An experimental model to investigate the targeting accuracy of MR-guided focused ultrasound ablation in liver

High Intensity Focused Ultrasound (HIFU) is fundamentally a propagating wave capable of producing localized thermal lesions [1‐3]. A large variety of US transducers, either single element or phased array, with continuous or pulsed-mode [4], have been described [5‐7] for this non-invasive approach, proving the growing interest in the scientific and medical communities.

The highest temperature elevation induced by the HIFU beam is expected to occur around the focal point, but every penetrated tissue in the pathway of the wave will also be heated at various levels depending on acoustic and physiological properties and on its relative position from the transducer. Cavitation, boiling or an increased absorption coefficient in the tissue may induce a thermal drift of the ablative lesion towards the transducer [8]. Far field of the transducer can also be affected by thermal heating, especially if a reflective interface is present [9]. Refraction on pre-focal interfaces [10], spatial misregistration of the device or geometrical distortion of MR guidance images may also yield lateral offsets of the effective position of the focal point.

Magnetic resonance guided HIFU (MRgHIFU), in particular using near real time Proton Resonance Frequency Shift (PRFS) thermometry (MRT), is at present the only FDA approved HIFU technology [11]. This approach has been clinically investigated to treat malignant tumors using prototype or off-label devices originally dedicated to uterine fibroid ablation [12‐20].

Theoretically, there are no limitations on the lesion size treatable by MRgHIFU [21]. Sequential MRgHIFU elementary sonications were evaluated clinically [20], indicating the technical feasibility despite the long treatment times [12‐22]. Volumetric MRgHIFU ablation has also been investigated, demonstrating that a large and uniform ablation zone can be rapidly obtained using continuous sonication and mechanical displacement of the transducer [23, 24], interleaved electronic and mechanical displacement of the focus [25] or fully electronic steering of the beam [6]. Moreover, studies using ultrasound guidance proposed different methods to increase the lesion size, such as: inertial cavitation that significantly enhanced the heat deposition at the HIFU focus [26, 27], focus splitting for larger coagulations regions [28] or using a multifocal thoroidal transducer that produced composite lesions [29, 30].

Effective spatial control of an induced thermal lesion during rapid volumetric ablation should be considered a major safety issue [9]. Together with the capacity to create large, rapidly achieved lesions, any volumetric sonication paradigm must also guarantee accurate spatial control of the HIFU ablative lesion, with predictable position of the ‘gravity’ center of the ablative region. To our best knowledge, available reports are missing providing experimental data to support that volumetric thermal ablations are indeed centered on a specific predefined target in 3D, i.e. a histological “gold standard” [31].

Different studies describing ballistic markers for HIFU ablations have already been published. Melodelima et al[32, 33] proposed a tumor-mimicking model made by injecting a warm solution at 65°C that polymerizes in hepatic tissue and forms a typically 1 cm size discrete lesion that was detectable by ultrasound imaging and gross pathology. That model was a large unperfused 1 cm hyperechoic lesion on sonograms (US reflecting) and the polymerized solution had attenuation coefficient of 0.39 dB.cm⁻¹ at 1 MHz, as compared to 0.68 dB cm^-1 for liver parenchyma [34]. Recently, Pichardo et al[35] proposed using the same agarose-based solution for MR targeting and thermal monitoring during HIFU sonication on living static tissue (rabbit muscle). The model applied to static muscular tissue in rabbits induced significant edema probably due to intrinsic temperature of the injected solution, while the animals were sacrificed immediately post-treatment. Therefore, the final dimensions of the lesion and its spatial relation to the ballistic model were not assessed histologically.

In order to improve and/or to assess spatial control of ablation, we present here an experimental method to create a user-defined ballistic target that can be used as an absolute reference marker prior to any HIFU trials. The marker must be irreversible, MR compatible and detectable both in pre-operative MRI and in post-mortem histology and has to enable the evaluation of the spatial control achieved during MR-guided HIFU ablation of liver lesions, in particular concerning the lesion’s ‘center of gravity’. At the end of the procedure, the contrast agent/stain must be stably encapsulated in the intercellular space, where perfusion, diffusion and/or metabolism are limited or suppressed.

The optimal parameters for marker fixation and MR detection were determined to be 2 W RF power during 30 sec, resulting in 60 J energy deposition, with a 18 G electrode, and 4 W RF power during 15 sec, also resulting in 60 J energy deposition, with a 16 G electrode. A higher power level was in general suboptimal as it triggered impedance cutoff and subsequently smaller lesions. Figure 1d shows T1-weighted MR images obtained ex vivo after monopolar RF ablation using the home-built electrode, with and without local injection of the mixture of methylene blue and Gadolinium chelate. The marker created by localized energy deposition and fixation of the contrast agent is clearly visible under MR and can be conveniently used to pre-define the zone to be ablated by HIFU, centered on the so-created target. The size of the histological marker ranged between 2.5 and 4 mm (Figure 1e). To illustrate the marker’s usefulness in assessing the spatial accuracy of the thermal lesion, we present in Figure 1f-h three independent examples of ex vivo thermal lesions obtained by volumetric MRgHIFU (see the whitened ellipsoidal area inside the fresh ex vivo sample), centered on the ballistic target (high intensity blue spot) within 2 mm spatial accuracy. Some spreading of the methylene blue staining along the muscle fibers was unavoidable when slicing the fresh meat across the central plane of the lesion. However, the color contrast between the native marker and the spread staining is evident.

During the in vivo interventional procedure of marker fixation, the ultrasonography clearly visualized the tip of the RF electrode before ablation (Figure 2a), and the boiling cavitation during RF ablation (Figure 2b,c). The precise position of the induced ablation could be chosen before RF coagulation, based on initial images where the RF electrode tip is visible. This confirmation in real time is very useful, making it possible to change the needle’s position to the preferred point and then fixate the marker. The occurrence of the “white cavitation” cloud of bubbles is a direct indication that RF ablation was effective. The “wash-out” of the cloud bubbles required approximately one minute after the end-point of RF coagulation, as seen in Figure 2d. After the bubbles wash out, no ultrasound indication of the marker’s presence could be observed, with standard (retro-diffused, or back-scattered) ultrasound imaging performed in the frequency range 2-4 MHz, although the experiments were conducted in optimal conditions ex vivo where the marker was positioned a few centimeters only from the imaging transducer. The imaging wavelength at 4 MHz being approximately 10 times smaller than the marker itself, missing of any marker signature cannot be attributed to acoustic diffraction. Therefore this is interpreted as no significant change in the acoustic impedance in tissue because of the marker, within the experimental resolution. This evidence is not covering the thermal effects involving the absorption of the HIFU beam on the marker, therefore specific MRgHIFU experiments were conducted, as described in Appendix.

Figure 3 displays in vivo MR T1-weighted images pre- and 7-days post-marker fixation and HIFU therapy in static tissue, in rabbit thigh. The ballistic target is clearly seen prior to the HIFU treatment in each section through the 3D reconstruction (a-c). The apoptotic and necrotic area formed 7 days after treatment is visible as a hypo intense signal zone surrounded by a hyper intense rim (d-f) and corresponds to the planned HIFU ablation. Methylene blue marker trapped in tissue by thermo coagulation was clearly detected in different animals after either 3 weeks (Figure 3h) or 8 months (Figure 3g) of post-mortem formalin fixation. Frame g is showing an eccentric lesion with respect to the ballistic marker as a result of per-operatory HIFU-induced muscle contraction, detected on line during the treatment (data not shown). Frame h is illustrating a thermal lesion precisely centered on the ballistic marker when intra-operatory tissue motion did not occur. The size of the final HIFU ablations (mean values measured on 4 rabbits), measured at Day 7, after Gd injection, were: in plane long axis = 30 mm, short axis = 21 mm and height (parallel to the HIFU beam) = 45 mm.

The ballistic target model was easily detectable in vivo in pig liver, as a strong hyper-signal in T1w 3D VIBE images (Figure 4) immediately after fixation, due to the trapped Gd-based contrast agent. The small volume of RF coagulated tissue and the trapped tracers induced no measurable magnetic perturbation in the phase maps acquired with the PRFS temperature-sensitive GRE-EPI sequence during HIFU sonication. This behavior enables the direct use of reference-free PRFS-calculated temperature maps [46, 47] without any correction specific to the marker. Seven days after the marker procedure, the thermal lesions induced by RF ablation could be identified in freshly cut liver samples (Figure 5a,b). Macroscopic slices indicated a blue-to-green colored volume of coagulated tissue (frame a) for the methylene blue stained ballistic marker and a whiter zone (frame b) for a case without methylene blue. No methylene blue was detected outside the RF-ablated tissue, indicating normal clearance of the un-trapped staining. Post-mortem macroscopic illustration (frames c and d) and microscopic histology (frame e, same lesion as shown in frame d) of the ballistic marker (inside the lobe – c, and at the periphery of the lobe - d and e) after 3 weeks of formalin fixation clearly indicates the ablated zone surrounded by healthy tissue. Histopathology indicated a complete disorganization of hepatocytes and the absence of red cells due to the sinusoids destruction in the center of the marker. More peripherically, intra-cellular edema was found. The lesion was sharply delineated by a few hundred microns thick fibro-inflammatory rim. The rest of the hepatic parenchyma had a normal aspect, with clear delineation of lobuli and no visible side effects due to the interventional procedure. Please note that no HIFU sonication was performed in the examples illustrated in Figures 4 and 5. For in vivo studies, the marker was imaged (T1w VIBE MR sequence) in situ at 10 and 80 minutes after RF procedure, and no alteration was detected between these 2 time points.

An in vivo ballistic marker followed by HIFU ablation in pig liver is shown in Figure 6. Temperature (a-c) and thermal dose maps (d-f) in three different planes indicate the thermal build-up accumulation at the end of the HIFU sonication. Seven days after the treatment, after Gd administration, the final lesion (g-i) could be easily identified in T1w VIBE images. Moreover, the lesion was also identified in post-mortem macroscopic (j) and microscopic analysis (l), having a similar ablation shape. In frame j the two ablated regions are easily distinguished: the blue arrow indicates the RF coagulated marker (identified by the presence of methylene blue), positioned peripheral to the HIFU ablation, delimited by a dotted red line. Microscopic histology (frame l) revealed a clear difference between the ablated zone and healthy tissue, with normal cells showing clear striation by hematoxilin-eosin staining, as well as peripheral nuclei, while areas of coagulation necrosis appeared with lighter staining and showing a complete loss of lobular structure. A 3D reconstruction (Osirix Software) of the hepatic lobe containing the final lesion (same as in frame j) is shown in frame k, indicating the HIFU lesion and the RF marker at the periphery. The shortest distance between the centers of the two lesions was 5 mm. The calculated volume of the HIFU lesion was 1.935 cm³, while the volume of the created marker was 0.0235 cm³ in the presented case.

We describe here a new method to create a user-defined ballistic target that can be used as an absolute reference marker, based on trapping a mixture of gadolinium chelate and methylene blue (hystology stain) and using sharply localized RF energy deposition, causing immediate cell death and coagulation. The gadolinium-based contrast agent was easily detectable by pre-operative MR imaging; therefore, this method provided a minimally invasive and robust ballistic target, mimicking a small ‘virtual’ tumor, and can be conveniently used as a target for MRgHIFU ablations. According to pathologists, the methylene-blue does not affect the microscopic histology staining, while it is trapped within the coagulated tissue. The large majority of pre-clinical and clinical studies using MR-guided HIFU technology are reporting temperature elevation below 70°C, since approaching the boiling regime is likely to produce local disruption of the magnetic susceptibility [9, 43] and thus errors in PRFS MR thermometry. A sharp RF coagulation is required whenever the HIFU treatment does not result in immediate necrosis, because the HIFU-induced apoptosis in the treated area does not prevent wash out of the staining until all metabolism ceases. We demonstrated that methylene blue was trapped by immediate coagulation of the tissue using localized RF energy deposition, and that any un-trapped staining was cleared all around, including from inside the HIFU-ablated tissue. Staining was detectable immediately after animal sacrifice (see Figure 5), after 3 weeks of formalin fixation (see Figure 6) and also after eight months of formalin fixation (see Figure 3). This demonstrates a highly stable histology marker, compatible with semi-chronic follow up protocols, where the living tissue may deform during the time the animal is allowed to survive after sonication. No local or systemic complications related to the marker were noticed in 10 animals during the survey period of 7 days. The minimally invasive interventional procedure observing sterile conditions should be at low risk.

During RF ablation, boiling bubbles in the tissue were easily detected with ultrasonography, during approximately one minute after the end-point of RF ablation in pig liver (see Figure 2). Beyond this transient period, the created target did not modify the local echogenicity of tissue in the range 1.57 - 4.4 MHz, nor the local magnetic susceptibility (neutral versus PRFS thermometry) and induced no observable side effects in tissue. The marker involves tissue coagulation and injection of a small volume of stable liquid state solutions, while its size is comparable to the wavelength of HIFU. These conditions yield the minimal acoustic perturbation of HIFU beam – if any – by a marker that is conveniently visualized with conventional resolution T1-weigthed MR sequences. If no HIFU ablation of tissues surrounding the marker is performed, the small coagulative volume of the marker itself is very likely to be reabsorbed by the tissue in the long term. Due to real time US visualization of the electrode tip while the operator manipulates the needle towards the investigated zone, a major advantage of this ballistic marker is the ability to choose the precise position of the induced ablation. If the needle is not at the right place, the operator can alter the position until the optimal position is reached before delivering the RF energy. Another advantage of this model can be considered the possibility of creating multiple markers in the same organ, due to the time effective procedure and to their point-like aspect. The RF electrode can be repositioned in the organ, or removed completely, changed and inserted again in the body using another entry point. The effectiveness of ablation is confirmed with US detection of white cavitation, while staining is confirmed by MR imaging of trapped Gd from the same mixture.

Correct evaluation of what tissue is ablated by HIFU sonication and precise determination of HIFU-induced lesion position versus the treatment planning cannot be done if the animal is sacrificed immediately post-treatment. Therefore, for our study, the animals were followed up for 7 days. Histological analyses both macro- and microscopic clearly distinguished the two types of lesions (ballistic marker and HIFU ablation). This can indicate, post-treatment, whether the HIFU lesion was centered on the target or not. Moreover, various prefocal thermo-acoustic effects (like thermal drift induced by cavitation or differential absorption in the traversed tissues) may be evidenced with this approach.

Accurate targeting in abdominal organs during breathing is a challenging problem. Unlike previous reports dealing with static conditions during the sonication [32, 35], our sonications in pig liver were performed under continuous breathing and gating. Internal organ drift between targeting and sonication steps [48], acoustic aberration of the HIFU beam versus the theoretical beam forming in homogeneous tissue and geometric distortion of MR images sequence-specific are inherent limits to the achievable accuracy for centering the lesions on the T1-contrast marker. Here, a larger population for in vivo quantitative assessment of the targeting accuracy with the available MRgHIFU device was not defined as a goal.

The presented method demonstrated no risk of tissue contamination with air bubbles and no irregular heating pattern was observed on PRFS temperature maps during HIFU sonication, as opposite to [35]. Importantly, the needle and the syringe were mounted together and filled with liquid solution prior to the incision. The interventional procedure to fix the marker here was safe and successfully performed in all cases within approximately 15 minutes. No specific training was requested for the interventional radiologist. The very low volume of injected liquid was delivered at room temperature and no physical or chemical transformation will occur spontaneously after some critical period of time. Thus, there is no constrained timing for executing the interventional procedure. This is particularly important to allow the interventionist sufficient time to accurately perform the image guided positioning of the electrode. Tissue effraction is minimal and direct assessment of the ballistic accuracy of millimeter-range (i.e. elementary) HIFU lesions aiming to overlap the marker location is meaningful due to the small size of the marker itself, unlike the previous model [32, 35].

There are different applications where the presented marker model may be suitable. The presented ballistic model is expected to be a convenient procedure to validate the accuracy of HIFU ablation in any organ, static or mobile. Beside tissue-dependent or device-dependent targeting errors, the entire problem of motion tracking errors in moving targets can further tackled.

Other applications of our affixed marker may involve minimally invasive tissue ‘tattooing’ as landmarks for diseased tissue/tumor or healthy structures at risk to be delineated for surgery, for instance “missing metastasis”. Current preoperative marking techniques to guide radiotherapy or surgery include guide wire locators (with the associated risks of migration, transection and scheduling conflicts), radioactive injections [49, 50] and charcoal suspensions [51], which are invasive. Radiotherapy is often guided using external tattoos and internal clips in tumor beds or at biopsy sites. Substituting the methylene blue staining with a radio-opaque agent (e.g. iodine-based) may render the marker visible to X-Ray techniques.

HIFU is fundamentally a propagating wave with a limited precision of a-priori prediction of the focus position in tissue during sonication. We presented here an experimental method to investigate the targeting accuracy of MR-guided focused ultrasound ablation, by fixing a ballistic target detectable with both pre-operative MR and post-mortem histology. The marker’s ultrasound conspicuity was below the detection threshold on a clinical ultrasound imager. The marker was established irreversibly, induced no changes in local magnetic susceptibility and its size was comparable to the HIFU wavelength. It showed no side effects in tissue and can be considered as a long-term stable histology staining method.