml}$ [Fig. 4(b)]. The experimental Doppler velocity data were calculated by using the Kasai autocorrelation velocity equations [Eqs. (1) and (2)].³⁵ As expected, the measured average Doppler velocity increased with increasing the nanoparticles concentration and with increasing the applied solenoid voltage.

Fig. 4

Average measured Doppler velocity versus the (a) SPIO nanoparticles concentration and (b) the solenoid voltage. LR, linear regression. The error bars represent the fluctuation between the minimum and the maximum values.

3.2.

Histological Comparison of Melanoma Tissue Samples With and Without Injected Nanoparticles

Prior to the MM-ODT image acquisition, histological analysis was conducted to confirm the existence of melanoma cells and magnetic nanoparticles. Hematoxylin and eosin (H&E) dye reagent was used for identifying melanoma cells, and Prussian blue dye reagent was used for identifying magnetic nanoparticles. Figure 5(a) shows the H&E-treated melanoma tissue and Fig. 5(b) shows the magnified view of the dashed rectangular area in Fig. 5(a). The H&E reagent is the most commonly used histological staining reagent and it contains an acidic dye, eosin, and a basic dye, hematoxylin. It stains melanin structures in purplish blue. The left portion of the tissue region in Fig. 5(b) has a purplish blue stain owing to a high melanin pigment content characteristic of melanoma cells, while the right portion of that same tissue region has a very slight purple stain, owing to a low melanin pigment content. Hence, the H&E control experiment confirmed the presence of melanoma cells. Melanoma tissues treated by the Prussian blue reagent were used for identifying the implanted nanoparticles, as shown in Figs. 5(c) and 5(d). Prussian blue is a dark blue pigment with the idealized formula ${\mathrm{Fe}}_7{(\mathrm{CN})}_{18}$ and it is not soluble in water. The Prussian blue reaction involves treating sections with acid solutions of ferrocyanides. If any ferric ions are present in the tissue, they combine with the ferrocyanides, forming a bright blue stain. A negative staining response is observable in Figs. 5(c) and 5(d), stemming from the absence of iron particles.

Fig. 5

Histological images of the mouse skin tissue with and without nanoparticles injection. (a) H&E-treated melanoma tissue without nanoparticles, (b) a magnified view of the dashed rectangular region in (a), (c) Prussian-blue-treated melanoma tissue without nanoparticles, (d) a magnified view of the dashed rectangular region in (c), (e) H&E-treated melanoma tissue with nanoparticles injected into it, (f) a magnified view of the dashed rectangular region in (e), (g) Prussian-blue-treated melanoma tissue with nanoparticles injected into it, and (h) a magnified view of the dashed rectangular region in (g).

We repeated the above experiment under the same experimental conditions for the melanoma tissue sample into which nanoparticles were injected, for confirming the molecular interaction between the melanoma cells and the injected magnetic nanoparticles. The dashed ellipses in Figs. 5(e) and 5(f) indicate blood vessels, which were identified after applying the H&E reagent. The purplish blue color around the blood vessels indicates the presence of live melanoma cells. The brownish regions surrounding the melanoma cells represent the injected nanoparticles.

In Figs. 5(g) and 5(h), the melanoma tissue regions are surrounded by dark blue boundaries as a result of the reaction between the SPIO magnetic nanoparticles and the Prussian blue reagent. Therefore, Figs. 5(g) and 5(h) confirm the molecular interaction between the melanoma cells and the magnetic nanoparticles, manifested as an obvious variation of color, because it verifies the applicability of MM-ODT signals to detecting particular melanoma regions.

3.3.

In Vivo Magneto-Motive Optical Doppler Tomography Imaging of Melanoma Tissue

To gain a better understanding of the process, we performed in vivo imaging of a mouse skin that did not contain nanoparticles. As shown in Fig. 6(a), red-colored MM-ODT signals were obtained with a noise region as a result of the animal body motion during breathing. Second, nanoparticles (concentration of $15\mathrm{mg}/\mathrm{ml}$) were injected into the mouse skin for characterizing the necessity of the contrast agents. Figure 6(b) shows the results of this experiment. Although the Doppler signals owing to the motion of the injected nanoparticles were identified (magenta color), the noise region was observed as well. Limited frame acquisition speed, frequency mismatch, and motion of the mouse body during breathing can be considered as the main causes underlying the appearance of this noise.

Fig. 6

MM-OCT and MM-ODT images of in vivo melanoma tissue. (a) An MM-ODT 2-D image of melanoma tissue without nanoparticles, (b) an MM-ODT 2-D image of melanoma tissue with nanoparticles injected, (c) an MM-OCT M-mode image of melanoma tissue with injected nanoparticles showing the Doppler shift variation for the nanoparticles, and (d) an MM-ODT M-mode image of melanoma tissue with injected nanoparticles showing the color change in the Doppler signal. White vertical bars: $500\mu \mathrm{m}$ and white horizontal bars: 20 ms.

The effect of these limitations was removed by performing the M-mode analysis. Doppler shift variations associated with the injected nanoparticles were observed in the MM-OCT image [Fig. 6(c)]. Subsequently, a dramatic reduction in the intensity of noise signals and optimal Doppler signals (blue-colored Doppler signals) in response to the periodically applied magnetic force can be identified in Fig. 6(d) throughout the area of $780\times 500\text{pixels}$ in the axial and temporal dimensions, respectively. This result is of sufficient quality to allow us to conclude that the MM-ODT M-mode images yield more detailed information on the Doppler shift compared with the B-mode images. Therefore, MM-ODT M-mode images are more versatile for detecting melanoma tissue.

As before, we measured the average Doppler velocity for different SPIO nanoparticle concentrations and different values of the induced solenoid voltage. The concentrations that we considered were 0, 5, 10, and $15\mathrm{mg}/\mathrm{ml}$. A volume of $100\mu \mathrm{l}$ from each diluted solution was injected into the melanoma lesion and held for 20 min before acquiring the corresponding MM-ODT image. Due to the volume of the melanoma tissue region ($9.05\mathrm{mm}\times 4.23\mathrm{mm}\times 2.00\mathrm{mm}$), $100\mu \mathrm{l}$ of SPIO nanoparticles solution had to be injected to obtain a sufficient distribution of nanoparticles homogeneously even though the tissue swelled slightly. The dependence of the average Doppler velocity on the induced solenoid voltage was examined by increasing the voltage from 11.25 to 90 $V_{p-p}$, in steps of 11.25 V. In these experiments, the concentration of the injected SPIO nanoparticles was $15\mathrm{mg}/\mathrm{ml}$, and the injected volume was $100\mu \mathrm{l}$. Each concentration and voltage experiment was repeated five times, and the average values are plotted in Figs. 7(a) and 7(b). Similar to the previous demonstration, these results confirmed that the Doppler velocity can be increased by increasing both the applied solenoid voltage and the SPIO nanoparticles concentration.

Fig. 7

Average Doppler velocity versus (a) the SPIO nanoparticles concentration and (b) the applied voltage. LR, linear regression. The error bars represent the fluctuation between the minimum and the maximum values.