Noninvasive magnetic resonance-guided focused ultrasound for tendon disruption: an in vivo Animal study

Abstract Purpose Surgical resection of the tendon is an effective treatment for severe contracture. Magnetic Resonance-guided Focused Ultrasound (MRgFUS) is a non-invasive ultrasonic therapy which produces a focal increase in temperature, subsequent tissue ablation and disruption. We evaluated MRgFUS as a clinically translatable treatment modality to non-invasively disrupt in vivo porcine tendons. Material and methods In vivo Achilles tendons (n = 28) from 15–20kg Yorkshire pigs (n = 16) were randomly assigned to 4 treatment groups of 600, 900, 1200 and 1500 J. Pretreatment range of motion (ROM) of the ankle joint was measured with the animal under general anesthesia. Following MRgFUS treatment, success of tendon rupture, ROM increase, temperature, thermal dosage, skin burn, and histology analyses were performed. Results Rupture success was found to be 29%, 86%, 100% and 100% for treatment energies of 600, 900, 1200 and 1500 J respectfully. ROM difference at 90° flexion showed a statistically significant change in ROM between 900 J and 1200 J from 16° to 27°. There was no statistical significance between other groups, but there was an increase in ROM as more energy was delivered in the treatment. For each of the respective treatment groups, the maximal temperatures were 58.4 °C, 63.3 °C, 67.6 °C, and 69.9 °C. The average areas of thermal dose measured were 24.3mm2, 53.2mm2, 77.8mm2 and 91.6mm2. The average areas of skin necrosis were 5.4mm2, 21.8mm2, 37.2mm2, and 91.4mm2. Histologic analysis confirmed tissue ablation and structural collagen fiber disruption. Conclusions This study demonstrated that MRgFUS is able to disrupt porcine tendons in vivo without skin incisions.


Introduction
Contractures reduce the range of motion (ROM) of joints and are associated with deformity, pain, and disability [1][2][3].While conservative treatments like physiotherapy or casting exist, they are often unable to achieve prolonged increases in ROM in severe contractures [3][4][5][6][7].While effective, surgical resection of the tendon is invasive and can be associated with increased morbidity, costs, extensive rehabilitation, and recurrence, requiring subsequent surgeries [8][9][10][11].Many patients do not respond to conservative treatments or are not appropriate surgical candidates.
High-intensity focused ultrasound (HIFU) is a noninvasive ultrasound therapy that uses extracorporeal transducers to focus ultrasound waves into a small ellipsoid within the body.The transducer emits sound waves which penetrate through tissue and converge at the targeted region leading to a focal increase in temperature and subsequent tissue ablation and disruption [12][13][14].When HIFU is coupled with Magnetic Resonance Imaging (MRI) to visualize, plan, and monitor treatment efficacy, it is referred to as Magnetic Resonance-guided Focused Ultrasound (MRgFUS).To avoid ablation of unintended tissue, real-time tissue temperature is monitored using proton resonance frequency shift MR thermometry (PRFS-MRT) [15,16], which has been demonstrated experimentally to produce thermal maps with temperature accuracies of 1 � C [16].The non-ionizing nature of MRgFUS allows for treatments to be safely performed without long term sequelae [17].MRgFUS is clinically approved for treating uterine fibroids, bone metastasis and prostate cancer [18].Research into musculoskeletal applications of HIFU is in early stages.Ablation of tendons is rapid and efficient due to the tissue's high acoustic attenuation coefficient of 2.9 dB/(MHz cm) [19].
The primary objective of this study was to evaluate MRgFUS's potential as a clinically translatable treatment modality to noninvasively disrupt in vivo porcine tendons.Secondary goals were to assess treatment characteristics such as maximal temperature, temperature spread, skin and intervening soft tissue damage, and to evaluate tissue histology to further guide treatment optimization.

Animal preparation
Animal procedures were performed following protocols approved by the Animal Care Committee at the Hospital for Sick Children (Toronto, Ontario, Canada) under AUP47783, which was approved by the research ethics board prior to experimentation for the humane treatment of animals.Achilles tendons (n ¼ 28) from healthy 15-20kg Yorkshire pigs (male and female) were selected due to anatomical and physiological similarities to the human musculoskeletal system.No exclusion criteria were set.After a one-week acclimatization period and environment enrichment at our animal facilities, pigs were given an intramuscular pre-anesthetic injection of Acepromazine/Atropine/ketamine (0.1/5/11.9 mg/kg), intubated, and ventilated with inhalant anesthesia (2.5% isoflurane in 2 L oxygen) for the duration of experiments.Hair over the hind limbs was removed using an electric trimmer and depilatory lotion.A 22 G angiocatheter was inserted to deliver maintenance fluids (0.9% saline with dextrose).Animal temperatures were stabilized to approximately 37 � C using a water-circulating vinyl blanket.Rectal temperatures were monitored using a temperature probe and vital signs were monitored using an MR-compatible system.

Goniometry
A manual standard goniometer was used to measure the maximal dorsiflexion of the ankle joint before and after treatment.Maximal dorsiflexion was assessed with the knee fixed in three positions (0 � , 90 � , and 135 � flexion as shown in Figure 1) to account for the effect of potential tightness of the gastrocnemius-soleus complex [20,21].This measurement was performed by a senior orthopedic surgery resident who received clinical training and participated in clinical research using this assessment technique.Two non-clinical senior lab members witnessed the assessments to ensure rigorous measurements.

MRgFUS setup, planning, and treatment
The animal was positioned feet first and laterally on the MRgFUS table with the femur directly over top of the acoustic window and the bottom ankle in a dorsiflexed position.Acoustic coupling was achieved by layering degassed reversed osmosis water, a 3.5 mm ultrasound gel pad (Aquaflex, Parker Laboratories, Fairfield, NJ, USA), and degassed ultrasound gel.A 3 T Achieva MRI system (Philips Healthcare, Best, the Netherlands) and a Sonalleve three-channel pelvic coil were used to acquire images for treatment planning, bubble scan verification, and real-time temperature monitoring with PRFS-MRT.The thermometry sequence used in this platform is provided by the vendor as specified in the FDA-approved clinical instructions and validated experimentally by other groups [16,22].It has a temperature uncertainty of less than 1 � C and a spatial accuracy determined by the imaging resolution of 1.5 � 1.5 � 5.5 mm [16,[23][24][25].The experimental setup is shown in Figure 2 and pulse sequence parameters are summarized in Table 1.
Prior to MRgFUS treatment, bubble scan verifications were performed to eliminate any source of near-field thermal injury.Repositioning and repeated bubble scans were performed, when necessary, until an adequate setup and bubble scan was agreed by 2 authors.MRgFUS treatment was delivered using a clinically-approved Sonalleve V1 MRgFUS system (Profound Medical Inc., Ontario, Canada) with transducer frequency ¼ 1.2 MHz, number of elements ¼ 256, surface diameter ¼ 128 mm, radius of curvature ¼ 120 mm, and focal spot size of 1.5 � 1.5 � 9.2 mm (assuming linear wave propagation) when a  2 mm diameter treatment cell is employed [26].Treatments consisted of 30 s sonications at powers of 20, 30, 40, or 50 W. Powers reported are calibrated by the vendor as acoustic powers.This corresponded to treatments of 600, 900, 1200, or 1500 J respectively.A total of 28 Achilles tendons were randomly assigned to one of these treatments using a random number generator with seven tendons in each group.No untreated control group was used as this was a feasibility study, and no a priori sample size analysis was done as the sample size was dictated by budgetary constraints.Two treatments were delivered to each tendon to ensure coverage of the full cross-sectional width of each Achilles tendon (Figure 3).Due to slight variability in the dimensions of tendons from different animals, some tendons had small areas of overlap between the two treatment cells.This was deemed acceptable for the purpose of this study, which aimed at complete coverage to prove the feasibility of tendon release.New T2weighted MR images were acquired after the first treatment to plan for the second treatment (Table 1).A region of edema was present following the first sonication, making it possible for the second treatment cell to be positioned to maximize coverage of the width of the tendon and minimize overlap on the already treated region.Furthermore, the new image ensured that any potential movement of the animal would not offset the targeting of the second treatment.

Post-treatment verification and necropsy
After the final treatment tendons were scanned using the same MR planning protocol.Ankle joints were then ranged to their maximum range of motion.Post-treatment goniometry was performed as described above.Animals were euthanized with sodium pentobarbital (120 mg/kg).Necropsy and gross examination of the tendons were performed to verify the primary end-point of tendon rupture, which was subsequently confirmed with histological analysis.The cold ischemic time between euthanasia and formalin fixation was approximately 15-30 min.A binary logistic regression, receiver operating curve (ROC), and ANOVA and Tukey posthoc tests were calculated using R-Studio (Posit Software, Boston, MA) to compare the frequency of tendon rupture when varying amounts of energy were delivered to the tendon.R-studio was used for all statistical analysis.

Thermal lesion analysis
The thermal area affected by the treatment was measured using the thermal dose approach which predicts a 100% lesion probability at the value of 240 cumulative equivalent minutes at 43 � C (240CEM) [27,28].The Sonalleve system displays the 240CEM area on the coronal plane of treatment.Although this does not represent the treatment volume, it estimates the relative size of lesions per treatment as demonstrated in Figure 4.The analysis of thermal lesion size was performed at three-time points: immediately following the end of the ablation sonication, one minute after the end of the sonication, and five minutes after the end of the sonication.The area measurements from the Sonalleve display were performed manually using ImageJ by two authors blinded to the treatment groups [29].ANOVA and Tukey post-hoc tests were performed to compare the average thermal lesion area at each energy, to indicate the degree of thermal spread at higher and lower-powered treatments.

Treatment temperature Analysis
Temperature monitoring (Figure 5) was performed continuously with a dynamic scan time of 1.7s using PRFS-MRT with a zero-order drift correction to assess the heating of tissue both on and off target with in-plane spatial accuracy of 1.5 mm during treatment, and for a continuous five-minute cooling period after the end of the sonication [30].Analysis was performed   immediately at the end of the sonication, as well as both 1 and 5 min(s) after the end of the sonication to assess the maximum temperature and to monitor tissue cooling.Ensuring tissue temperatures return to physiologic baseline (37 � C) prior to beginning further sonication is important to avoid inaccurate thermal maps and to ensure the safety of the procedure.To calculate the maximum temperatures the temperature within the hottest 3 � 3 voxel region surrounding the maximum temperature were averages.This assessment gives an indication of the temperature in the hottest region of the tissue.Analysis was performed independently by two authors blinded to the treatment groups.The Welch Two Sample t-test was used to assess the relationship between treatment energy and tissue temperature.

Histology and skin necrosis analysis
Following tendon gross examination, they were immersed in 10% neutral buffered formalin, cross-sectioned through the treatment area, embedded in paraffin, sectioned at 4 mm, and stained with hematoxylin and eosin for analysis.Tendons were inspected for disruption, as evidenced by discontinuities in their fibers, by two authors blinded to the treatment groups.Gross and histologic analyses were used to verify tissue ablation and structural collagen fiber discontinuation.
Gross examination of the dermal and subdermal tissue overlying the treatment area was performed following euthanasia to assess undesired thermal spread.Visual inspection also assessed whether the injury was more mild hyperemia, or also included a region of coagulation which would indicate severe tissue damage.The total injury area was measured using ImageJ by two authors blinded to the treatment groups [29].ANOVA and Tukey post-hoc tests were performed to compare the average area of skin necrosis resulting from different treatment parameters.Because this was not a survival study, wound healing, tissue regeneration, and scar formation were not examined [31].

Tendon disruption analysis
Due to joint manipulation during ranging, an audible pop, along with tactile feedback was noted in tendons which were subsequently confirmed to be ruptured during the gross anatomic and histological analysis.In tendons without disruption, no audible pop or tactile feedback was present despite continuous and stressed ranging, and there was no evident fiber disruption visible during necropsy and histology.The primary endpoint was tendon disruption which  was confirmed during necropsy and histology.The number of tendons disrupted varied depending on the energy delivered during sonication (Table 2).A binary logistic regression was performed using energy as a non-negative continuous independent variable and a number of ruptures as a binomial dependent variable [31].This regression demonstrated a p-value of 0.018 with a positive coefficient of 0.22.The 95% confidence interval of the regression coefficient is 0.034 and 0.41.The odd ratio was 1.25 with a confidence interval from 1.04 to 1.50.This binomial logistic regression was compared against a null-model or intercept-only model to evaluate the overall model fit [31].A likelihood ratio chi-square test yielded a p-value of 0.0001, demonstrating that energy was an independent variable The ROC was calculated as shown in Figure 6.The area under the curve was 0.92 (p ¼ 0.002) indicating an excellent fit with a cutoff point of 900 J [32,33].Using the Youden Index, the optimal sensitivity and specificity were 0.91 and 0.83 while giving sensitivity and specificity the same priority [34].

Goniometry Analysis
The changes in ankle joint ROM following MRgFUS treatment are presented in Table 2 and Figure

Thermal dose analysis
The average areas of thermal dose measured were 24.3mm 2 , 53.2mm 2 , 77.8mm 2 and 91.6mm 2 for sonication energies of 600, 900, 1200, and 1500 J respectively.Statistical significance was found between energies of 600 J and 1200 J (p < 0.0002) and between energies of 600 J and 1500 J (p < 0.00003) demonstrating a significant increase in thermal lesion area as treatment energy was increased.Results are summarized in Table 3 and Figure 8.

Temperature analysis
Using the average temperatures surrounding the hottest voxel, the maximal temperatures were 58.4 � C, 63.3 � C, 67.6 � C, and 69.9 � C for energies of 600, 900, 1200, and 1500 J respectively.ANOVA analysis yielded a significance value of p < 0.001.Five minutes following treatment, the maximal temperatures were 37.5 � C, 37.8 � C, 39.1 � C, 40.3 � C for energies of 600, 900, 1200, and 1500 J respectively.No  statistical significance between treatment groups was achieved at this time, which is expected as tissues return to physiologic temperature.Further details are in Table 4 and Figure 9; a temperature of the hottest voxel versus time curve is shown in Figure 10 showing heating during treatment and the temperatures for the five-minute cooling period.

Histology and skin injury analysis
Gross tissue and histologic examination of tendon rupture confirmed the findings in Table 2. Figure 11 illustrates an example of rupture of a sonicated tendon as evidenced by the discontinuation of collagen fibers in the zone of thermal coagulation.In unruptured sonicated tendons, we observed minimal thermal changes and no tendon fiber disruption.Histologic analysis of tendons confirmed tissue ablation and structural discontinuation of collagen fibers in ruptured tendons.Gross tissue and histologic examination of skin injury demonstrated that the average areas of skin necrosis were 5.4, 21.8, 37.1, and 91.4mm 2 for energies of 600, 900, 1200, and 1500 J respectively.Thermal injury area was not statistically correlated to treatment energy, likely due to the large standard deviations (especially at high energies).Findings regarding the characteristics of each thermal injury are summarized in Table 5.

Discussion
We demonstrated that in vivo porcine Achilles tendon disruption is possible using MRgFUS ablation.MRgFUS could be a new methodology to disrupt contracted tendons to increase the ROM of affected joints without skin incisions, soft tissue dissection, or the risks of surgery.This study used MRgFUS treatment energies of 600, 900, 1200 and 1500 J corresponding to powers of 20, 30, 40, and 50 W delivered for 30 s, respectively.20 W was selected as the lowest power for this study as the typical MRgFUS test shot of 10 W for 10s used to assess targeting produces minimal tissue denaturation in most soft tissues.The upper limit power of 50 W for 30s was selected as preliminary studies showed complete tendon disruption beyond this energy.Our primary goal of tendon disruption was verified tactilely during ranging, through visual examination during necropsy and histology, and indirectly through an increase in range of motion.As energy delivered increased from 600 J to 1500 J, the proportion of tendons disrupted increased from 29% to 100%.Results suggest that energies between 900 J and 1200 J were sufficient to cause mechanical tendon disruption.Since we mechanically stretched the tendons to rupture them, we were not able to measure the volume solely caused by ablation.We used the area of the 240CEM region at 5 min following sonication as an indicator of the extent of tissue affected.The average areas measured were 24.3mm 2 , 53.2mm 2 , 77.8mm 2 and 91.6mm 2 for energies of 600, 900, 1200, and 1500 J respectively.Maximal temperatures achieved were 58.4 � C, 63.3 � C, 67.6 � C, and 69.9 � C for the same energies respectively.As expected, parameters such as area, and temperature increased as more energy was delivered.While 5 min of cooling was sufficient to ensure tissue returned to 37 � C before further treatments for treatments at 20 and 30 W, it was insufficient for higher powered  treatments at 40 and 50 W. Longer temperature monitoring is therefore necessary in future work, to ensure that thermal maps of subsequent treatments are not biased by elevated baseline tissue temperatures.
Goniometry was performed to assess the change in ankle ROM following treatment with the knee fixed at three different angles.All treatments caused increases in the ROM, but when treatment groups were compared the only statistically  significant difference was between 900 and 1200 J, when the knee joint was fixed at 90 degrees.Despite confirmation of tendon disruption and differences in the area of thermal dose, the lack of statistically significant differences in ROM between groups most likely stemmed from the inaccuracies of manual goniometer measurements over the significant soft tissue envelope of porcine legs.Although goniometry is cited as the gold-standard method for assessing ROM clinically, more accurate measurements would have been possible using X-Ray imaging of the joint or mechanical platforms specialized in joint measurement [35].Despite the small area of the 240CEM boundary, we still observed skin necrosis with areas averaging 5.4mm 2 , 21.8mm 2 , 37.2mm 2 , and 91.4mm 2 for energies of 600, 900, 1200, and 1500 J respectively.It was observed that higher energy caused more skin necrosis.Analysis revealed that most thermal injuries were either posterior or in the far field, with just a few in the near field.Our feasibility study was not designed to abort treatments even when the temperature map expanded beyond the tendon and onto the skin, which is the most likely source of thermal injury.Additionally, bubble scan verifications to ensure sufficient coupling of the tissue to the acoustic window were effective at mitigating near-field energy absorption due to air trapped at the tissuegel pad interface.With the use of active cooling offered by new FUS systems, and with additional measures such as custom molds to have more precise coupling between the leg and the gel pad, the occurrence of near-field injury can be further minimized.Excessive tissue heating in the posterior and far-field direction could be reduced by using lower treatment powers, or shorter treatment durations.Clinically, the treatment should be aborted when real-time temperature monitoring shows significant heating of the skin.However, as a feasibility study whose primary goal was not safety, we did not stop treatments despite observing undesired thermal spread.
All animals were sacrificed immediately following experimentation.Future survival studies will characterize the healing process of tendons following the FUS therapy.Other studies have characterized tendon healing following thermal treatment and found that fibrosis may occur during healing which would negate the benefits of tendon disruption [36,37].While the recurrence of contracture following FUS is unknown, recurrence does occur following surgical treatment [38].Because this study used a porcine model, differences in the effect of FUS ablation on human contracted tendons could be caused by factors affecting tendon organization like activity level, sex and endocrine hormone exposure, previous traumas, exposure to physiotherapy, and conditioning [39].Many of these heterogeneities in tendon characteristics could be overcome using real-time temperature monitoring with PRFS-MRT and visualization of the area for 240CEM.
In summary, this study demonstrated that MRgFUS can disrupt porcine tendons in vivo without any skin incision.Between 85% and 100% of tendons were disrupted following sonications of 900 J (30 W for 30 s) and 1200 J (40 W for 30 s).At higher treatment energies, the maximum temperature achieved and the area of 240CEM increased, but the occurrence of unintended side effects, such as thermal damage and skin necrosis also increased.

Figure 1 .
Figure 1.Goniometry measurement of the ankle range of motion with the knee stabilized in (a) full extension, (b) 90 degrees of flexion, and (c) 135 degrees flexion, While the angle of the ankle at maximal dorsiflexion is measured.

Figure 3 .
Figure 3. T2-weighted MRI image of the Achilles tendon used in the treatment planning of MRgFUS sonication.Treatment position is verified in (a) sagittal, (b) axial, and (c) coronal views.

Figure 5 .
Figure 5. T2-weighted MRI image of the Achilles tendon with proton Resonance frequency shift MR thermometry (PRFS-MRT) at the end of MRgFUS sonication showing the maximum temperature achieved during sonications.Temperature mapping corresponds to the scale displayed on the right for PRFS-MRT in (a) sagittal, (b) axial, and (c) coronal views.

Figure 4 .
Figure 4. T2-weighted MRI image of Achilles tendon with area Outlining 240 cumulative equivalent min at 43 � C (240CEM) at three time points: (a) immediately after treatment, (b) 1 min after treatment, and (c) 5 min after treatment.
7. For example, with the knee flexed in a position of 135 � , 600 J sonications caused an increase in ankle ROM of 14 � while 1500 J sonications caused an increase in ankle ROM of 30 � .ANOVA and Tukey post-hoc tests were performed for each of the 3 knee angles but only the measurement with the knee at 90 � flexion showed a statistically significant change in ROM between 900 J and 1200 J (p ¼ 0.017).

Figure 6 .
Figure 6.(a) Optimal cutoff point and distribution by class of a number of tendon disruptions by energy delivered and (b) receiver operator curve with optimal sensitivity is of 0.91 with a specificity of 0.83 while giving sensitivity and specificity the same priority.

Figure 7 .
Figure 7.The difference in ankle ROM following MRgFUS treatment with a knee at (a) full extension (0 degrees), (b) at 90 degrees, and (c) at 135 degrees.

Table 4 .Figure 9 .
Figure 9. Maximum temperature measured at three different time points (1) immediately after treatment, (2) 1 min after treatment, and (3) 5 min after treatment for each energy delivered per sonication.

Figure 8 .
Figure 8. 240 CEM areas at three different time points (1) immediately after treatment, (2) 1 min after treatment, and (3) 5 min after treatment for each energy delivered per sonication.

Figure 10 .
Figure 10.Temperature-time curves showing the hottest voxel temperatures during treatment and for five minutes of cooling for (a) a 20 W treatment and (b) a 40 W treatment.

Figure 11 .
Figure 11.(a) Gross anatomy is a ruptured tendon with an area of tendon rupture.(b) Histology of ruptured tendon with fiber disruption in the zone of thermal coagulation.(c) Magnification in the border between normal collagen fiber and zone of thermal coagulation.(d) Magnification in border between a zone of thermal coagulation and the absence of collagen fibers due to disruption.

Table 1 .
Summary of MRI and PRF-Based MRT pulse sequence parameters.

Table 2 .
Summary of tendon disruption for each energy delivered per sonication.

Table 3 .
Summary of 240 CEM area for each energy delivered per sonication.Area of lesion treatment as per 240 CEM (mm 2 )