Interstitial magnetic thermotherapy dosimetry based on shear wave magnetomotive optical coherence elastography.

While magnetic thermoseeds are often utilized in interstitial magnetic thermotherapy (iMT) to enable localized tumor ablation, we propose to extend their use as the perturbative source in magnetomotive optical coherence elastography (MM-OCE) so that the heat-induced elasticity alterations can be 'theranostically' probed. MM-OCE measurements were found to agree with indentation results. Tissue stiffening was visualized on iMT-treated porcine liver and canine soft tissue sarcoma specimens, where histology confirmed thermal damages. Additionally, the elasticity was found to increase exponentially and linearly with the conventional thermal dosage metrics and the deposited thermal energy, respectively. Collectively, a physiologically-meaningful, MM-OCE-based iMT dosimetry is feasible.

systems, iMT has the advantage of inducing heat remotely and wirelessly. Clinical trials of iMT for oncology have previously been performed on prostate and oral cancers [3][4][5].
In any kind of cancer treatment, the goal is to provide the maximum amount of damage to the malignant tissues while sparing or minimizing the collateral injury to the surrounding healthy tissues. Often, the maximum therapeutic dose that is allowed is limited by the damage to the healthy tissues [9]. Therefore, an accurate evaluation of the thermal dosage is important. Traditionally, the thermal dosage applied to the tissue has often been estimated via the cumulative equivalent minutes at 43 C° (CEM43). This metric converts different temperature-time history during thermotherapy to a biological isoeffect at 43 C°, an empirically determined "break" point where the survival rates of various cell lines transition from one constant to another [10,11]. However, CEM43 does not fully account for factors such as the thermotolerance effect (which can be an intrinsic biological property, or developed by step-down heating, multiple dose delivery, or exposure to radiation) and the influence of high-temperatures (> 47 C°) that are often seen in ablation [10]. In addition, the assessment of the cross-sectional temperature profile is often challenging [12].
Interestingly, biomechanical properties, such as the viscoelasticity of the tissues, can also be altered because protein denaturation and coagulation can be induced along with temperature elevation [12][13][14][15]. This suggests that biomechanical properties could be an alternative physiologically-meaningful indicator of the thermal dosage applied in thermotherapy. Elastography is a technique that probes and/or maps the mechanical characteristics of the tissue with biomedical imaging modalities such as ultrasound, magnetic resonance imaging (MRI), or optical coherence tomography (OCT). Among these clinically available imaging modalities, the finest cross-sectional structural details (μm-scale) and axial displacement sensitivity (nm-scale) are provided by OCT [16][17][18][19]. Tissue elasticity can be described by the longitudinal stress-to-strain ratio, or the Young's modulus, meaning that the stiffness of the sample can be inferred by exerting a certain force and measuring the induced displacement. In optical coherence elastography (OCE), subtle tissue displacements are commonly probed by phase-sensitive OCT, and further spatial and/or temporal analysis of the displacements can allow for elasticity measurements (through quasi-static and/or dynamic OCE) [17].
Magnetomotive optical coherence elastography (MM-OCE) is a branch of dynamic OCE that vibrates tissue with an oscillating magnetic force and probes the temporal tissue displacements for further mechanical assessment [20][21][22][23][24][25]. Magnetic nanoparticles (MNPs), acting as internal mechanical perturbative sources, can be delivered to the tissues uniformly and induce motions when exposed to an external magnetic field gradient. Subsequently, either the natural or resonant frequency can be extracted from the induced "magnetomotion" and be utilized to infer the elasticity of the homogeneous or heterogenous tissues [20][21][22][23][24]. In addition, by confining the MNPs to a small cylindrical region at high concentration, a more localized magnetic force can be exerted to induce shear waves, where the wave propagation velocity can be utilized for stiffness quantification [25]. However, the distribution of MNPs in tissues is often hard to control, let alone confining the MNPs into a cylindrical region. Moreover, localized magnetic force was less achievable with the large dimensions of a formed MNP mold (Ø ~3 mm) [25]. In contrast, readily accessible magnetic thermoseeds for iMT have well-defined geometries (typically rod-shaped) and are directly inserted at the tumor site. The small dimensions of thermoseeds (Ø ~1 mm [2,[4][5][6]) can also allow for a highly localized mechanical excitation. Therefore, we aim to examine the feasibility of performing shear wave MM-OCE with a rod-shaped magnetic thermoseed serving as an internal, localized perturbative source. Subsequently, the capability of monitoring iMTinduced stiffness changes is further investigated on both normal porcine liver and cancerous soft tissue sarcoma (STS) specimens. Clinically, hepatic tumors are one of the most common diseases amenable for curative ablation treatments [26] and thermotherapy has been reported to improve treatment outcomes for STS [27].
Using a single platform for both therapeutic and diagnostic applications allows for theranostic function. For elastography-based thermotherapy monitoring applications, ultrasound elastography (UE) has been coupled with high-intensity focused ultrasound to assess the alteration of tissue stiffness after treatment on excised chicken breast [28] and in vivo human prostate tumors [29]. Additionally, RFA probes have been utilized to both ablate liver tissues and serve as a compressor to allow for ultrasound strain imaging [30]. A mechanical actuator can also be attached to an RFA electrode or microwave antenna to induce shear waves for mapping the elasticity of the treated tissue [31]. Using OCE, our group has previously proposed and demonstrated the use of MNPs for both magnetic fluid hyperthermia and MM-OCE, where a significant shift in resonant frequency was observed in the mechanical spectrum of the treated tissue specimen, qualitatively indicating tissue stiffening [24]. Here, the theranostic function is implemented with a thin embedded magnetic rod, which can be utilized in both iMT as a highly localized heating source (therapeutic), and in shear wave MM-OCE as an internal vibrating source for quantitative elastic measurements (dosimetric/diagnostic).

Sample preparation and histology
Tissue-mimicking phantoms with different elasticity were prepared by mixing polydimethylsiloxane (PDMS (Clearco Products, Pennsylvania)) with curing agents RTVA and RTVB (Momentive, New York) with a mass ratio of PDMS:RTVA:RTVB ranging from 30:10:1 to 100:10:1 (stiff to soft). Subsequently, 1 mg/g of titanium dioxide (TiO2) was added to the PDMS solution as scatterers for OCT contrast prior to sonication (>1 hr). Afterward, the sonicated PDMS solution was poured into a Petri dish and heated for >6 hr at 80 C°. Finally, homogeneous PDMS phantoms (Ø ~34 mm, thickness ~8 mm) of various stiffness were produced. To simulate an ablated tissue, a heterogenous phantom was also created (a stiff cylindrical inclusion at the center surrounded by soft medium). A stiff PDMS (PDMS:RTVA:RTVB 30:10:1) was first made, trimmed into a thin cylinder (Ø ~7 mm), and placed at the center of a Petri dish. Subsequently, the surrounding area was filled with a soft PDMS solution (PDMS:RTVA:RTVB 100:10:1) to create a soft environment. The phantom was heated again for another >6 hr. All phantoms were separated from the Petri dish wall before imaging to minimize boundary effects.
Ex vivo porcine liver tissues were trimmed to rectangular cuboid shapes with dimensions (length × width × height) of ~28 × 28 × 8 mm 3 . In addition, one specimen of excised canine soft tissue sarcoma with dimensions (length × width × height) of ~17 × 9 × 4 mm 3 was also utilized. For both PDMS phantoms and biological tissues, a ~1 cm long carbon steel wire (Malin, Ohio) was inserted in depth at the center to serve as both the magnetic thermoseed for iMT and the vibration source for MM-OCE. The steel wire has a diameter of ~220 µm, which is less than that of a 25-gauge syringe (~510 µm), radioactive seed (~500-800 µm [32]), and RFA probe (~1.5 mm [31]).
After imaging, the tissues were fixed in neutral-buffered formalin for >6 hr before histological processing. The tissues were embedded in paraffin, sectioned at 5 μm using a microtome (Leica Microsystems, Germany), and stained with Masson Trichrome (Sigma-Aldrich, Missouri) for collagen visualization.

iMT system setup
The iMT system was built by integrating a pulse generator, heat inductor, water chiller, and a magnetic coil (inner Ø ~3 cm, 13 turns over 10 cm length) on a portable cart ( Fig. 1(c)). The tissue specimen was placed within the coil, where the long-axis of the magnetic thermoseed (rod) was situated in parallel to the induced magnetic field ( Fig. 1(b), 1(d)). To heat the porcine liver specimens, the magnetic field strength ranged from 18.1 -36.0 kA/m (and 0 kA/m for control), oscillated at 64.7 kHz for 4 min during one iMT treatment. In addition, a canine soft tissue sarcoma (STS) specimen underwent two iMT treatments, both with a magnetic field strength of 25.9 kA/m. The first and second iMT were administered for 2 and 4 min , respectively.

Temperature monitoring and CEM43 quantification
A thermal camera (FLIR, Oregon) was utilized to monitor the temperature change distributed across the tissue surface during iMT once every minute. Subsequently, the conventional thermal dosage estimation was performed based on the cumulative equivalent minute at 43 C° (CEM43), determined by [10,11]: Here t is the time instant from 0 to f t , t ∆ and T are the temporal interval (1 min) and the average temperature between two successive time points, respectively, and R is the empirically determined constant related to the cell killing rate. Note that the measurable temperature was limited to the sample surface.

Shear wave MM-OCE system setup and data acquisition
An OCE system requires a mechanical excitation source to delicately vibrate the tissue, and an OCT imaging system to detect the responding tissue displacements. Here, an alternating magnetic force was exerted on the magnetic thermoseed inserted in the tissue via a customized magnetic solenoid and programmable power supply. The magnetic solenoid produced a magnetic field with a peak-to-peak strength of ~4.09 kA/m at 5 mm away from the sample. To produce a sinusoidal magnetic force, the driving voltage carrying a square root of sine waveform was sent to the power supply. The magnetomotive modulation frequency was kept low (~460 Hz) to prevent additional heating during MM-OCE measurement. The chamber containing the coil was continuously perfused with water as a cooling agent during magnetic excitation. The spectral-domain OCT imaging system consisted of a superluminenscent diode (Thorlabs, New Jersey) with a wavelength of ~1325 ± 50 nm as a light source, and a spectrometer with an InGaAs line-scan camera (Goodrich, North Carolina) at the detection arm. The axial and lateral resolution were ~8 μm and ~16 μm, respectively. The displacement sensitivity of the system at the line scan rate utilized (~92 kHz) was ~3 nm (optical distance). M-B mode data were acquired, where multiple M-scans (over ~22 ms for the liver samples and ~4 ms for the STS specimen) were collected at different lateral locations. Note that as the STS specimen was comparatively small (detailed in Section 2.1), less magnetomotive modulation cycles were provided to the STS specimen to reduce the excitation energy and hence minimize the boundary effect. The lateral pixel resolution was ~11 μm.

Shear wave MM-OCE data processing
The harmonic stimulating force that originated from the centered magnetic rod (which generated a magnetic field gradient predominantly in the axial direction) induced a cylindrical shear wave that propagated along the radial distance ( r ) through the sample, and vibrated in the axial direction with a modulated angular frequency ( ω ) over time ( t  [34,35]) and mass density ρ (~970 kg/m 3 for PDMS and ~1040 kg/m 3 for soft tissues). Note that although Rayleigh waves could possibly be generated during the magnetic excitation as well, we did not take its effect into account as the difference between Rayleigh and shear wave velocities was as small as 5% [36]. Similar to our previous study [25,37], standard OCT processing (i.e. background subtraction, k-linearization, dispersion correction, and Fourier transform) was performed on each M-scan, where axial displacement was extracted from the phase difference between adjacent M-scans (note that the phase derivative was calculated in the complex form, where median filters were applied to both the real and imaginary parts prior to taking the angles). Afterward, bandpass filtering at the modulated frequency ω was performed on the single- phase gradient and wave propagation velocity, s c , was only observed on the mechanically heterogeneous sample. Subsequently, a cross-sectional Young's modulus ( E ) map could be obtained, where thresholding was performed based on both the intensity and the goodness-offit of the phase gradient (R 2 > 0.8).

Tissue-mimicking phantoms
OCE maps of both homogeneous and heterogeneous PDMS phantoms are visualized in Fig. 2 , where a higher Young's modulus ( E ) value was observed on phantoms fabricated with a higher curing agent ratio. In addition, obvious mechanical contrast was seen from the heterogeneous sample. The average Young's moduli obtained from the two-dimensional OCE maps for each homogeneous phantom were compared to those from mechanical indentation testing with a spherical indenter (Ø ~2.5 mm), where each value of E was obtained by fitting the force-displacement curve to the Hertzian model. As seen in Fig. 3, the two measurements correlated well with each other (Pearson's r = 0.965, N = 16) and have an almost one-to-one correspondence (mean difference = 0.558 kPa), suggesting a comparable performance between the two approaches.

iMT-treated ex vivo porcine liver
Tissue elasticity of iMT-treated porcine liver specimens, before and after therapy, were imaged with the described MM-OCE method, where the magnetic thermoseed served as both the heating and mechanical perturbative source. Both shear wave propagation and the E maps of each liver specimen are visualized, where a significant increase in both the shear wave propagation velocity ( s c ) and value of E were only observed from the iMT-treated specimen (Fig. 4(III), 4(V), and Visualization 2), along with an obvious temperature elevation near the thermoseed (Fig. 4(I)). Interestingly, there is a gradual spatial change observed in the stiffness (i.e. higher values of E near the thermoseed) in Fig. 4(III). The corresponding Masson Trichrome-stained histological slices (Fig. 4(IV)) also showed an obvious 'burn mark' at the site near the heat source, where a loss of collagen (absence of blue staining) can be found. In contrast, the control tissue (field strength = 0 kA/m) did not show any changes in the surface temperature, s c , E map, or histology. Collectively, it is suggested that the spatially varying tissue stiffening was a result of iMT-generated thermal damage. The altered elasticity was quantified by the ' E ratio', which was obtained by normalizing the average Young's modulus after iMT over that before iMT (i.e. after iMT before iMT Comparing this E ratio with the corresponding CEM43 metrics, an exponential correlation (Pearson's r = 0.799, N = 27) was observed ( Fig. 5(a)). Interestingly, a similar correlation was previously reported in the literature, where the shear modulus ratio obtained after water bath heating of rat leg muscle was exponentially associated with CEM43 [15]. In contrast, the ratio of the average OCT intensity after and before iMT has a value close to unity for all data (OCT intensity ratio: 0.98 ± 0.02, N = 27), implying negligible variation between treated and untreated scenarios. Note that the E , CEM43, and OCT intensity values compared here were all averaged over the same area. In addition, our porcine liver results manifested as a sharp increase in the E ratio starting from CEM43 between 8.5 × 10 −2 -4.6 × 10 2 min, which was partially comparable to the thermal damage threshold reported in the literature. To date, the lowest CEM43 reported to cause thermal damage in canine livers was 9.9 min, while thermal coagulation was observed on porcine livers at CEM43 = 320 min [38,39]. Tissue necrosis was shown in rabbit and porcine livers at CEM43 between 2 × 10 5 -8.6 × 10 11 min [39], which overlaps with the CEM43 coverage that corresponds to higher E ratios in our data.
To investigate the relationship between the power ( P ) absorbed by the ferromagnetic thermoseed (rod) from the axially parallel magnetic field and the corresponding elasticity change induced, the values of the E ratio were further compared to the square of the applied field strength ( | P H ∝ , holds validity given our initial assumption that the geometrical and electromagnetic properties of the thermoseed remained constant during iMT [40]. The result (Fig. 5(b)) shows that the E ratio is linearly related to 2 0 | | H (Pearson's r = 0.938, N = 27), which physically relates the stiffening to the thermal energy generated.

Ex vivo canine soft tissue sarcoma (STS)
Other than normal tissues, the feasibility of evaluating iMT dosage based on shear wave MM-OCE was also performed on a canine STS specimen. Again, the shear wave propagation ( Fig. 6(V) and Visualization 3) and elasticity map (Fig. 6(II)) of the tissue was visualized before and after each iMT treatment. Surface temperature profiles were recorded during iMT ( Fig. 6 (III)) and Masson Trichrome staining was performed after the treatments (Fig.  6(IV)).
Similar to the porcine liver, STS tissues stiffened after iMT. Additionally, a monotonic increase in the average E value was observed as more doses were provided. In fact, a higher value of E (18.0, 44.2, and 140.9 kPa for before, after first iMT, and after second iMT treatments, respectively) corresponded to a higher CEM43 value (7.8 × 10 4 and 3.1 × 10 9 for the first and second iMT treatments, respectively). Unlike the porcine liver data, a higher mechanical heterogeneity was observed in the treated STS tissues, which potentially reflects the expected heterogeneity of the local anatomy or the thermal sensitivities across the tumor microenvironment. Nevertheless, an increase in local values of E was observed across the tissue. In addition, a generally lower collagen content was present near the heat source, similar to the observation in the liver specimens, suggesting thermal damage.

Conclusions
A magnetic-rod-based shear wave MM-OCE has been successfully developed and utilized to visualize and quantify the iMT-induced thermal damage at the tissue cross-section for both normal liver and soft tissue sarcoma (STS) specimens. Young's modulus ( E ) quantified with the proposed method correlated well with that from mechanical indentation tests, as demonstrated on PDMS phantoms. On iMT-treated tissue specimens, a significant increase in tissue elasticity was observed while the OCT intensity exhibited negligible changes. Furthermore, Young's modulus maps of the iMT-treated porcine liver and canine STS tissues exhibited patterns similar to their corresponding Masson Trichrome-stained histological sections, where there was an apparent loss of collagen structures (due to thermal coagulation) near the localized heating sites. For the liver specimens, changes in elasticity were found to be exponentially related to the conventional temperature-based thermal damage metrics (CEM43) and linearly proportional to the power absorption of the ferromagnetic thermoseed. Collectively, the results suggest that the elasticity alteration probed by shear wave MM-OCE could be exploited to serve as a physically and physiologically meaningful thermal dosage metric for iMT.
OCT-based assessment of thermal damage in biological tissues has been widely investigated, where heat-induced alterations can be inferred from tissue morphology, backscattering, and/or birefringence [41][42][43]. While OCE has previously demonstrated its capability of detecting thermally-altered biomechanical properties and dynamics for cartilages undergoing water-bath heating [44], little has been explored for cancer treatment monitoring on soft tissues. MM-OCE, as demonstrated here, has not only provided an opportunity for elasticity-based thermal dosage evaluation, but has also highlighted the theranostic function of the magnetic thermoseed, which allows for both therapeutic and dosimetric functions. Expanding from this study, it would be interesting to examine the performance of MM-OCEbased iMT dosimetry using a temperature-regulated magnetic thermoseed, where the heating efficiency diminishes above the Curie temperature of the ferromagnetic alloy [4,5]. Changes of the biomechanical tissue properties in the nonlinear elastic regime, or in more anisotropic tissues, can be investigated as well, where the latter is permitted by the inherent rod-shaped geometry of the thermoseed [45]. Additionally, theranostic functions can hypothetically be enabled with OCE in other types of thermotherapy. For instance, laser ablation dosage can potentially be assessed with laser-induced elastic wave characteristics or deformation amplitudes [46,47]; ultrasound-ablated tissues can likely be evaluated with acoustic radiation force OCE [48].
In clinical practice, ultrasound, computed tomography (CT), and MRI are currently utilized to guide the placement of RFA probes and evaluate ablation outcomes [49]. With previously demonstrated intraoperative imaging capabilities [50], an OCT-assisted clinical environment is also foreseen, where handheld or catheter-based OCT can be utilized to both detect the local tumor microstructure and/or image the altered mechanical response of tissue through morphological alteration [50][51][52]. By incorporating magnetomotive components into a handheld or catheter-based OCT platform, in situ MM-OCE-based thermal dosage evaluation may be feasible in the future, serving as a physiologically-meaningful and complementary approach for real-time thermal treatment dosimetry.