Tumour hyperthermia and ablation in rats using a clinical MR‐HIFU system equipped with a dedicated small animal set‐up

Abstract

Purpose: We report on the design, performance, and specifications of a dedicated set‐up for the treatment of rats on a clinical magnetic resonance high intensity focused ultrasound (MR‐HIFU) system.

Materials and methods: The small animal HIFU‐compatible 4‐channel MR receiver volume coil and animal support were designed as add‐on to a clinical 3T Philips Sonalleve MR‐HIFU system. Prolonged hyperthermia (T ≈ 42°C, 15 min) and thermal ablation (T = 65°C) was performed in vivo on subcutaneous rat tumours using 1.44 MHz acoustic frequency. The direct treatment effect was assessed with T₂‐weighted imaging and dynamic contrast enhanced (DCE‐) MRI as well as histology.

Results: The developed HIFU‐compatible coil provided an image quality that was comparable to conventional small animal volume coils (i.e. without acoustic window), and a SNR increase by a factor of 10 as compared to the coil set‐up used for clinical MR‐HIFU therapy. The use of an animal support minimised far field heating and allowed precise regulation of the animal body core temperature, which varied <1°C during treatment.

Conclusions: The results demonstrated that, by using a designated set‐up, both controlled hyperthermia and thermal ablation treatment of malignant tumours in rodents can be performed on a clinical MR‐HIFU system. This approach provides all the advantages of clinical MR‐HIFU, such as volumetric heating, temperature feedback control and a clinical software interface for use in rodent treatment. The use of a clinical system moreover facilitates a rapid translation of the developed protocols into the clinic.

• Ablation
• binary feedback control
• hyperthermia
• small animal HIFU
• volumetric heating

Introduction

Magnetic resonance guided high intensity focused ultrasound (MR‐HIFU) is a promising technique for controlled non‐invasive tumour treatment by selective tissue heating [1], [2]. By focusing the mechanical ultrasound waves, ultrasonic energy can be deposited deep within the body without harming surrounding tissue, which is not possible with other thermal therapy methods (e.g. laser, radiofrequency, microwave or cryothermal). Clinical application of HIFU has been accelerated with the introduction of image guidance, especially MR imaging [3]. The latter offers excellent soft tissue contrast and spatial resolution that aids in planning the HIFU treatment and also evaluating the treatment outcome [4]. More importantly, MR has the ability to map the local temperature distribution within the targeted location by dynamic acquisition of MR phase images [5], [6], which can serve as feedback to the ultrasound control unit. Using MR‐controlled ultrasonic heating, the local temperature can be slightly increased and maintained at a level of hyperthermia (T = 41°–47°C), or the temperature can be raised further, leading to cell death (thermal ablation) [7–9]. The thermal dose at which cell death occurs varies for different types of tissue [10–12].

Tumour thermal ablation with MR‐HIFU is currently being introduced to the clinic to treat both benign (e.g. uterine fibroids [13], [14]) and malignant tumours (e.g. breast [15], [16]). In contrast to benign tumours, the clinical treatment of malignant tumours demands complete coverage. There, treatment with high spatial and thermal precision is more of a critical issue as malignant tumour border zones need precise treatment to prevent local recurrence and to minimise damage to surrounding healthy tissues [17]. To achieve complete coverage of large tumours, volumetric ultrasonic heating combined with multiplane MR‐based temperature mapping allows for depositing a well‐defined thermal dose within a predefined treatment volume [18]. Hyperthermia‐induced drug delivery is another promising application that can benefit from the temperature feedback made possible by the MR‐HIFU technology. For this application, a tumour needs to be heated to a constant temperature of around T = 42°C for a prolonged time period [19], [20].

The clinical translation of malignant tumour ablation and hyperthermia‐induced drug delivery using MR‐HIFU requires thorough pre‐clinical testing in large cohorts together with therapy monitoring methods. For this, studies performed with small animal models such as mice and rats are preferred over larger animals (e.g. rabbits, pigs) for both ethical and practical reasons, as well as the extensive knowledge and availability of antibodies and tumour models in rodents, for example xenografts in immune compromised strains. Rabbit tumour therapy studies are limited by the restricted choice in well characterised transplantable tumour models (currently VX2 only [21], [22]). Furthermore, the natural body temperature of rabbits is slightly higher than the human body temperature (39°C as compared to 37°C), which, although used [23], [24], does not make them ideal candidates for use in mild hyperthermia studies. Finally, therapeutic outcome can be evaluated more easily in small animals using standard small animal imaging equipment for positron emission tomography (PET) or single photon emission computed tomography (SPECT) studies as compared to similar evaluations in larger animals.

Most focused ultrasound studies in rodents are performed using manual positioning systems, which limits the type and amount of exposures [25], [26], and/or are performed under ultrasound guidance [27], [28], which lacks an accurate temperature feedback. Chopra et al. [29] were the first to present a MR‐compatible prototype system for ultrasound treatments in rodents under MR guidance with a piezoelectric 3‐axis positioning system. Meanwhile, more systems have been developed including phased array systems in high field small bore animal scanners [30].

In contrast to studies in rodents, MR‐HIFU studies in pigs and rabbits have the advantage that a clinical system can be used for therapy. Most clinical MR‐HIFU systems provide automated transducer positioning with a phased‐array transducer and MR‐temperature feedback combined with a software control interface enabling accurate planning and heating of the entire target area to a predefined temperature without re‐positioning the animal. Furthermore, the thermal dose delivered to the target tissue is calculated online and can, in addition to the measured temperature change, be used as feedback to guide therapy. The use of a phased array transducer allows complete electronic control over the ultrasound field, which could for example be exploited for fast volumetric heating of the target tissue [22], or for compensation of motion or beam distortion. Finally, the use of a clinical system during pre‐clinical studies facilitates a rapid translation of the developed protocols to clinic.

MR‐HIFU treatment and treatment monitoring of large animals or humans differs from treatment and treatment monitoring in rodents due to the target size. First of all, a high spatial resolution (approximately 1 mm) is required for accurate temperature feedback in rat‐sized tumours (typically around 10 × 10 × 10 mm³). Increasing the spatial resolution of the temperature feedback on the clinical MR‐HIFU system would immediately affect the signal‐to‐noise ratio, which in turn would increase the measured temperature uncertainty [31]. Furthermore, high resolution anatomical images are required for MR therapy assessment post small animal treatment. The use of conventional (phased array) small animal MR volume coils would offer the required image quality; however, most of these coils do not have an acoustic window and are therefore not HIFU compatible.

Another difficulty in treating small animals is to control the exit of the ultrasound wave. The acoustic beam still contains a large part of its initial energy after the target due to the small pathway travelled through tissue. Reflection of this energy would result in uncontrolled far field heating. Finally, the focal point dimensions in combination with the heat diffusion and limited heat capacity of small animals makes it a challenge to treat a small part of the animal while doing minimal harm to healthy tissues and keeping the animal body temperature constant.

We recently showed the use of a dedicated set‐up for hyperthermia‐induced drug delivery treatment of rats on a clinical MR‐HIFU system [32]. Here we report on the design, specification, and performance of this set‐up and show its use for both hyperthermia and ablation treatment of rat tumours. We show that the different features of the clinical system can be used for treatment of rats and potentially mice. The developed set‐up consisted of a small animal HIFU‐compatible MR receiver coil and animal support. The use of a dedicated coil resulted in a higher signal‐to‐noise ratio and/or increased spatial resolution for accurate temperature mapping and therapy monitoring in small animals, while the animal support set‐up minimised far field heating and allowed precise regulation of the animal body temperature. The performance of the combined system was tested in a proof‐of‐concept study, showing that volumetric tumour ablation as well as maintained tumour hyperthermia using a binary feedback temperature control algorithm was possible in rats. The short‐term treatment effect was monitored with standard dynamic contrast enhanced (DCE) MRI in combination with T₂‐weighted anatomical imaging and histology.

Methods

MR‐HIFU

Pre‐clinical HIFU treatments were performed using a clinical MR‐HIFU platform (Philips Sonalleve®, Vantaa, Finland) integrated into a 3T human MR scanner (Philips Achieva®, Best, the Netherlands). The HIFU platform included a HIFU transducer embedded in a MR table top, a RF‐generator cabinet and a workstation for therapy control. The 256‐element spherical shell (12 cm radius of curvature, 13 cm aperture) phased‐array HIFU transducer (Imasonic SA, Besançon, France) could be mechanically translated with five degrees of freedom (three translational and two rotational). The focal point (1 × 1 × 9 mm³ at −6 dB, measured using a needle hydrophone in water at 1.44 MHz) could be steered electronically allowing volumetric heating of the target area. The clinical MR‐HIFU platform was equipped with a single loop RF coil element integrated inside the table top surrounding the acoustic window, combined with a multi‐element coil that could be placed on top of the pelvic area of the patient.

Small animal set‐up

In order to comply with the demands of small animal treatment monitoring (high spatial and temporal resolution), a dedicated small animal HIFU compatible MR receiver coil (Philips Healthcare) was designed and developed. The coil was a 4‐channel volume coil designed for full body, high quality imaging of rats or mice (maximum weight 400 g). The multichannel volumetric design allowed for parallel imaging (sensitivity encoding, SENSE [33]) which could be utilised to decrease the image acquisition time. Mutual coupling of the individual coil elements was prevented by capacitive decoupling of the bottom two elements, geometrical overlap between laterally adjacent upper and lower elements, and high impedance preamplifiers for the upper two elements. High‐pass filtering at the MR receive coil prevented possible interference of the low frequency electrical signals from the ultrasound driver chain with the MR radio frequency signal.

To create an acoustic pathway for the ultrasound beam from the transducer to the inner part of the small animal coil, the coil had an open structure (Figure 1A) and was partially immersed in demineralised degassed water. To this end, the coil was made fully waterproof and was carefully tuned to compensate the effects of the dielectric properties of the water. A pool that fitted onto the acoustic window in the MR‐HIFU table top was designed to support and fixate the coil (Figure 1B). The bottom of the pool was made of thin transparent polyester film (Mylar®), which was chosen based on thickness and acoustical properties to minimise reflections and refractions of the ultrasound when entering the pool. The space between the MR‐HIFU table top and the pool was filled with a thin layer of degassed water to ensure acoustic coupling between the pool and the acoustic window of the table top. A water circulation system was installed against the sides of the pool to heat the pool water. This set‐up resulted in an accessible sonication volume of approximately 50 cm³. A thin copper strip, placed in the middle of the acoustic window at the bottom of the coil, was physically shared by the bottom two elements of the coil in order to further improve SNR. The strip was designed to minimise its impact on the ultrasound beam and did not noticeably influence the focal spot shape [34].

Figure 1. (A) The acoustic window in the bottom of the small animal HIFU‐compatible MR coil (dimensions 10 × 10 cm²) with the copper strip in the middle. (B) The coil fixed inside the water pool and installed on the patient bed of the MR scanner. (C) Side view of the animal bed design. The space between the foil layers (indicated by the arrow) was filled with degassed water and ultrasound gel. (D) Top view of the animal bed showing the details of the design.

To achieve good coupling of the ultrasound beam into the animal and to minimise far field heating, the target area was submerged in a mixture of degassed water and ultrasound gel. Therefore a cradle was designed that consisted of two layers of thin plastic foil (Saran wrap, Dow, Terneuzen, the Netherlands) stretched in a rectangular plastic frame (Figure 1C). The volume between the two layers of foil was filled with a mixture of degassed water and ultrasound gel (Aquasonic 100, Parker, Fairfield, NJ). The ultrasound gel was added to minimise water convection caused by local heating. An acoustic absorber with a pyramidal structured surface (Aptflex F28P, Precision Acoustics, Dorchester, Dorset) was placed on the top foil layer to eliminate far field heating. The pyramidal structured surface of the absorber ensured that the reflected ultrasound did not converge to a focus. A small part of the absorber was covered with a thin isolating layer of rubber material to protect the animal skin from damage due to direct absorber heating. At the top end of the frame, a rat anaesthesia inhalation mask (Minerve, Esternay, France) was placed. A water circulation system was installed on top of the absorber for temperature regulation of the animal (Figure 1D). Finally, the rat was positioned on top of the circulation tubing with the target area submerged below the rest of the body through an opening in the absorber and top plastic layer (Figure 2).

Figure 2. Schematic drawing of the small animal set‐up. The bars indicate the position of the four different coil elements. The bar (green/red) below the tumour is shared by the bottom two elements of the coil. P, posterior; L, left; R, right.

Animal protocol

Hyperthermia or ablation treatment was performed on tumour‐bearing female Fisher344 rats (n = 10, age 5–7 weeks). Tumours were inoculated by subcutaneous injection of 5·10⁶ GS 9L cells (rat glioma, HPA Cultures, Salisbury, Wiltshire) on the hind leg. The tumours had a minimum volume of 400 mm³ at the time of treatment. General anaesthesia was induced and maintained by inhalation of isoflurane (1–3%) in medical air (flow 0.4–0.6 L/min). All animals received RIMADYL® (carprofen, 3 mg/kg bw) as precautionary pain suppressor prior to HIFU treatment. The lower part of the animal was shaved using depilatory cream (Veet, Hoofddorp, the Netherlands). The skin on the target area was covered with degassed ultrasound (US) gel (Aquasonic 100) after which the animal was positioned inside the small animal MR‐HIFU set‐up. Vital parameters were monitored throughout the entire treatment, including respiration frequency using a small balloon sensor (Graseby, Smiths Medical, St Paul, MN) and body temperature using a rectal temperature probe (Neoptix, Québec City, Canada). At the end of the experiment, animals were euthanised by tissue perfusion fixation and the treated area was dissected and conserved for histology. All animal experiments were approved by the animal welfare committee of Maastricht University (the Netherlands) and were in compliance with the guidelines set by the institutional animal care committee, accredited by the National Department of Health.

MR acquisition

The HIFU therapy was planned on a T₂‐weighted 3D image set acquired with a turbo spin echo scan (TSE‐factor: 12, TR/TE: 1500/40 ms, FOV: 150 × 99 × 100 mm³, voxel size: 0.6 × 0.6 × 1.0 mm³, SENSE‐factor: 1.5 (in both RL and AP direction), number of averages: 2, acquisition time: 9:22 min). Subsequently, a susceptibility sensitive scan (T1W 3D steady‐state gradient‐echo, TR/TE: 15/12 ms, FA: 10°, FOV: 280 × 280 × 40 mm³, voxel size: 1.1 × 1.1 × 2.0 mm³, slices: 20, acquisition time: 3:56 min) was acquired to check for presence of air bubbles in the ultrasound beam path.

Direct treatment effects were monitored by comparing T₂‐weighted and dynamic contrast enhanced T₁‐weighted MR images acquired before and after HIFU treatment. The T₂‐weighted images were acquired with a TSE scan at the tumour location (TSE‐factor: 7, TR/TE: 7644/50 ms, FOV: 40 × 60 × 50 mm³, voxel size: 0.5 × 0.5 × 0.6 mm³, slices: 20, rest slabs: 2, acquisition time: 1.39 min (respiratory triggered)) before treatment, immediately after treatment (within 15 min) and 1 h post treatment. The T₂‐weighted images acquired before treatment were used to determine tumour size in a slice by slice approach. DCE‐MRI was performed 1 h before and 30 min after treatment using the low molecular weight contrast agent gadopentetate dimeglumine (Gd‐DTPA, Magnevist®, Bayer Schering, Berlin, Germany). A pre‐contrast reference scan was performed (single shot 3D gradient‐echo, TR/TE: 4.1/1.74 ms, FA: 3°, FOV: 80 × 80 × 42 mm³, voxel size: 0.7 × 0.7 × 0.7 mm³, number of averages: 4, acquisition time: 28.4 s) to calculate a pre‐contrast T₁ map required for quantitative analysis of the uptake kinetics [35]. Gd‐DTPA was injected via a tail vein catheter using an infusion pump (0.2 mmol/kg bw, 0.2 mL/min) after the tenth dynamic scan during a dynamic scan series (single shot 3D gradient‐echo, TR/TE: 4.1/1.74 ms, FA: 13°, FOV: 80 × 80 × 42 mm³, voxel size: 0.7 × 0.7 × 0.7 mm³, number of averages: 1, dynamic scan time: 10.7 s, number of dynamics: 90, acquisition time: 16:05 min).

During ablation therapy, the temperature change in the target region was monitored by continuous acquisition of three adjacent slices perpendicular to the beam axis centred on the target area (RF‐spoiled gradient echo with EPI readout, EPI‐factor: 7, TR/TE: 52/20 ms, FOV: 250 × 250 mm², voxel: 1.4 × 1.4 × 4.1 mm³, fat suppression: spectral pre‐saturation with inversion recovery (SPIR), SENSE‐factor: 1.8, SENSE‐direction: RL, number of averages: 2, dynamic scan time: 5.0 s). During hyperthermia therapy, the temperature change in the target region was continuously monitored in one slice placed perpendicular to the beam axis and centred to the target area, which was acquired with the same acquisition parameters as for the ablation therapy but with the number of averages increased to eight, resulting in a dynamic scan time of 5.9 s.

The temperature maps were calculated online using proton resonance frequency (PRF) shift thermometry [6] and displayed on the therapy control workstation. The temperature images were corrected for baseline drift by subtracting the average apparent temperature change calculated from all voxels in the temperature maps that were located at a sufficient distance from the heating trajectory (3 cm from the trajectory centre) to be considered non‐heated. The SNR of the temperature mapping sequence was calculated by dividing the mean signal intensity of an ROI by the standard deviation of the background noise in the same ROI. The background noise was measured by acquisition of a noise dynamic scan. Thermal dose maps of the target area were calculated online based on the obtained temperature images, according to the Sapareto‐Dewey equation [36], with a unit of equivalent minutes at T = 43°C (EM). During heating, the evolving 30 EM and 240 EM thermal dose contours were shown as overlay on anatomical images of the leg and target area. Thermometry and total thermal dose calculation was continued for 3 min after the end of sonication or until the temperature in the target area had returned to normal body temperature.

Volumetric MR‐HIFU therapy

Volumetric heating of the target tissue was achieved by electronic steering of a single focus point along a circular trajectory (diameter = 4 mm, referred to as the treatment cell) perpendicular to the direction of the HIFU propagation, as described by Köhler et al. [18]. The trajectory was divided into eight points, of which the order of sonication was chosen to maximise the distance between successive points, ensuring even temperature rise over the entire circle. Each location was sonicated for 50 ms before switching (<10 µs) to the next position. Ultrasound was applied as continuous wave ultrasound with a frequency of 1.44 MHz. Low power test sonications (acoustic power: 5–10 W, duration: 20 s) were performed prior to the treatment and used to correct for misregistration between the heated area and planned treatment location. For prolonged hyperthermia treatment, the maximum temperature reached in the target area was used as input for a binary feedback control algorithm which was a modified version of the algorithm described by Enholm et al. [37]. Here, the feedback temperature was compared with pre‐defined threshold temperatures (T_min = 41°C, T_max = 43°C) and used to switch the transducer power on or off accordingly. In all the hyperthermia and ablation treatments, tumours were treated once using a single ellipsoidal treatment cell of approximately 4 mm in diameter and 10 mm in length. The treatment cell was positioned in the tumour and heated to an average temperature of T = 42°C for 15 min using 8 W acoustic power (I_SATA ≈ 117 W/cm²) or shortly to T = 65°C using 30 W acoustic power (I_SATA ≈ 441 W/cm²) for hyperthermia and ablation experiments, respectively.

DCE‐MRI analysis

The uptake kinetics of Gd‐DTPA before and after HIFU treatment were analysed on a voxel‐by‐voxel basis using the 2‐parameter Tofts and Kermode model [38]. The model was fitted to the voxel concentration–time curves measured in the treated area and used for the calculation of quantitative parametric maps of the transfer constant k^trans (min⁻¹), the rate constant k_ep (min⁻¹) and the volume of extra‐vascular extracellular space (EES) per unit volume of tissue v_e (−) using an IDL‐based software tool (IDL 6.3, ITT Visual Information Solutions, Boulder, CO). The transfer constant is a measure for the volume transfer between the blood plasma and the EES. Here, it was assumed that the transfer constant was equal to the blood plasma flow per unit volume of tissue due to high vascular permeability for the low molecular weight contrast agent used. The rate constant was defined as the flux rate constant between EES and blood plasma, or k_ep = k^trans/v_e. The arterial input function was described by a bi‐exponential model function as proposed by McGrath et al. [39]. The corresponding model descriptors were determined in independent experiments in rats. A region of interest (ROI) was manually drawn on each slice containing tumour tissue. The non‐perfused tumour volume was calculated before and after HIFU treatment from the number of voxels for which k_trans < 0.04 min⁻¹. This threshold value was based on empirical findings.

Histology

Tumours and surrounding muscles were collected for histological analysis. The tissue samples were split in half to allow for using both paraffin and cryo‐sections. For the paraffin sections, the tissue was fixated in 4% formaldehyde solution for at least 24 h before embedding in paraffin. The embedded tissue was cut into 5‐µm thick slices and stained with haematoxylin and eosin (H&E) to analyse general morphology. For cryo‐sections, the tissue was snap frozen to T = −40°C using 2‐methylbutane (Acros Organics, Geel, Belgium) and stored at T = −20°C. The frozen tissue was cut into 6‐µm thick slices and air dried briefly at room temperature to allow for nicotinamide adenine dinucleotide (NADH) diaphorase staining to analyse cell viability [40]. The cryo‐sections were incubated at T = 37°C for 1 h in a Gomori‐Tris‐HCl buffer (pH 7.4, 10 mL) containing β‐NAD reduced disodium salt hydrate (N8129, Sigma‐Aldrich, St Louis, MO, 10 mg) and nitro blue tetrazolium (N5514, Sigma‐Aldrich, 10 mg). Specimens were then washed and mounted for analysis.

Statistical analysis

Statistical analysis was performed using the Statgraphics software package (Statgraphics Centurion XVI version 16.0.05, StatPoint Technologies, Warrenton, VA). All data were expressed as mean plus or minus standard deviation. DCE‐MRI results obtained before and after MR‐HIFU treatment were tested for normality and compared using one‐way analysis of variance (ANOVA). In case of equal variances, additional post‐hoc analysis was performed with a paired, two‐tailed t‐test. A p value of less than 0.05 was considered statistically significant.

Results

Small animal set‐up

A dedicated HIFU compatible small animal MR receiver coil (Figure 1A, 1B) was developed and used as an add‐on to a clinical MR‐HIFU system. High resolution anatomical images were obtained and used in therapy planning and assessment (Figure 3A, 3B). For temperature mapping in small animals, a spatial resolution of 1.4 × 1.4 × 4.1 mm³ was obtained while providing a high in vivo SNR of around 258 ± 131 in rat tumour (n = 4) and around 20 ± 14 in rat muscle (n = 2) using the temperature‐sensitive EPI sequence (Figure 3C). These SNRs correspond to theoretical temperature uncertainties ≤0.3°C in the targeted tissue [41], [42]. When the multi‐element coil set‐up provided for human uterine fibroid treatment monitoring was used, SNRs of around 21 ± 11 and 4 ± 2 (n = 2, corresponding to theoretical temperature uncertainties ≤1.7°C) were obtained in tumour and muscle tissue respectively, using the same scan parameters.

Figure 3. (A) Sagittal T₂‐weighted full body image acquired to plan the therapy. (B) Detailed transversal T₂‐weighted image at tumour location acquired with the sequence used for immediate evaluation of the treatment effects. (C) Magnitude image of the temperature mapping sequence in a scan plane perpendicular to the HIFU beam at the tumour location. T, tumour; L, leg; F, foot.

The standard deviation of the temperature change measured with the small animal coil over 15 min in the absence of heating was 0.09°C in a phantom bottle (SNR ≈ 890) and 0.17°C in vivo in tumour tissue (Figure 4). The small but stable decrease in the in vivo temperature after 400 s (Figure 4) represents a small drift in the animal body temperature, indicating the difficulty of keeping the animal body temperature completely stable. Overall, a very good phase and temperature measurement stability was found over time.

Figure 4. Stability of the temperature change measurements over 15 min, measured in a phantom and in in vivo tumour tissue without heating.

The acoustic absorber placed above the submerged target area limited far‐field heating as well as propagation of the exiting ultrasound beam through the entire animal body. This approach in combination with the water circulation system enabled a steady body temperature of 36.8 ± 0.4°C to be maintained in rats during treatment. No skin or muscle damage was observed at the side of the leg where the ultrasound beam exited the body, indicating controlled exiting of the ultrasound wave. As proof of principle, five rats were treated with hyperthermia and five rats with ablation using the developed set‐up (Table I).

Table I.  Overview of the performed animal experiments.

No.	Treatment	Body temperature before treatment (°C)*	Maximum body temperature during treatment (°C)*	Maximum temperature in treatment cell (°C)	Heating rate during start up (°C/s)	Tumour dimensions (l × w × d) (mm)^a	Tumour volume (mm^3)^b	Tumour‐to‐ muscle SIR^c	Treatment time (s)^d	Δ NPV (mm^3)^e
1	Hyperthermia tumour	37.1	37.3	42.7	0.09	8.1 × 6.6 × 14.5	483	3.39	900	33
2		37.0	37.4	43.5	0.12	10.6 × 9.7 × 14.1	741	3.58	900	145
3		37.0	37.2	43.6	0.12	7.4 × 11.1 × 18.5	845	3.33	900	32
4		36.8	37.1	43.1	0.10	7.2 × 8.6 × 11.1	556	3.15	900	187
5		37.3	37.8	43.8	0.15	11.9 × 14 × 15.7	1260	3.77	900	185
6	Ablation tumour	37.2	37.8	65.1	0.51	11 × 8.5 × 15.9	973	3.16	100	857
7		37.2	37.4	64.7	0.53	7.9 × 10.8 × 14.3	756	3.26	66	471
8		36.5	36.7	66.5	0.51	10.3 × 8.8 × 12.6	645	3.41	72	481
9		37.0	37.3	67.4	0.47	10.3 × 10.3 × 12	849	3.79	68	296
10		37.4	37.8	68.0	0.37	7.8 × 11 × 12.4	546	2.96	126	237

SIR, signal‐intensity‐ratio.

*Measured using a rectal temperature probe.

^aMaximum diameters (length × width × depth) measured on T₂‐weighted MR images acquired before HIFU treatment.

^bMeasured by segmentation on T₂‐weighted MR images acquired before HIFU treatment.

^cMeasured from T₂‐weighted images acquired before HIFU treatment.

^dFor hyperthermia experiments, the time at the average temperature of 42°C. In ablation experiments, the total time that the tumour was sonicated.

^eIncrease in non‐perfused volume (NPV) of tumour tissue (k^trans < 0.04 min⁻¹) directly after treatment.

**Not applicable.

Tumour hyperthermia

In the hyperthermia experiments, the temperature in the treatment volume was successfully elevated. The use of the binary feedback algorithm resulted in a stable cycling temperature development around T = 42°C for 15 min (T_min ≈ 40.8°C, T_max ≈ 43.8°C, Figure 5A). Temperatures slightly above and below the defined thresholds were observed due to the dynamic scan time of the temperature mapping sequence. As the under‐ and overshoot did not exceed 1.8°C, the temporal resolution of 5 s that was obtained in this study was considered feasible.

Figure 5. Temperature distribution during (A) hyperthermia on tumour and (B) ablation on tumour. Plots show respectively an image of the temperature mapping during treatment, in which the colours indicate the temperature distribution and the red contour indicates the tumour area, an anatomical image of the animal leg with tumour with an overlay of the thermal dose contour (acquired at the end of the therapy, orange line indicates 30 EM dose contour, white line indicates 240 EM dose contour; F, feet; M, lower leg muscle; A, planned location of the treatment cell centre). Furthermore, the evolution of the maximum temperature in time and the spatial distribution of the temperature elevation are plotted. The dotted lines indicate the end of the sonication in the temporal temperature profiles and the size of the treatment cell in the spatial temperature profiles. For the tumour ablation, an extra plot is included showing the spatial distribution of the thermal dose are shown. FH, feet–head; LR, left–right.

Only minor changes were visible on the T₂‐weighted images acquired pre‐ and post‐hyperthermia in treated tumour (average change in signal intensity: −1.4 ± 7.3%, surrounding muscle: 1.0 ± 3.6%) (Figure 6A). A similar result was obtained for T₂‐weighted images obtained directly after hyperthermia treatment (average change in signal intensity: 5.3 ± 7.3%, surrounding muscle: 2.3 ± 3.7%) as compared to 1 h post‐HIFU hyperthermia treatment. Kinetic modelling of the tissue contrast agent uptake showed a minor decrease in k^trans (Δk^trans = −0.017 ± 0.013 min⁻¹, p value: 0.02), however no change in k_ep (Δk_ep = −0.022 ± 0.035 min⁻¹, p value: 0.22) or v_e (Δv_e = −0.089 ± 0.083, p value: 0.09) was found when averaged over the entire tumour including the treated volume in all animals (Figure 7). Some inter‐animal variability was found in the increase in non‐perfused tumour volume (k^trans ≤ 0.04 min⁻¹) (Table I). On average, the non‐perfused volume increased by 116.4 ± 78.9 mm³ after tumour hyperthermia treatment (average tumour volume: 809 ± 285 mm³), which also affected the tumour averaged transfer constant k^trans.

Figure 6. T₂‐weighted images, T₁‐weighted DCE‐MRI images and parametric maps of k^trans [min^− 1], k_ep [min⁻¹] and v_e [–] of the treated area before and after (A) hyperthermia treatment (data animal 2) and (B) ablation treatment of the tumour (data animal 7). A red ROI is manually drawn around the tumour in the parametric maps. The orange arrow points at local oedema visible after HIFU ablation treatment on the T₂‐weighted scan. hyp, hyperthermia; ab, ablation.

Figure 7. Change in pharmacologic parameters per animal study after HIFU treatment. For treatment parameters see Table I.

Histological analysis using NADH‐diaphorase staining for cell viability showed viable tumour tissue as well as viability of the muscle tissue surrounding the tumour (Figure 8A). On H&E staining, the hyperthermia‐treated tumour appeared similar to control tumour tissue; showing a tight packing of cell nuclei and a large amount of blood vessels (Figure 9A).

Figure 8. NADH‐diaphorase staining after (A) tumour hyperthermia and (B) tumour ablation. In the hyperthermia experiments, all tissue is viable. In the ablation experiments a clearly defined area of non‐viable tissue is visible. The black box indicates the areas from which the high magnification images were taken. v, viable tissue; n, non‐viable tissue.

Figure 9. H&E staining after (A) tumour hyperthermia, (B) 1.5 h post tumour ablation and (C) 4 days post tumour ablation. The black box indicates the areas from which the high magnification images were taken.

Tumour ablation

In the ablation experiments, the lethal thermal dose region perpendicular to the beam axis direction corresponded well with the dimensions of the treatment cell (diameter: 4 mm) (Figure 5B). Along the beam axis, comparison of the area of induced thermal damage to the treatment cell size could not be performed due to the spatial resolution of the temperature monitoring in that direction (three slices of 4.1 mm thickness, as compared to the treatment cell length of 10 mm). No correlation was found between the heating rate and the tumour‐to‐muscle signal intensity of T₂‐weighted images (Table I). Tumour ablation treatment and associated necrosis resulted in local oedema formation in the surrounding muscle, as visible on T₂‐weighted images (change in signal intensity surrounding muscle: −33.6 ± 23.5%, Figure 6B). The area of tissue necrosis in the tumour could not be distinguished on the T₂‐weighted images (change in signal intensity tumour: 1.2 ± 3.6%).

Comparison of the DCE‐MR images before and after ablation treatment gave a good impression of the affected area. The effect of the ablation was largest on transfer constant k^trans, which showed a significant decrease when averaged over the entire tumour including the treated volume (Δk^trans = −0.025 ± 0.016 min⁻¹, p value: 0.03, Figure 7). Outside the treated volume, the k^trans did not show a significant change (Δk^trans = 0.0062 ± 0.018 min⁻¹, p value: 0.36). As a result, an increase in non‐perfused volume (NPV) of 371 ± 123 mm³ was found after ablation treatment using a 4‐mm diameter treatment cell (Table I).

Two animals were allowed to recover after ablation to monitor the effect of the treatment on the animal well‐being and tumour growth. These animals were scanned using MR imaging again 4 days post therapy. Both animals showed shrinkage of the tumour volume compared to before treatment as determined from T₂‐weighted MR images (animal 6: −493 mm³, animal 10: −73 mm³) and showed normal behaviour, including use of both legs. Partial tumour ablation was also seen in histological analysis with NADH‐diaphorase staining (Figure 8B). A sharp demarcation was found between viable and non‐viable tissue at the edge of the ablated region. NADH‐diaphorase staining of the muscle tissue directly surrounding the treated part of the tumour showed a superficial layer of necrosis in the muscle tissue, which extended up to 2 mm maximum into the muscle tissue. The effect of tumour ablation was poorly defined by H&E staining, showing varying results directly after tumour ablation (Figure 9B). Infiltration of neutrophiles into the tumour lesion was observed 4 days after ablation treatment, and cell damage was more pronounced in the treated part of the tumour (Figure 9C).

Discussion

The objective of this study was to develop a dedicated add‐on set‐up for MR‐HIFU treatment of rats in a clinical system and to show that the combined set‐up allows full use of the clinical MR‐HIFU features (e.g. temperature feedback, volumetric heating, binary feedback controlled hyperthermia) for small animals. These features have been extensively studied separately by others in rabbit [22–24], pig [18], [43] and human subjects [44] in clinical systems, and in rodents in dedicated small animal systems [45]. Here, the combination of rodents treated in a clinical MR‐HIFU system is demonstrated.

Due to its robustness, PRF‐shift thermometry is the most widely used method for monitoring temperature changes during clinical MR‐HIFU therapy. However, the available MR receive coil of the clinical system was designed for use in the treatment of human uterine fibroids and did not provide sufficient image quality at the high resolution required for an accurate therapy assessment and temperature monitoring of the HIFU therapy in small animals. A dedicated HIFU‐compatible small animal MR coil was therefore developed, that can be mounted on the MR‐HIFU table top. This way, one order of magnitude image SNR improvement was obtained (increase from 21 to 258 for in vivo tumour tissue), which resulted in high resolution anatomical images and a high temperature accuracy at a spatial resolution sufficient for small animal temperature monitoring (1.4 × 1.4 × 4.1 mm³, three slices). Using the small animal coil, the SNR in muscle tissue remained sufficient for accurate temperature feedback (SNR 20, corresponding to a theoretical temperature uncertainty ΔT ≈ 0.3°C), as compared to the human coil set‐up (SNR 3.7, ΔT ≈ 1.7°C). In other words, to obtain the same SNR using the standard human coil set‐up as compared to the small animal coil, the voxel size would have to be increased to, for example, 3.5 × 3.5 × 8 mm³, which is far too large for use in small animal temperature monitoring. In clinical practice, a voxel size of 2.5 × 2.5 × 7 mm³ is used with the standard human coil set‐up, resulting in a relative temperature uncertainty of ΔT ≈ 0.5°C in a porcine thigh at 1.5 T [18].

The actual measured temperature standard deviation was higher than the theoretical temperature uncertainty that could be expected based solely on the obtained SNR (in a phantom: 0.09°C compared to <0.01°C estimate; in vivo tumour: 0.17°C compared to <0.03°C estimate). This could indicate that at the high SNR obtained, other factors like motion, flow, and field drift become more important in determining the temperature uncertainty.

Fast parallel acquisition, enabled by the 4‐channel design of the coil, resulted in an acceptable temporal resolution of 5.0 s, despite the many phase encodings required for acquisition of high resolution images. Altogether, the thermometry measurements acquired with the small animal coil were sufficient for use in a previously described binary feedback algorithm [37], which was here used for controlling the prolonged hyperthermia in a small treatment volume. The online calculated temperature information from the treatment cell was used to determine whether to continue sonicating and heat the target further or to briefly switch off the transducer power to allow the target tissue to slightly cool down. The use of this algorithm resulted in slow controlled fluctuation around a predefined temperature (Figure 5A) without requiring a priori knowledge of tissue‐specific parameters such as perfusion, energy absorption, heat diffusion or attenuation.

To manage the entering and exiting of the ultrasound beam into the tissue, the target area was carefully positioned and submerged which formed a sufficiently large area for ultrasound contact. The animal was protected from far‐field heating using an acoustic absorber, which resulted in excellent animal body temperature management during treatment (36.8 ± 0.4°C). The latter is especially important for the development of protocols for hyperthermia‐induced local drug delivery from thermo‐sensitive carriers.

The current small animal set‐up was specifically designed to treat subcutaneous tumours on rat extremities. The similarity to clinical cases could be further improved by using a non‐subcutaneous tumour model. Although not specifically designed for it, the use of the HIFU‐compatible small animal coil could be extended to treat other body parts, e.g. liver or pancreas, of mice and rats. However, for treatment of such areas, the acoustic absorber cannot be used to eliminate far‐field heating. This will complicate treatment planning as well as tight regulation of the overall animal body temperature, especially in case of ablation experiments. Furthermore, respiratory and cardiac motion will become an issue in treating abdominal areas.

Coagulated tumour tissue could not be distinguished from viable tumour tissue based on T₂‐weighted scans acquired up to 1 h post treatment, and T₂‐weighted imaging is therefore not considered to be well‐suited for direct assessment of the success of tumour therapy in this rat model. However, T₂‐weighted imaging was considered useful for general tumour demarcation, imaging of oedema, and corresponding inflammatory response after ablation treatment.

Cheng et al. [46] described the use of permeability parameters k^trans and v_e to predict subtle thermal histopathological change. Therefore, DCE‐MRI was here used as the method for direct therapy assessment. A 2‐parameter Tofts and Kermode model was used to estimate the kinetic parameters on a voxel‐by‐voxel basis. The total blood plasma volume (v_p) was not taken into account in the DCE modelling, which may have affected the estimates of k^trans because the intravascular signal could have contributed to a large proportion to the observed tissue signal. However, in this work the kinetic parameters calculated from measurements in the same animal directly before and after treatment were compared. This was considered feasible since errors caused by the contribution of the arterial input function model would affect both estimates to the same extent. The contribution of the intravascular signal however, could have been affected by the HIFU ablation treatment. This effect was not taken into account.

DCE‐MRI was used to calculate the NPV of tumour tissue before and after treatment by setting a threshold for the transfer constant. Thirty minutes post hyperthermia treatment, the average values of the kinetic parameters across the tumour tissue showed a minor decrease in k^trans when the data were averaged over all treated animals (Figure 7). As a result, an increase in NPV was found after hyperthermia treatment in some individual animals (Table I). These results suggest that the vasculature was damaged or destroyed in some parts of the tumour, resulting in local signal decrease, and compensation in other parts of the tumour by an increased perfusion. The effect of hyperthermia treatment on tumour perfusion has been studied extensively. Most studies show an initial perfusion increase [47], but there are also studies that report a decrease in perfusion after hyperthermia treatment which they explain by destroyed tumour vasculature due to local heating above the threshold for vascular damage [48], [49]. For example, heating to 43.5°C was shown to induce a perfusion increase during the first 30 min followed by a pronounced reduction in perfusion up to 24 h after treatment [50].

All animal tumours that were treated with ablation showed a substantial increase in NPV (Figure 7). The post‐treatment NPV depended on whether the heat diffusion region surrounding the treatment cell was located entirely within the tumour or partly outside the animal, as well as treatment duration (Table I). Overall, the NPV varied between 43% and 88% of the pre‐treatment tumour volume. In this study, no long‐term follow‐up was performed after treatment. Instead, standard imaging methods like DCE‐MRI and T₂‐weighted imaging together with histology were used to assess the treatment effect. Although long‐term follow‐up is a prerequisite to determine the therapy efficiency, the study presented here primarily aimed to prove functionality of the small animal set‐up and was not designed to look at long‐term effects of ablation and hyperthermia treatment.

Histology of the treated tissue was performed to analyse general morphology (H&E) and cell viability (NADH‐diaphorase). No tissue alterations were observed on H&E‐stained hyperthermia‐treated tissue. For ablated tumour tissue, H&E gave varying results and remained inconclusive for defining the coagulated area directly after therapy, probably due to local heat fixation of the treated tissue. The latter causes the treated tissue to appear normal directly after therapy, as previously observed in several clinical HIFU tumour therapy studies [51], [52]. Four days post therapy, the heat‐fixed region started to degrade and lose structure, and infiltration of neutrophiles into the lesion became visible on H&E staining (Figure 9C). NADH‐diaphorase staining showed cell viability for the muscles and tumours treated with hyperthermia, and the skin surrounding these areas, while identifying a clearly defined zone of coagulative necrosis after ablation. Due to the sharp demarcation between viable cells and coagulated tissue NADH‐diaphorase staining can be considered very useful in HIFU therapy assessment.

Conclusion

In this paper a method was presented for focused ultrasound treatment of rat tumours using a clinical MR‐HIFU system. To that aim, a dedicated HIFU‐compatible small animal MR coil and animal support were developed that could be used as add‐on to clinical MR‐HIFU systems. This approach provided all the advantages of clinical MR‐HIFU, such as volumetric heating, temperature feedback control and a clinical software interface for use in rodent treatment. Protocols for stable hyperthermia and ablation treatments of subcutaneous tumours were presented that can be used in future studies in large cohorts of animals for the development and optimisation of temperature‐induced drug delivery strategies as well as for finding optimal imaging methods for post therapy assessment of malignant tumours. The translation of the HIFU protocols and biomarkers from pre‐clinical research to human applications can be accelerated with the use of the clinical system during the pre‐clinical development.

Acknowledgements

The authors would like to thank Caren van Kammen, Maastricht University, for her help with the animal studies and her animal handling expertise. Furthermore the authors acknowledge Chatu Jayadewa (Therapeutic Ultrasound Team headed by Gail ter Haar, Institute for Cancer Research, Sutton, Surrey) for useful discussions on histology of US‐treated tissue. This research was performed within the framework of the Center for Translational Molecular Medicine (www.ctmm.nl), project VOLTA (grant 05T‐201) and within the framework of the European Commission project Sonodrugs (NMP4‐LA‐2008‐213706).

Declaration of interest: The authors Edwin Heijman, Max O. Köhler, Mika Ylihautala, Gösta J. Ehnholm, Arjan W. Simonetti, and Holger Grüll are employed by Philips.