Anatomical-based model for simulation of HIFU-induced lesions in atherosclerotic plaques

Abstract

Purpose: The aim of this study was to simulate the effect of high intensity focused ultrasound (HIFU) in non-homogenous medium for targeting atherosclerotic plaques in vivo. Materials and methods: A finite-difference time-domain heterogeneous model for acoustic and thermal tissue response in the treatment region was derived from ultrasound images of the treatment region. A 3.5 MHz dual mode ultrasound array suitable for targeting peripheral vessels was used. The array has a lateral and elevation focus at 40 mm with fenestration in its centre through which a 7.5 MHz diagnostic transducer can be placed. Two cases were simulated where seven adjacent HIFU shots (∼5000 W/cm², 2-s exposure time) were targeted on the plaque tissue within the femoral artery. The transient bioheat equation with a convective term to account for blood flow was used to predict the thermal dose. The results of the simulation model were then validated against the histology data. Results: The simulation model predicted the HIFU-induced damage for both cases, and correlated well with the histology data. For the first case thermal damage was detected within the targeted plaque, while for the second case thermal damage was detected in the pre-focal region. Conclusion: The results suggest that a realistic, image-based acoustic and thermal model of the treatment region is capable of predicting the extent of thermal damage to target plaque tissue. The model considered the effect of the wall thickness of large arteries and the heat-sink effect of flowing blood. The model is used for predicting the size and pattern of HIFU damage in vivo.

• Atherosclerosis
• HIFU
• numerical simulation
• therapeutic ultrasound
• thermal ablation

Introduction

Atherosclerosis is a systemic disease that affects large- and medium-sized arteries. The disease starts as an endothelial dysfunction, which initiates recruitment of inflammatory cells and lipids within the arterial wall [1]. This, together with smooth muscle hyperplasia, leads to the development of slowly growing atherosclerotic plaques that become clinically symptomatic when they cause significant haemodynamic stenosis.

Various modalities are readily available for the treatment of atherosclerosis. These treatment options can be classified broadly into ‘conservative’ vs. ‘interventional’ treatments. Conservative treatment, known as ‘best medical management’, entails modifying patient’s life style and risk factors for the disease, including controlling hyperlipidaemia and hypertension. Interventional treatment includes both surgery and minimal interventional percutaneous procedures. The degree of luminal stenosis and patient’s symptomatology, among other factors, determine the best treatment option for each patient [2]. Each treatment option has its own limitations and risks. For example, best medical management can be used to treat carotid artery stenosis. However, this treatment is not recommended as stand-alone treatment for plaques causing more than 60% luminal stenosis [3]. On the other hand, carotid endarterectomy, angioplasty and stenting can carry the risk of causing distal embolism, stroke or even death during the procedure [2]. As a general rule, the use of minimal invasive procedures has the potential to reduce the complications that may arise when more invasive interventions are utilised. In this context High intensity focused ultrasound (HIFU) may present a novel option for treatment of localised atherosclerotic plaques in terms of providing a completely non-invasive treatment paradigm for ablation of atherosclerotic plaques. Current clinical applications of HIFU for pancreatic cancer and future challenges are described in Xiaoping and Zheng [4]. An evaluation of the value of the treatment of malignant tumours of the bony pelvis using ultrasound-guided HIFU ablation is presented in Yang et al. [5].

A review of the current unmet clinical needs in addition to technical and clinical studies to accelerate the development of treatment of liver cancer using HIFU is described in Aubry et al. [6].

The theory behind investigating the use of HIFU for ablation of localised atherosclerotic plaques is that plaque growth is a dynamic process that entails neovascularisation and angiogenesis to support the build-up of the viable plaque tissue. Thus, ablation of the plaque substance may result in thrombosis or occlusion of these neo-vessels. This in turn may slow down or even stop the progressive growth of these plaques. In addition, the ablated plaque may heal by fibrosis. Fibrotic tissue may retract over time resulting in relative shrinkage of the plaque substance. Taking uterine fibroids as an example, the ablated fibroid tissue shrinks over time; and this brings symptomatic relief of patients’ complaints.

HIFU beams can be precisely focused within a small focal volume, resulting in a rapid rise of the local tissue temperature. This temperature elevation causes localised tissue damage through coagulative necrosis [7,8]. With proper choice of operating frequency and transducer design, this can be achieved without significant biological damage to the intervening tissue [9,10]. The advent of image guidance modalities such as MRI and ultrasound has allowed for completely non-invasive thermal therapy procedures [11–14]. Different applications of image-guided HIFU therapy for potential medical applications have been proposed earlier [15,16] in addition to a computational model for ultrasound tomography [17]. The computational modelling of these therapies has become an important research tool, allowing the physician to improve monitoring during treatment. Furthermore, modelling allows the prediction of thermal lesion prior to the treatment which can be of great benefit in treatment planning. Several numerical models already exist for simulation of the treatment process [18,19] and predicting the outcomes of some other procedures. In addition, some groups have studied the non-linear thermal effect of the wave propagation [20–22] and the vascular damage induced by ultrasound [23,24].

Several guidelines currently exist for treatment of atherosclerosis, including medical, surgical and interventional treatments [3]. We have been investigating the use of HIFU to target localised atherosclerotic plaques in vivo. We demonstrated the feasibility of inducing HIFU thermal lesions within atheromatous plaques in the femoral arteries of familial hypercholesterolaemic (FH) swine in vivo [25]. In this study, precise thermal damage was induced within the plaque tissue with no incidence of plaque rupture or endothelial damage. The protection of the endothelium is most probably attributed to the heat-sink effect of the blood flow in the artery, which resulted in a thin (∼200 μm) layer of protected tissue (endothelium and part of the tunica intima). No significant side effects or adverse events were reported in this study [25].

Imaging and histological studies of peripheral vessels in the FH swine model have clearly demonstrated the heterogeneity of the tissues surrounding the target plaque tissue [25]. The HIFU beam goes through regions of skin, fat and muscle in addition to the connective tissue around the vessel. These studies also showed that diagnostic ultrasound, when carefully evaluated before treatment, can produce sufficient anatomical contrast to differentiate between these tissue types. This information can be used to define a heterogeneous propagation model for the HIFU beam and a heterogeneous model for the transient bioheat transfer equation. Furthermore, the arteries and veins in the treatment region can be easily identified on a diagnostic ultrasound. This provides an opportunity to account for their effect on the resulting heating patterns.

The availability of a realistic numerical model for the HIFU propagation and temperature evolution in the presence of tissue heterogeneity and directional blood flow would be highly advantageous for the potential treatment of atherosclerotic plaques. It provides an opportunity for pretreatment planning to allow for treatment optimisation. It also allows for retrospective analysis of the treatment results to help the physician decide post-treatment protocol. In this paper we present results from a heterogeneous acoustic and thermal model derived from pretreatment ultrasound images of the treatment region. In addition to the tissue heterogeneity, the thermal model contains a convective term accounting for blood flow in the target vessel. Both models utilise the finite difference time domain (FDTD) discretisation of their respective partial differential equations (PDE). The model geometry is obtained based on an image segmentation of a pretreatment ultrasound image of the target region. The model accounts for the water bolus, skin, fat, muscle, and connective tissue in addition to target vessel lumen. The acoustic and thermal properties of the different tissues within the treatment region are obtained from the literature.

Two simulations of actual HIFU treatments targeting the femoral artery in the FH swine model are presented in this paper. In one scenario the model showed that all HIFU shots have been able to produce lesions at the intended target plaque without damage to the endothelium or intervening tissue in the path of the HIFU beams. In the second scenario the model predicted a pre-focal lesion near a vessel in the path of the HIFU beam. In both cases the model predictions were consistent with histological evaluations of the treated regions.

Materials and methods

Ultrasound pretreatment imaging

A conventional diagnostic ultrasound imaging (Sonix RP, Ultrasonix, Richmond, BC, Canada) was conducted through the 7.2-MHz diagnostic transducer (L14-5/38, Ultrasonix) to visualise the artery and the plaque before the targeting procedure. This was also performed to collect data regarding the depth and the diameter of the artery in addition to the thickness of the subcutaneous fat region. An investigation of the feasibility of targeting atherosclerotic plaques by HIFU in four swine was shown in Shehata et al. [25]. In addition, the procedure of targeting the plaques was proven under general anaesthesia and different ablation protocols were used to meet the study objectives.

Histology procedure

Gross pathology examination/necropsy and tissue collection

The animals underwent a target necropsy and the proximal end of each treated artery was exposed. Lactated Ringer’s solution was used to flush the treated vessel followed by a perfusion fixed in situ with 10% neutral buffered formalin (NBF) for more than 10 min. A surrounding tissue section of 1 cm of untreated proximal and distal vessel was harvested in addition to the artery. The proximal end of each vessel was sutured with a label and placed in 10% NBF.

Tissue trimming

The vessel was fixed and cross-sectioned through the whole vessel length. A photograph was taken of the treatment sites section and recorded on the tissue trimming form while the sample was placed in the tissue processor for paraffin infiltration.

Histological processing of target specimens

The samples were placed in paraffin after being removed from the tissue processor and sectioned through the entire block at 200-µm intervals. Slides of 4 µm were cut and stained with haematoxylin and eosin (H&E) for evaluation.

Histological evaluation

Each section was evaluated using light microscopy including analysis of skeletal muscle and muscle wall lesion severity.

Integrated dual-mode ultrasound array (DMUA) system for image-guided lesion formation

The DMUA system described in Casper et al. [26] was used for targeting the femoral artery in the FH swine model. This system is capable of imaging and delivering therapeutic HIFU simultaneously, which results in inherent registration between the (imaging and therapy) coordinate systems. This ensures accurate targeting and monitoring of the lesion formation process. The integration of both the diagnostic imaging transducer and the DMUA transducer, shown in Figure 1 extends the imaging field of view (FoV) of the image guidance system [25,26]. This study was approved by institutional animal care; animals were fasted for at least 12 h before the procedure. Swine were placed in supine position and a 7.2-MHz linear array imaging probe was used to evaluate the atherosclerotic burden in the arteries before the treatment process using DMUA.

Figure 1. The HIFU transducer. The 3.5-MHz DMUA showing a central fenestration for integration of a diagnostic ultrasound transducer.

The DMUA used herein operates at 3.5 MHz and has a central fenestration through which a diagnostic transducer (HST15-8/20) is applied as shown in Figure 1. Specifications of the DMUA used are summarised in Table 1.

Table 1. DMUA specifications.

Number of channels	32 × 2
Radius of curvature	40 mm
Centre-to-centre spacing	1.5 mm
Kerf (inter-element spacing)	0.2 mm
Height of the elements	30 mm with 16.5 mm central fenestration

Simulated treatments

We simulated two cases for targeting atherosclerotic plaques within the posterior wall of the femoral artery of FH swine. In each case the treatment consisted of seven HIFU shots with the foci spaced 1 mm, circumferentially at the periphery of the posterior wall. The intensity was approximately 5000 W/cm² with 2-s exposure duration for each shot. The wait time between shots was approximately 2 s. Electronic steering was used to generate the shots without the need to move the transducer mechanically.

Case 1

The plaque burden was assessed using diagnostic ultrasound prior to HIFU treatment. The HIFU beam passed through the skin, fat, muscles, and perivascular connective tissue before crossing the arterial lumen to reach the targeted plaque in the posterior wall of the femoral artery as shown in Figure 2. The animal was sacrificed 3 days after treatment for histological evaluation of the target (plaque) tissue and surrounding structures. This case was an example of a successful treatment as evidenced by histological evaluation.

Figure 2. Segmentation of the ultrasound image from the treatment region. Segmentation lines outline the main anatomical structures. S, skin; F, fat; M, muscle; A, artery; CT, connective tissue.

Case 2

The plaque burden was assessed prior to the treatment using diagnostic ultrasound. The saphenous artery (proximal branch of the femoral artery) path in front of the femoral artery to come within the path of the incident HIFU beams possibly causing scattering and distortion of the therapeutic HIFU beam as shown in Figure 3. The animal was revived for 30 days after the treatment to observe any short-term effects of the HIFU treatment. The animal did not exhibit any loss of function or other adversary effects during this 30-day period prior to sacrifice. Histological evaluation in this case showed a pre-focal lesion proximal to the saphenous artery and represents an example of collateral damage that could have been predicted by numerical modelling.

Figure 3. Ultrasound scan of the treatment region from case 2. The saphenous artery (Sp) traversed superficially to the femoral artery (FA) to come within the path of the incident HIFU beam.

Numerical model

A finite difference approximation of partial derivatives was used in the numerical solution of the PDEs. The FDTD method is a powerful tool for solving a broad range of PDEs [27]. The FDTD heterogeneous model for acoustic and thermal tissue response was used; the model is based on the wave equation and the transient bioheat transfer equation. This was achieved after segmentation, ∼4 min, of the ultrasound image to yield accurate information about the regional anatomical details and by using the values of Table 2 [28,29]. A level set method described in Li et al. was used for segmentation of the different tissue regions. This method is based on the intensity in- homogeneities model of the images. This was achieved by defining, in the neighbourhood of each point, a local intensity-clustering criterion, which is then used to give a global criterion of image segmentation. The level set method is able to simultaneously segment the image and estimate the bias field that accounts for the intensity inhomogeneity of the image. This can be used for intensity inhomogeneity correction.

Table 2. Acoustic and thermal values by medium [28,29].

Type	Density kg/m^3	Specific heat J/kg °C	Thermal conductivity W.m^−1/°C	Speed of sound m/s	Attenuation dB/cm/MHz
Skin	1109	3530	0.4	1400	1.15
Fat	960	2973	0.24	1476	0.36
Connective tissue	1525	2372	0.39	1613	1.00
Muscle	1090	3600	0.42	1547	2.00
Blood	1050	3800	0.55	1627	0.10
Blood vessel wall	1102	3400	0.48	1550	1.15
Atherosclerotic  plaque	1320	2900	0.49	1600	1.20

The simulation accounted for the thickness and nature of the skin, subcutaneous fat, muscles, perivascular connective tissue and the blood vessel within the path of the HIFU beam as illustrated in Figure 2, and shows an ultrasound image of the target femoral artery (A) with the skin (S), fat (F), muscle (M), and connective tissue (CT). A continuity boundary condition was assumed between different tissue types in the numerical simulation.

To maintain the stability of this model the temporal step follows the Courant, Friedrichs and Lewy (CFL) condition [27]. This condition defines the changes of the time step with respect to the changes in the mesh size. This condition is necessary for the convergence of the used FDTD method and hence for obtaining a stable solution. The numerical method is stable if the numerical solution is bounded and no oscillations exist (independent of the excitation). The electronic steering of the HIFU beam was considered in this simulation, as described in Ebbini [31] and Almekkawy [32].

Acoustic model

Mathematical methods

The propagation of the ultrasound field is governed by the wave equation. We used this equation to model the HIFU field from the DMUA. This model followed the Euler and continuity equations [33,34]. (1) where û, p, and c, denote the particle velocity, pressure and speed of sound, respectively. The absorption coefficient ‘γ’ is used as an indication of the absorptive loss in the medium which can be related to the attenuation coefficient ‘σ’ through the complex wave number k as in the following equation (2) where c′ is the dispersive wave velocity. In absorptive media, both c and σ are frequency dependent in absorptive media, but because the HIFU beam is narrowband, we will assume that they are independent of frequency over the bandwidth of interest. The two equations in (1(1) ) can be coupled to get the second order partial differential equation as follows. (3) In this paper, we solve the decoupled equations.

A modification has to take place, according to [33], in Equation 1(1) using a stretched coordinate. This new coordinate will make it suitable to solve the PDE equation numerically in the PML region which is used as an ABC. The stretching is chosen as (4) as ω is the angular frequency and g is expressed as (5) and . The term represents the loss in the PML while is a scaling factor. Equation 1(1) should be modified by using the stretched coordinates and by replacing ▿ with ▿_g, we will get the following. (6) These two equations are used in the whole computational domain with for the interior region. The spatial and temporal partial derivatives are approximated by a Taylor series expansion for each node in the computational grid; this grid consists of spatial dimensions using uniform spacing in addition to temporal dimension Δt with uniform spacing. A perfect matched layer as a material ABC was applied at the computational edge of the grid to eliminate the reflections from the outer boundary of the computational domain. A higher absorption is achieved by using PML as an absorbing boundary condition. To include the PML in the absorptive media for the acoustic waves, the computational domain, consisting of N_x, N_y, N_z cells in 3D, is partitioned into a normal interior region and the boundary PML region. After modification of Equation 1(1) by using the stretched coordinates as described, and after using the FDTD method, the equations should be approximately (7) After rearrangement, the time stepping equation for u_x is equal to (8) where For the pressure field p (9) The time stepping equation for p is approximately equal to (10) where The PML region is responsible for the absorption of the outgoing propagated wave; it consists of cells with an increasingly quadratic attenuation profile toward the outer boundary as follows (11) where q = 2 as we are using a quadratic profile and is the value at the centre of the first cell. The simulation was conducted at a 3.5-MHz central frequency and the simulation grid was set to ≈0.04 mm. The simulation HIFU exposure was conducted for 100 µs and was applied for 600 cycles.

A perfect matched layer

The finite difference equation is applied at interior and boundary nodes. In certain cases it is important to extend the solution domain beyond the natural boundaries of the physical problem to improve the expression of the boundary condition. This extension is called the perfect matched layer (PML) which has an important role in numerical simulations of acoustic wave propagation in inhomogeneous media. This is achieved by artificially adding boundary conditions in finite difference methods to minimise the reflection caused by the edges of the computational domain. The PML was first introduced by Berenger [35] as a material absorbing boundary condition (ABC) for electromagnetic waves. We followed Liu and Tao [33] to account for the coupling of loss from a PML and the regular absorption loss in acoustic wave propagation as shown in Equations 4–11(4) .

Thermal model

The transient bio-heat transfer equation is used to govern the thermal effects in tissue with a convective term, as described in Saperto and Dewey [36]. The temperature model is described by the following equation: (12) where T, C, ρ, K, w, v, Q are the temperature, specific heat, density, thermal conductivity, perfusion rate, blood flow velocity, and the external heating (HIFU) source, respectively. The term ▿.K ▿T models the thermal diffusion, while w_bC_b (T_a−T) models the effect of perfusion. The boundary condition of the tissue domain is conducted through body temperature (37 °C) at the domain borders. The temperature initial condition is assumed to be (37 °C). The simulation was conducted at a 3.5-MHz central frequency with an approximately corresponding wavelength of 0.42 mm and the simulation grid was set to one tenth of a wavelength. A forward-time central-space (FTCS) scheme has been used with applying upwind differencing (class of numerical discretisation) for the convection component [37]. An explicit method was used and a forward difference for the time derivative, a backward difference for the first spatial derivative and a centred difference for the second spatial derivative is as follows: (13) The numerical error is in the order of Δt and Δξ. In order to meet the stability criteria as indicated previously, the temporal step followed the CFL condition [27] to achieve the convergence of the difference approximation. Thermal dose

The overall performance of the thermal dose calculation is evaluated by using an empirical exponential relation between the tissue temperature and the exposure time required to coagulate the tissue used to predict the thermal dose (in equivalent minutes at 43 °C) [36] (14) where TD is the thermal dose which relates the time required for coagulating the tissue at temperature T to the equivalent time at TD43. The size of the necrosis is calculated based on the threshold value of 240 min at 43 °C [38]. All thermal dose calculations are normalised to this standard.

Results

The acoustic 2D field simulation accounted for the phased array DMUA concave geometry [26] of 3.5 MHz 32 × 2 elements and has a central fenestration through which a diagnostic (HST15-8/20) transducer probe is applied. DMUA images and the diagnostic transducer images are provided together for identifying target locations and monitoring the dynamics of the lesion formation in space and time. We examined a single artery with the focus at the posterior wall of the artery with the artery running along the transverse direction of the transducer. We simulated the temperature increase and the lesion formation after each shot. The numerical simulation modelled the domain of the targeted plan with the water-filled bolus used to couple the diagnostic transducer and the HIFU transducer to the animal.

The transducer, with elementary pitch of 1.5 mm and 0.2 mm inter-element spacing, was used to target the femoral artery of the adult swine through a water bolus. The simulation grid spacing was set to one tenth of the acoustic signal wavelength in water with 1300 computational grid points. The acoustic and thermal simulation parameters are shown in Table 2. The computational time is approximately 8 min for the 2D acoustic simulation and 3 min for the 2D thermal simulation for each target point in a quad processor with 12 GB memory. The area of induced HIFU thermal damage was defined as the tissue area that reached a thermal dose equivalent to 43 °C for 240 min.

In both cases, the number of shots was seven shots per plane and the duration time for each exposure was 2 s followed by a ∼2 s cooling period for a total time of 26 s per plane. Both cases tolerated the procedure well, with no evidence of considerable changes in vital signs or haemodynamic instability. In addition no dissection or rupture occurred in the artery during or after targeting with the HIFU beam. The arteries were displaced slightly during the targeting period as a result of the acoustic radiation force of the HIFU beam but they returned back to their original position retaining normal pulsatility upon stoppage of the HIFU beam. There was no report of disability during the period of recovery and animals showed normal activity until the day of euthanasia.

Case 1

An ultrasound image of the target femoral artery was acquired and segmented using the level set method as described in Li et al. [30]. The segmentation contours and the different tissue types are shown in Figure 2. This figure shows the B-mode image, using the L14-5/38 transducer, of the domain including the artery and vein in which the artery lies superficial to the vein. Running the simulation program showed that HIFU, at the aforementioned exposure parameters, was able to induce damage in the substance of the targeted plaque. The simulation showed that the individual HIFU lesions overlapped inside the substance of the treated plaque, forming a confluent zone of thermal damage. HIFU-induced thermal injury within plaques showed distinguishable histology characteristics. Necrotic cores containing cellular debris is shown in the targeted plaques. Reactive inflammation was noted as evident by infiltration of the periphery of the lesions by neutrophils and other inflammatory cells.

There was unequivocal injury in the adjacent skeletal muscle. In addition there was no evidence of HIFU-induced damage in the targeted healthy arterial wall or microscopic injury of the endothelium. From the histological examination of the targeted arteries shown in Figure 4(B), the lesions are partly overlapped to form a confluent ablated zone approximately measuring 6.92 mm in length and 0.93 mm wide as detected by histology evaluation. As shown in Figure 4(A), the extension and configuration of the HIFU-induced thermal damage, as predicted by the simulation, matched the actual damage zone with approximately 6.5 mm in length, and 1.1 mm in width.

Figure 4. Correlation between the simulation results and histology. (A) Simulation result of the thermal dose within the targeted atheromatous plaque. Note the intact intima (black arrows) overlying the targeted plaque. (B) Thermal damage within the targeted plaques as shown from histology (H&E stain).

Another important outcome predicted by the proposed model was that the temperature distribution over the intima was lower than at the focus, within the plaque itself. Therefore, the intima was kept intact without any significant damage. No signs of luminal thrombus or injury of the endothelium lining the artery were detected. Endothelialisation was confluent with total circumferential coverage. Figure 4(B) shows the intact intimal endothelial lining overlying the targeted plaque. This was in keeping with the results predicted by the simulation model. The results of our simulation correlated well with the histological evaluation, which revealed thermal injury within the plaque tissue consistent with the HIFU shots applied.

Case 2

In this case the aim was to target an atherosclerotic plaque within the posterior wall of the external femoral artery partially obscured by another artery on the HIFU path in large (100–140 kg) swine. The simulation model showed evidence of pre-focal damage in the vicinity of the saphenous artery, while the temperature rise within the targeted plaques was below the threshold for thermal damage. Histology analysis was again consistent with the simulation results. Lesion localisation in histology slides was done by identifying the arranged skeletal muscle markers through the microscopic evaluation. As shown in Figure 5(A); histology analyses showed evidence of fibrosis (scarring) in the connective tissue adjacent to the saphenous artery, while there was no unequivocal HIFU-induced thermal damage in the targeted plaque. The fact that no unequivocal HIFU lesions were detected within the targeted atherosclerotic plaques can be explained by the orientation of the saphenous artery in relation to the femoral artery during treatment. Figure 6 shows that the saphenous artery, together with its perivascular connective tissue, is directly superficial and in front of the femoral artery within the path of the HIFU beam. It also shows the arterial pulsation detected from the two-dimensional M-mode and strain imaging techniques [39,40] used to capture the transients of tissue motion/deformations according to sub-therapeutic HIFU beams.

Figure 5. Correlation between the simulation results and gross histology. (A) Simulation result of the thermal dose within the prefocal region, the arrow points to the prefocal damage. (B) Prefocal thermal damage within the same region as shown from gross histology. FA, femoral artery; Sp, saphenous artery.

Figure 6. B-mode image of the treated region while targeting the femoral artery. The strain imaging on the right hand side shows arterial pulsations indicative of existing arteries in the region of interest. FA, femoral artery; Sp, saphenous artery.

A 3.5-MHz phased array was used to generate sinusoidally-modulated intensity level HIFU beams near blood vessels. This technique is implemented on a Sonix RP scanner that acquires the two-dimensional RF data at a high frame-rate (>500 fps with limited frame sizes covering the vicinity of the HIFU focal spot). The system is capable of detecting the thermal and mechanical tissue response due to HIFU-induced strains. Orientation was attributed to slight tilting of the transducer during treatment. With this orientation, the saphenous artery became aligned with the femoral artery as compared to its position in the pre-procedural ultrasound scanning, as shown in Figure 3. Thus, the saphenous artery may have masked significant HIFU energy from reaching its original target within the plaque in the femoral artery. The simulation program showed that HIFU induced damage in the pre-focal region instead of the target area. In both cases, the HIFU lesions formed, as predicted by our simulation, are consistent with the actual damage determined based on histological evaluation.

The limitation of this study is that while the simulation results registered well with tissue histology regarding lesions configuration, extension and overlap in addition to demonstration of collateral damage in the pre-focal region, it was difficult to accurately correlate the ‘distances’ between the simulation and the histology slides. Tissue deformation, either as a result of tissue stretching during cutting for tissue processing, or tissue shrinkage due to the effect of formalin are unavoidable. This results in a slight mismatch between the distances generated by the simulation program with those of histology.

Discussion

We have previously demonstrated that HIFU can be delivered safely and with sufficient energy to induce thermal damage within atherosclerotic plaques in vivo [25]. However, for the proposed approach to gain acceptance, we needed to explore the behaviour of the HIFU lesions within the targeted plaques. Simulation programs such as the 2D acoustic and 2D thermal models proposed here should help to predict the extent and pattern of thermal damage induced within the targeted plaques. Our experience so far shows that, even for the same exposure parameters, the size and extension of the HIFU lesions within the plaque are not consistent. This may be attributed to differences in plaque composition, blood flow parameters or other anatomical constraints in the path of the HIFU beam. The proposed simulation model considered the effect of wall thickness of large arteries and the heat-sink effect of flowing blood. Furthermore, we integrated the anatomical details obtained from an ultrasound scan of the region of interest within our simulation program. Thus, we took into consideration the attenuation of the HIFU beam caused by different anatomical regions. We consider histological analysis of the targeted plaques as the gold standard for evaluation of HIFU-induced damage.

The treated swine underwent necropsy by trained staff at the test facility. The treated arteries were perfused with lactated Ringer’s solution followed by 10% NBF. The entire block section of the artery and the surrounding muscles was then harvested and placed in 10% NBF. After fixation, arterial segments believed to contain targeted plaques were placed in cassettes and left in tissue processor for paraffin infiltration. Samples were then serially sectioned through the entire block at 200 µm intervals. Slides were then further cut at 4 µm and stained with H&E stain [25].

We had a unique opportunity to validate the results of our simulation program against those of histology. For case 1, the extension and configuration of thermal damage predicted by the simulation program correlated well with the actual HIFU-induced thermal damage detected within the targeted plaque on histology. Of note, running the simulation program clearly showed the pattern by which the discrete thermal lesions overlapped to create such a confluent zone of damage. This pattern can be explained by the thermal conduction within the fat-rich plaque that caused the adjacent discrete HIFU lesion to overlap. The fact that the intimal endothelial lining was kept intact can be explained by the heat-sink effect of the flowing blood in the targeted vessel that appears to protect the intima. An intact overlying endothelium is an important safety determinant of this procedure because any intimal damage, if accidentally induced, can carry the risk of inducing localised thrombosis or even serious distal embolism.

For case 2, the saphenous artery with the surrounding connective tissue passing in front of the femoral artery increased the pre-focal attenuation of the HIFU beam, preventing the delivery of enough HIFU power to the inside of the targeted plaque. This resulted in unintended pre-focal damage, with apparent failure to induce significant damage within the targeted plaque. This case clearly demonstrates the importance of analysis of the treatment region before treatment. In our experience, the pre-procedural diagnostic ultrasound scanning provided essential sufficient data about the treatment region. This information can help in treatment planning in view of the anatomical constraints within the path of the HIFU beam. Both safety and efficient control are essential requirements for any proposed application to be implemented clinically in the future. Real-time feedback mechanisms need to be developed to provide instantaneous information regarding the ongoing treatment. However, we also believe that simulation models, like the one we are proposing herein, are essential to predict the outcome of the treatment in advance. This kind of prediction can give the treating physicians a higher degree of confidence in predicting how the treatment process will proceed. Furthermore, if the feedback data can be implemented in such simulation programs in real-time settings, this can provide physicians with enhanced visual perception and immediate evaluation of the ongoing treatment process. One of the limitations of our study for investigating targeting localised atherosclerotic plaques by HIFU is the limited number of treated plaques so far. We have previously reported in detail the results of treatment of four FH swine [25]. So far, we have not encountered any incidence of plaque rupture or distal embolism. While these results are promising in terms of demonstrating the safety profile of the proposed approach, we still need to take them with caution. Not all plaques have the same tissue composition. While some plaques have a more stable nature, other plaques may have significant central necrosis and friable fibrous caps. The latter plaques, known as ‘vulnerable plaques’, are more susceptible to rupture during the interventional procedures. Thus, more studies are still needed to investigate ablation of a wider spectrum of plaques. We have performed our experiments on FH swine not only because of close similarities in disease pathogenesis to humans but also due to analogous pattern of disease progression [41,42]. Based on this similarity with human atherosclerosis, we are optimistic that the successful results can be duplicated in humans in the near future.

Conclusion

The results provide an early validation for the feasibility of using image-based modelling of the acoustic and thermal field in heterogeneous tissues. Image-based modelling could play a critical role in treatment planning when HIFU is used in non-invasive precision lesion formation. In the first case, the simulation model correctly predicted the protection of the intima by the heat-sink effect due to blood flow in the targeted vessel. This could be significant when non-invasive HIFU is used in the treatment of vulnerable plaques. In the second case the model shows the formation of a lesion in the pre-focal plan instead of the target area. In this case, the retrospective analysis benefited from real-time image data to help explain the collateral damage and the apparent failure to form a lesion at the target. The retrospective analysis can provide added value to the quality of treatment follow-up based on risk analysis of collateral damage and confidence level of therapeutic outcome at the target.

Declaration of interest

Emad Ebbini is a consultant for the International Cardio Corporation, the company sponsoring this research. This relationship has been reviewed and managed by the University of Minnesota in accordance with its conflict of interest policies. The authors alone are responsible for the content and writing of the paper.