Neuronavigation-guided Transcranial Histotripsy (NaviTH) System

Abstract

This paper describes the development of the first neuronavigation-guided transcranial histotripsy (NaviTH) system and associated workflow for transcranial histotripsy ablations. The NaviTH system consists of a transcranial 360-element, 700 kHz transmit-receiver capable histotripsy array, a clinical neuronavigation system and associated equipment for co-registering the patient and histotripsy transducer, and therapy planning and targeting software systems. A workflow for NaviTH treatments, including pre-treatment aberration correction, was developed. Targeting errors stemming from target registration errors (TRE) during the patient-to-histotripsy transducer co-registering process, as well as focal shifts due to skull-induced aberrations were investigated and characterized. The NaviTH system was demonstrated in treatments of two < 96 h post-mortem human cadavers and in experiments in two excised human skullcaps. NaviTH was successfully used to create ablations in the cadaver brains as confirmed in post-treatment MRI. A total of three ablations were created in the cadaver brains and targeting errors of 9 mm, 3.4 mm and 4.4 mm were observed in corpus callosum, septum, and thalamus targets, respectively. Targeting errors were found to be primarily due to TREs resulting from transducer tracking instrument (TTI) design flaws and imperfections in the treatment workflow. TTI design and workflow improvements reduced TREs to < 2 mm, and skull-induced focal shifts, following pre-treatment aberration correction, were 0.27 mm. Total targeting errors of the NaviTH system following the noted improvements were 2.46 mm. In conclusion, we have demonstrated the feasibility of the first NaviTH system in a human cadaver model. While accuracy improvements are still required, the proposed system has the potential to allow for transcranial histotripsy therapies without requiring clinically burdensome active MR treatment guidance.

Introduction

Transcranial magnetic resonance-guided focused ultrasound (tcMRgFUS) thermal ablation has been approved by FDA to treat essential tremors and Parkinson’s Disease (Jung et al. 2018; Coluccia et al. 2014; Elias et al. 2016, 2013; Lipsman et al. 2013; Wintermark et al. 2014). Image guidance during the tcMRgFUS procedure is achieved via real-time MR thermometry and allows visualization of the focal ablation spot (Chung et al. 1999). However, MR guidance requires a long scanner time in the MR bore, is expensive, and limits clinical operational space to the bore of the magnet. Low amplitude tcMRgFUS for blood-brain barrier (BBB) opening has been performed to deliver therapeutic agents to the target tissue (Hynynen et al. 2005). There are numerous tcMRgFUS BBB opening trials ongoing for the treatment of brain tumors (NCT03739905, NCT03321487, NCT04370665, NCT05565443). To step away from MR guidance due to difficulty in wide clinical use and expensive cost, neuronavigation-guided transcranial focused ultrasound (NaviFUS) for BBB opening has been shown to demonstrate high targeting accuracy of ~2 mm to 4 mm (Chen et al. 2021; Wei et al. 2013). Ongoing clinical trials with this system are currently in progress for glioblastoma (NCT03626896) and epilepsy applications (NCT03860298).

Transcranial histotripsy uses microsecond-length, high-pressure ultrasound pulses, delivered through the skull, to mechanically fractionate brain tissues via cavitation (Xu et al. 2004; Parsons et al. 2006). The rapid growth and collapse of histotripsy cavitation produce high strains and stresses in the tissues in the focal region which fractionates them and reduces them to liquefied acellular homogenates (Vlaisavljevich et al. 2016). Transcranial histotripsy has shown great promise as a neurosurgical tool by demonstrating minimal heating of the skull (<4 °C) while effectively fractionating targeted brain tissues (Gerhardson et al. 2017). In vivo safety has also been demonstrated in mouse brain tumor models, as well as in vivo treatments of normal porcine brains (Gerhardson et al. 2020; Lu et al. 2021; Choi et al. 2020). Histotripsy cavitation clouds also emit acoustic shockwaves that can be captured using transmit-receive capable histotripsy arrays. Sukovich et al have demonstrated the real-time cavitation localization capability of the transcranial histotripsy array through an excised human skull (Sukovich et al. 2020), which provides a real-time cavitation and treatment monitoring tool for transcranial histotripsy without requiring active MR guidance.

Transcranial MR-guided histotripsy (tcMRgHt) has been also developed and shown its capability in treatments of in vivo porcine brain delivered through an excised human skull (Lu et al. 2021). tcMRgHt presents the same disadvantages as tcMRgFUS due to the high cost and the long MRI scanner time required, operational restriction for clinicians, and the requirement of MR-compatible equipment. Neuronavigation guidance presents a less costly alternative to active MR guidance. Neuronavigation systems are in wide clinical use and rely on co-registering pre-treatment scans of the patient with surgical devices (e.g., biopsy needles) and actively updating the co-registration during procedures to help surgeons accurately guide surgical devices to the therapy target. These systems are compatible with existing surgical tools and have been the hallmark guidance tool for neurosurgery for the past three decades (Orringer et al. 2012).

In this study, we integrate our transcranial histotripsy system with a clinical neuronavigation system to develop the first neuronavigation-guided transcranial histotripsy (NaviTH) system. We then demonstrate the feasibility and efficacy of the system to create transcranial histotripsy ablations in the cadaver brain for the first time. We introduce A) components of the NaviTH system, B) the workflow for transcranial treatment using the NaviTH system, and C) the setup and methods to evaluate the targeting error of the NaviTH system. The system targeting error of the NaviTH was divided into 1) target registration error (TRE) due to co-registration error and 2) focal shift due to skull-induced aberration after aberration correction. The TRE describes the discrepancy between true histotripsy focus location and the target projected by the neuronavigation system. Lastly, the feasibility and targeting accuracy of the NaviTH system were demonstrated in whole-body human cadavers and excised human skullcaps on red blood cell (RBC) phantoms. To the author’s knowledge, this is the first study that integrates the neuronavigation system with transcranial histotripsy.

Materials and Methods

This Section is presented in 3 parts: 1) components of the NaviTH system, 2) NaviTH workflow, and 3) the NaviTH system targeting evaluation setup. Abbreviations used throughout the paper are listed in Table 1.

Table 1:

Abbreviations

CT	Computed Tomography
MRI	Magnetic Resonance Imaging
NaviTH	Neuronavigation-guided Transcranial Histotripsy
TTI	Transducer Tracking Instrument
CRF	Coordinate Reference Frame
RRS	Retro-reflective Sphere
OPS	Optical Positioning System
TRE	Target Registration Error
AC	Aberration Correction

Components of NaviTH System

There are three main hardware components to the NaviTH system (Fig. 1A): the histotripsy array, the neuronavigation system, and the transducer tracking instrument (TTI). Other supporting components include the coordinate reference frame (CRF), the co-registration wand, the stereotactic frame, the water coupling system, and a 3D positioning system.

Figure 1:

Picture of the Neuronavigation-guided Transcranial Histotripsy (NaviTH) system (A) and the 360-element transcranial histotripsy array. The OPS recognized the unique geometry of the CRF, the TTI, and the co-registration wand to project the patients’ medical images, the histotripsy array focus, and the tip of the wand, respectively, all in one virtual 3D space. This allowed accurate real-time tracking of all instruments.

Transcranial histotripsy array with transmit-receive capabilities

A 360-element, 700 kHz, transmit-receive capable hemispherical histotripsy array with a focal distance of 150 mm was used for both cadaver and excised human skullcap experiments (Fig. 1B). The peak per-element negative pressure output of the transducers was measured to be 1.7 MPa, corresponding to a peak negative pressure output, P-, of the whole array (by linear summation of the individual elements) of > 600 MPa. Note, as cavitation exists 100% of the time at pressure > 30 MPa, direct pressure measurements above this value cannot be made; any reference to pressures > 30 MPa hereafter are thus intended only to provide an estimate of available pressure headroom during treatment. The array elements can be used to both transmit ultrasound pulses for histotripsy and receive ultrasound signals such as the acoustic cavitation emission signals.

Neuronavigation system, Co-registration wand, and CRF

A StealthStation^™ S7 surgical navigation system (Medtronic, Lafayette, CO) was used for image guidance of transcranial histotripsy. The navigation system co-registers the histotripsy array to the patient’s skull based on pre-treatment CT scans as follows. First, pre-treatment CT scans of the patient’s skull are loaded onto the neuronavigation system. A stereotactic headframe is then mounted to the patient’s head. A CRF is then attached to the stereotactic headframe. Mounted to the CRF is a series of retro-reflective spheres (RRSs) arranged in a unique geometry that identifies CRF. The neuronavigation system uses a stereoscopic infrared camera, the optical positioning system (OPS), to track the 3D locations of the RRSs and identify the associated device based on their geometric configuration. Once the CRF has been mounted to the patient’s head and identified by the neuronavigation system, a neuronavigation tool (the co-registration wand) is touched to the patient’s head and dragged/scanned over it by the clinician. During this process, the neuronavigation system records the position of the co-registration wand’s tip to generate a 3D point cloud representative of the exterior surface of the patient’s head. This point cloud is then co-registered with the CT of the patient’s skull, after which the location of the patient’s skull with respect to the CRF is known. Once this is complete, the neuronavigation system can then relate the positions and orientations of other tracked surgical devices to the CRF, and thus to the patient’s anatomy.

TTI

To allow neuronavigation-based optical tracking of the histotripsy array, a custom TTI, consisting of a geometrically unique plastic frame and associated RRSs, was designed and fabricated. Geometrical device descriptions in the neuronavigation system effectively define the position of the RRSs with respect to the ‘active’ region of the surgical tool (e.g., if the surgical tool was a biopsy needle, the geometrical device description would define the positions of the RRSs with respect to the needle tip). The geometrical device description of the TTI was thus defined such that, when attached to the histotripsy array, the ‘tip’ of the instrument would be co-aligned with the geometric focus of the array.

Water Coupling Tank, Stereotactic Frame, and 3D positioning system

A custom water coupling tank (TTP, Hertfordshire, United Kingdom) was designed and fabricated for the cadaveric experiments. This system consisted of a water tank with a hole in the side to allow insertion of the patient’s head, and a 3D positioning system mounted above the tank to which the histotripsy array could be attached (Fig. 1A, lower left). A custom stereotactic headframe consisting of a mounting bracket for the CRF, four set screws for rigidly attaching it to the cadaver head, and a mounting plate at the bottom to rigidly attach it to the water coupling tank was also designed and fabricated.

Workflow

In this study, the feasibility and targeting accuracy of the NaviTH system was first investigated using two ≤ 96 hours post-mortem, whole-body cadavers and two excised human skullcaps obtained from the Anatomical Donations Program (Michigan Medicine, University of Michigan, Ann Arbor, USA). The workflow steps are described with reference to the cadaver head or the excised skullcap, and the six steps to the workflow are described in detail in the following subsections:

1. TTI & Pre-treatment MRI and CT scans

2. Co-registration via neuronavigation

3. Patient positioning

4. Aberration Correction (AC)

5. Treatment delivery

6. Post-Treatment evaluation

TTI & Pre-operation MRI and CT Scans

Before treatment, CT scans of the excised skullcaps / cadaver heads were acquired; in the cadaver experiments MR images of the patients’ heads were additionally acquired. CT scans for both the cadaver and excised human skullcap studies were obtained to allow for CT-based acoustic aberration correction using Discovery CT750 HD (GE Medical Systems, Chicago, IL, USA) with the following imaging sequence: Stealth Protocol with reconstruction, 0.625 mm (1.25 mm original resolution) isotropic resolution. The MRI scans were acquired using a 3T MRI scanner (Ingenia, Philips, Amsterdam, Netherlands) with the following imaging sequences (T1, T2, T2*, ADC, and DWI). 3D Axial T1 (0.49 mm × 0.49 mm × 1 mm resolution) scans were used to co-register with the CT images to identify anatomical structures within the brain for treatment targeting.

Following their acquisition, the pre-treatment MRI and CT scans were uploaded into the neuronavigation system, along with a TTI geometry file containing information regarding the locations of the RRSs attached to it. The TTI is attached to the histotripsy transducer and used to co-register it the pre-treatment MRI and CT based on the positions of the TTI and CRF detected by the OPS. The design of the TTI is critical to minimize TRE, and critical design considerations are the distance from the RRSs to the tip, the rigidity of the instrument, and the RRSs’ positions relative to the “target”.

Co-registration via Neuronavigation

The stereotactic headframe, along with the CRF mounted to it, were rigidly affixed to the cadaver head. A “planar blunt tool”, i.e., the co-registration wand from the neuronavigation system, was then used to trace the contour of the cadaver head. The OPS was used to monitored the position of wand tip and generate a point cloud representative of the skull’s exterior surface based on the tracked position of the tip. The point cloud was then co-registered to the pre-treatment CT/MR scans by the neuronavigation system and subsequently co-registered to the CRF. The histotripsy array was then co-registered to the patient’s head by the neuronavigation system based on the position of the TTI with respect to the CRF, thereby finishing the co-registration step. This process was followed in the excised human skullcap study with the following modification. Instead of a stereotactic headframe, the CRF and excised skullcap were rigidly affixed to an acrylic mounting plate (described later) that could be mounted in a known position / orientation in the histotripsy array.

Patient Positioning

After co-registration, the patient’s head, mounted in the stereotactic frame, was positioned inside the water coupling system. A mounting bracket at the base of the stereotactic frame was used to secure the cadaver head to the water coupling tank and hold the cadaver head stationary within the tank. The water tank was designed with a circular hole in one of its walls to allow placement of the cadaver head within (Fig. 1A). A rubber membrane with a central circular hole cut in it to allow insertion of the cadaver head was mounted in this wall of the tank. Once the cadaver head was placed through the hole in the membrane into the tank, skin glue was used to create a water-tight seal between the membrane and the cadaver head. Following placement of the cadaver head within the tank, the histotripsy array was attached to a custom 3D positioning system, consisting of 3 screw-driven linear positioners, that was mounted above the tank. The array was mounted to the positioning system via a rigid bracket at the ‘top’ of the front face of the array, and suspended in the tank akin to a sideways bowl. During the cadaver experiments, the histotripsy array was mechanically repositioned using the 3D positioning system under neuronavigation guidance, to target a location of interest inside the cadaver brain. For the excised skullcap study, the acrylic plate with the attached skullcap and CRF was placed inside the histotripsy array and manually repositioned under neuronavigation guidance to target designated regions of interest.

Aberration Correction (AC)

Once the patient/skullcap has been positioned inside the histotripsy transducer, the two-step aberration correction method described by Lu et al. (Lu et al. 2022) was performed to correct skull-induced, per-element acoustic phase aberrations; skull-induced amplitude aberrations were not corrected for in this study. In the first step, CT-based aberration correction was performed using Kranion (Sammartino et al. 2019), an open-source ray-tracing software developed for transcranial focused ultrasound applications, to reduce aberration-induced focal shifts and pressure losses resulting from the histotripsy pulses traveling through the skull. First, the geometry of the histotripsy transducer, as well as the pre-treatment CT images were uploaded into Kranion. The relative positions of the histotripsy transducer and the head/skullcap were then retrieved from the neuronavigation system using the system’s StealthLink^™ API and used to align the head/skullcap and the array within the Kranion software. Kranion was then used to perform ray-tracing simulations of the acoustic pulses from the histotripsy transducer through the head/skullcap to calculate the relative delays in their arrivals at the transducer’s geometric focus; for the present work, a constant speed of sound (2300 m/s) in the skull bone was assumed for CT-based AC as done in previous work (Lu et al. 2022). Correction delays were then calculated based on the simulation results and applied to the firing times of the individual histotripsy transducer elements to temporally align their arrivals at the array’s geometric focus.

Following the CT-based AC, cavitation-based AC was applied. The CT-based aberration correction delays were applied to the transmit timings of the array elements, after which histotripsy pulses were delivered to generate cavitation through the skull. The shockwaves emitted during the collapses of the generated bubbles were then collected using the elements of the array as receivers. Aberration correction delays were then calculated based on the measured arrival times of the shockwaves at the individual array elements. This set of aberration correction delays was then applied to elements’ firing times during treatments. A full description of this two-step aberration correction approach can be found in (Lu et al. 2022).

Treatment Delivery

After the two-step aberration correction, for the cadaver study, histotripsy treatment was delivered to a 1 cm³ cubic volume comprised of a series of focal points arranged in a hexagonally close-packed (hcp) lattice (lattice-point spacing: 1.1 mm) using electronic focal steering at 200 Hz pulse repetition frequency (PRF) with 50 pulses per location. For the excised human skullcap study, histotripsy was delivered to generate cavitation in 3 × 3 × 3 = 27 locations (X, Y, Z = [−5, 0, 5] mm) using electronic focal steering at 5 Hz PRF with 10 pulses per location. The cavitation locations for the excised human skullcap study were chosen to mimic the extremities of the 1 cm³ volume treatment used for the cadaver study. A fiber optic probe hydrophone (HFO 690, Onda, Sunnyvale, CA, USA) was used to estimate the pressure levels used for the excised human skullcap experiment. The electronic voltage level used to drive the histotripsy array elements for both the cadaver and excised human skullcap experiments were identical, and therefore the pressure generated in the cadaver study was assumed to be similar to those of the skullcap study. In the excised human skullcap study, the skullcaps were translated and rotated to mimic various patient positions possible during treatment. The P- achieved within the skullcaps varied between 39 MPa to 61 MPa depending on the locations of the skullcaps in the histotripsy array.

Post-treatment Evaluation

After treatment, for the cadaver study, MRI head scans were collected to evaluate the ablation generated by histotripsy using the same 3T MRI scanner with the same imaging sequences as the pre-treatment MRI.

System Error Evaluation

The targeting accuracy of the NaviTH system was tested in two human cadavers and two excised human skullcaps. Table 2 lists the average and standard deviation of skull thickness in millimeters of each specimen and the skull density ratio (SDR) calculated by Kranion. Each step of the workflow for this study is detailed below.

Table 2:

Cadaver and Excised Human Skullcap Information: skull density ratio (SDR) and thickness

Specimen	Average SDR	Average Thickness
Cadaver 1	0.72 ± 0.16	7.02 ± 2.03 mm
Cadaver 2	0.69 ± 0.18	7.24 ± 5.43 mm
Skullcap 1	0.44 ± 0.10	6.27 ± 1.98 mm
Skullcap 2	0.59 ± 0.11	7.19 ± 2.28 mm

The sources of targeting error of the NaviTH system are primarily composed of two parts: 1) co-registration errors and 2) focal shifts induced by the skull aberration, and the corresponding errors arising from each are referred to as TRE and focal shift, respectively. The focal shift here refers to the difference between the positions of the planned focal target location of the histotripsy array and the actual location where cavitation was generated (after applying aberration correction). It should be noted that the overall targeting error in the initial cadaver study (described below) was found to be large (~5.6 mm), likely due to issues with the initial TTI design (version 1 or v1) utilized in those studies. Following the initial set of cadaver studies, however, an improved TTI (v2) was designed and built to reduce these errors, but due to COVID-19-related cadaver availability issues, it could only be tested in the excised human skullcap studies. A comparative study of the NaviTH system using both TTIs, and evaluating the different sources of targeting error, was conducted using the excised human skullcaps and is described below. The overall targeting error of the new NaviTH system was finally evaluated in a tissue-mimicking red blood cell (RBC) phantom through the excised human skullcap, where overall targeting errors for both the cadaver and the excised human skullcap study were defined based on the positional distances between the intended target location (i.e., the location of the transducer focus predicted by the neuronavigation system) and the resulting ablation location in the cadaver brain / RBC phantom.

Human Cadaver Study

The initial feasibility of the NaviTH system was tested in the brains of two whole-body cadavers obtained ≤96 h post-mortem following the treatment workflow steps described above. In the first cadaver, the array was aligned under neuronavigation guidance to target a posterior region of the corpus callosum. In the second cadaver, two target locations were selected for ablation, one in the thalamus and one in the septum pellucidum between the lateral ventricles. The intended treatment volumes were outlined on the pre-treatment MRI scan in the neuronavigation system and recorded as the intended target locations. The ablation volumes were identified in post-treatment MRI as hypointense regions in T1 and apparent diffusion coefficient (ADC) weighted images. The pre- and post-treatment MRI scans of the cadaver brain were co-registered via the neuronavigation system, and the distance between the center location of the intended treatment volume and the central location of the actual treatment volume in the pre- and post-treatment images was measured to obtain the total targeting errors.

Excised Human Skull Study

After the cadaver study, the targeting accuracy of the NaviTH system was evaluated using two ex vivo human skullcaps. The skullcaps were leased from the Anatomical Donations Program (Michigan Medicine, University of Michigan, Ann Arbor, MI, USA). However, Skullcap 1 had to be returned before the end of the study due to lease period restrictions of the Anatomical Donations Program. The targeting accuracy evaluation was divided into measuring target registration error (TRE) and the focal shifts resulting from skull-induced acoustic aberrations. Finally, overall system targeting accuracy was evaluated in treatments delivered to RBC phantoms.

Target Registration Error (TRE) Evaluation

One major source of targeting error for NaviTH is the positional error between the actual focus of the histotripsy array and its projected location based on the TTI as calculated in the neuronavigation system. The design of the TTI can thus have a major impact on the target registration error.

TTI

The TTIs were equipped with RRSs that enable the neuronavigation system to optically track the geometric pattern of the spheres in space and therefore the projected ‘tip’ (i.e., the histotripsy array focus) of the TTI device. The OPS, (a customized version of the Polaris camera, Northern Digital, Inc, Waterloo, Ontario, Canada), which is part of the neuronavigation system, is used for capturing the optical images of the RRS pattern unique to the TTI. A good RRS pattern should be asymmetric, the average distance from the RRSs to the target is short, and the intra-sphere distances are large (West and Maurer 2004). We constructed two TTIs (Fig. 2) for use in these studies and their geometries are summarized in Table 3.

Figure 2:

TTI v1 (left) and v2 (right). The numbers on the images denote the locations of the RRSs in the instruments.

Table 3:

Feature comparison of the two TTIs

	TTI v1	TTI v2
RRSs locations respective to the histotripsy focus	Arbitrary	3 spheres co-aligned with histotripsy focus and all spheres in the same plane as the focus
# RRSs	4	5
Average intra-RRS distances	50 mm	55 mm
Average distance of the RRSs to the histotripsy focus	333 mm	370 mm

For perspective, the average distance from the RRSs to the array focus in the NaviFUS system (the neuronavigation-guided FUS system currently in use in clinical trials for BBB opening), is approximately 150 mm (Wu et al. 2018). The large average distances from the RRSs to the target in our TTI designs were due to the constraints associated with our current water coupling system (Fig. 3A). Namely, the histotripsy transducer is submerged in a water tank and the TTI and CRF both need to be outside the water tank to be visible to the neuronavigation OPS for accurate tracking.

Figure 3:

Cadaver (A) and excised human skullcap (B) experimental setups. (A) The histotripsy transducer was placed inside a large acrylic water tank. Outside the water tank were the TTI and CRF with their RRSs facing the same direction for the OPS to visualize both instruments clearly for accurate tracking of instruments. This is the same setup depicted in Figure 1A. (B) The excised human skullcap study TRE evaluation setup had a similar configuration where the CRF was attached to the co-registered skullcap, and the skullcap was mechanically co-registered to the histotripsy transducer via a laser-cut, acrylic plate. The histotripsy transducer was equipped with the TTI to allow tracking of the geometric focus relative to the skullcap. The TTI is mounted in the same position on the histotripsy array in both (A) and (B).

TRE Evaluation Study

The TRE was evaluated using two excised human skullcaps. To evaluate the TRE, a focus structure (Fig. 4A & B) was designed in CAD software (SolidWorks, Dassault Systems, Velizy-Villacoublay, France) and 3D-printed (J750, Stratasys, Rehovot, Israel). The focus structure was designed with a pointed protrusion at its center that was located at the true geometric focus of the array (Fig. 4B). A laser-cut, 1/4” acrylic plate was rigidly attached to the histotripsy array’s front plate via a series of screws around its periphery (Fig. 4C). The acrylic plate provided a common platform for mounting the skullcap, focus structure, and the array scaffold to achieve mechanical co-registration (Fig. 4C) of the respective components, providing ground truth physical locations of the array focus and the skullcap for later comparison with neuronavigation co-registration results. Prior to each treatment, CT scans of the skullcap with the acrylic plate and the focus structure attached to it were acquired to enable neuronavigation-based co-registration with the histotripsy transducer. For treatments, the acrylic plate with attached skullcap and focus structure were rigidly mounted on the array scaffold via thumb screws, thus providing mechanically co-registered ground truth positions of the array focus and skullcap. Once in place, the neuronavigation system was used to co-register the histotripsy transducer with the skullcap based on the positions of the CRF and TTI devices attached to the acrylic plate and histotripsy transducer respectively. The neuronavigation co-registration information was then retrieved from the neuronavigation system and transferred to Kranion to perform CT-based AC. The co-registration results were also sent to 3D slicer (an open source image processing software, slicer.org) (Fig. 4D) which was used to evaluate the TRE based on measurements of the distance between the focus structure in the CT and the geometric focus of the histotripsy array following the co-registration-based alignment of the CT scans with the histotripsy array (Fig. 5).

Figure 4:

TRE evaluation setup. (A) and (B) provide a detailed view of the focus structure, and the setup view (C)provides how the skull-acrylic plate setup can be positioned respective to the transducer front plate. Three L-brackets were used to attach the skull to the acrylic plate and the acrylic plate connected to the histotripsy transducer front plate for mechanical co-registration. The skull anchors refer to the holes in the acrylic plate that attaches to the L-brackets. (D) The Kranion/Slicer view superimposes the co-registered element positions, the co-registered target (i.e. intended focus), and the CT of the skull and focus structure. The ‘intended target’ depicted in (D) is the co-registered histotripsy focus projected from the TTI, viewed in 3D Slicer. Histotripsy array elements are denoted with E-#, where # is the number of the element.

Figure 5:

Calculating the TRE. The Kranion scene, which includes the skull-focus structure CT, the transducer element locations, and the projected transducer focus (i.e. intended target), is exported into 3D Slicer, and the TRE is calculated from the focus structure central point, (i.e. geometric focus or the true target) to the ‘intended target’.

The acrylic plate, which contained the skullcap and connected to the front plate of histotripsy transducer, allowed translation and rotation of the skullcap with respect to the transducer to simulate patient positioning as well as treatments at different anatomical locations. A laser-cutter (VLS 6.60 laser systems, Universal Laser Systems, Inc. Scottsdale, AZ, USA) and 3D printer were used to manufacture the acrylic plate and TTI, respectively; tolerances for parts manufactured using both devices were 0.254 mm (0.01”).

The acrylic plate was designed with mounting features that allowed for lateral translations of the skullcap with respect to the array of up to 57.15 mm with 19.05 mm increments (2.25 inches with 0.75-inch increments), rotations of up to 75 degrees (25-degree increments about the superior-inferior axis of the skull), and axial translations of up 12.7 mm (0.5 inches). During experiments, the skullcaps were positioned / aligned in the array at a series of different locations allowable by the described mounting system. Owing to geometric constraints (e.g., collisions between the skullcap and transducer) and differences in the sizes of the two skullcaps, the TRE could not be evaluated at all possible mounting configurations of the acrylic plate to the histotripsy transducer. For Skullcap 1, 14 and 12 alignment locations were used to evaluate TRE using TTI v1 and TTI v2, respectively. For Skullcap 2, 20 alignment locations were used to evaluate TRE with TTI v2.

Focal Shift Evaluation Study

For the second part of the excised human skullcap study, the focal shift between the geometric focus of the histotripsy array and the measured treatment location was evaluated. Using the features of the acrylic front plate described in the TRE Evaluation Study section, the positions of the skullcaps with respect to the histotripsy array were varied (N = 16). At each position, 2-step AC was acquired prior to treatment.

Focal Shift Measurement

To calculate the focal shift induced by the skullcap, cavitation clouds were imaged by two cameras (Chameleon and Flea, Point Grey, FLIR, Wilsonville, OR, USA) mounted orthogonally to each other above the skullcaps facing towards the focal region of the transducer (Fig. 6). Histotripsy cavitation clouds were generated in a 3 × 3 × 3 grid of points through the skullcaps. Camera images of cavitation generated in the freefield and through the skullcaps were acquired and the distances between the locations of generated cavitation in each case were calculated to evaluate aberration-induced focal shifts. A detailed description of the method for measuring the 3D positions of the cavitation bubbles via the two camera approach may be found in (Sukovich et al. 2020).

Figure 6:

Focal shift evaluation setup. Two cameras were set up orthogonally to each other to extract bubble cloud locations. After the 2-step aberration correction, histotripsy cavitation clouds were formed via electronic focal steering through 3 × 3 × 3 = 27 locations (X,Y,Z = [−5, 0, 5] mm).

Note: full evaluations of aberration-induced focal shifts were only completed using Skullcap 2. Due to lease-duration restrictions of specimens acquired through UofM’s Anatomical Donations Program, only focal shifts due to lateral translations could be evaluated in Skullcap 1 before it needed to be returned.

Overall Targeting Error

The experiment to assess overall targeting error was performed through Skullcap 2. A red blood cell (RBC) gel phantom (Maxwell et al. 2010) was prepared to evaluate the system targeting accuracy with TTI v2. An RBC phantom holder was designed and fabricated using 3D printing and mounted to the acrylic plate at the geometric focus of the array. The phantom had cylindrical geometry with a radius of 16 mm and a height of 25 mm. The RBC phantom holder contained fiducial structures to allow localization of the histotripsy ablation point with respect to the geometric focus of the array. Through its attachment to the acrylic plate, the RBC phantom was mechanically co-registered with the skullcap and histotripsy array as described previously. An image of this setup is shown in Figure 7.

Figure 7:

System targeting error evaluation setup by red blood cell (RBC) phantom. The experimental setup for evaluating overall system targeting error is shown in (A). The RBC phantom is placed at the geometric focus of the histotripsy array and within the ex vivo skullcap. The system targeting error was calculated by (B) comparing the intended target (i.e. neuronavigation target) and the center of the histotripsy ablation point in the post-treatment MRI.

The workflow for the RBC phantom treatment was similar to the cadaver workflow presented at the beginning of the Methods:

1. The phantom was imaged with a 7T MRI scanner (Varian, Inc., Palo Alto, CA, USA) using a T2-weighted fast spin-echo sequence (FOV = 25 × 25 × 30 mm, resolution = 0.2 × 0.2 × 1 mm, TR = 3.8 s, ESP = 15 ms, Kzero = 3 (TE = 45 ms)) at axial, coronal, and sagittal orientations to locate the fiducial structures in the phantom and therefore, localize the geometric focus of the array.

2. The physical skullcap and the skullcap CT containing the acrylic plate were co-registered to the CRF by the neuronavigation system.

3. The skullcap was positioned in the array. Once both the TTI and CRF were in the neuronavigation field of view, the intended target (i.e., histotripsy array focus) projected by the TTI was visible respective to the focus point of the focal structure.

4. Two-step aberration correction was applied.

5. The RBC phantom was placed inside the skullcap (Fig. 7A) and 200 histotripsy pulses were delivered to a single point at a PRF of 5 Hz.

6. Post-treatment MRI of the treatment RBC phantom was acquired.

The RBC phantoms (N = 7) were made as previously described to visualize histotripsy damage (Choi et al. 2020; Maxwell et al. 2010; Allen et al. 2017). 1.5% agarose gel was prepared (Maxwell et al. 2010) by mixing degassed saline and agarose powder (DSA20070, Dot Scientific, Burton, MI, USA). When the liquid gel had cooled to below 40 °C, it was mixed with fresh bovine blood (Dunbar Meats, Milan, MI, USA), poured into the RBC phantom holder, and allowed to solidify.

Following treatment, the targeting accuracy of the NaviTH system was evaluated by comparing the intended treatment target with the center of the ablation zone identified in post-treatment MRI (Fig. 7B). Treatments were delivered with the skullcap positioned in seven locations with respect to the array to evaluate targeting accuracy as a function of treatment location within the skullcap.

Results

Cadaver Experiment

The feasibility of the NaviTH system was demonstrated in two human cadavers < 96 hours post-mortem. Three ablations of 1 cm³ were created in three different locations (corpus callosum, septum pellucidum, and thalamus) in the brains of two cadavers. The ablation zones were identified by post-treatment MRI (Fig. 8). The shape and size of the ablation zones matched well with those of the intended treatment volume. 3D Axial T1 MRI best depicted the lesions created near or in the ventricle (Fig. 8A & B) due to its high resolution. ADC MRI images best depicted the histotripsy lesion in the brain parenchyma/thalamus (Fig. 8C).

Figure 8:

Cadaver experiments post-treatment MRI with NaviTH system. MR images best representing the lesions created in the cadavers are shown. A) presents a sagittal slice of the post-treatment T1 MRI of Cadaver 1. The corpus callosum was targeted. For Cadaver 2, two 1 cm³ lesions, one in the septum pellucidum and another in the thalamus, were generated (B & C). B) is a sagittal slice post-treatment T1 MRI of the septum pellucidum lesion and C) is an axial slice of the post-treatment apparent diffusion coefficient (ADC) MRI image of the thalamic lesion created in Cadaver 2. The red arrows point out the volume ablation created with NaviTH.

The overall targeting error for the two cadavers using the TTI v1 NaviTH system is presented in Table 3. The targeting error ranged between 3.4 mm to 9 mm with an average error of 5.6 mm. We attribute the source of this large targeting error to the design of TTI v1, which resulted in large TRE (described below), as well as imperfections in the workflow early in this pilot study.

TRE Evaluation

The TRE of the two different TTIs’ was evaluated through two ex vivo human skullcaps (Fig. 9). The TRE resulted from the errors due to patient-to-image co-registration, the TTI-to-array focus tracking errors, and transformation errors between the neuronavigation and Kranion image spaces. For Skullcap 1, the TRE was 8.27 mm (x = 3.97 ± 0.56 mm, y = 6.77 ± 2.45 mm, z = 2.62 ± 1.33 mm) for N = 14 locations tested using the TTI v1 versus 1.29 mm (x = 0.77 ± 0.51 mm, y = 0.43 ± 0.23 mm, z = 0.95 ± 0.25 mm) using the TTI v2 for N = 12 location tested. For Skullcap 2, TRE was only measured with the TTI v2 and was 2.49 mm (N = 20 locations, x = 2.06 ± 0.47 mm, y = 0.92 ± 0.57 mm, z = 1.05 ± 0.27 mm). This large improvement in the TRE is attributed to the difference in the geometry of the two TTIs, in particular the co-alignment of the 3 RRSs with the geometric focus of the histotripsy array in TTI v2.

Figure 9:

TRE measurements with the two TTIs. The Skullcap 1 TRE by TTI v1 and v2 were 8.27 mm and 1.29 mm, respectively, and the TRE of Skullcap 2 by TTI v2 was 2.49 mm.

Focal Shift Due to Skull-Induced Aberrations

Lateral (XY) offsets of the cavitation clouds due to skull-induced acoustic aberrations were measured in Skullcap 1, and lateral and axial (XYZ) offsets of the clouds were measured in Skullcap 2 (Fig. 10). The total offsets were 0.28 mm (x = 0.13 ± 0.11 mm, y = 0.25 ± 0.15 mm) in Skullcap 1, and 0.27 mm (x = 0.09 ± 0.03 mm, y = 0.11 ± 0.03 mm, z = 0.24 ± 0.14 mm) in Skullcap 2, and no bias of the error was observed. In Skullcap 1, the lateral average focal shift (0.28 mm) was larger than observed in Skullcap 2 (0.15 mm). Correspondingly, Skullcap 1 had a lower SDR (Table 1), indicating more skull heterogeneity.

Figure 10:

Focal shifts due to the skull-induced acoustic aberrations measured with two orthogonally placed cameras. CT-based and cavitation-based AC were performed, and histotripsy cavitation clouds were formed at locations in a 3 × 3 × 3 grid. The focal shifts were assessed based on the measured distances between the centroids of the cavitation clouds in the free-field and transcranial cases. Only the lateral (XY) focal shifts could be assessed through Skullcap 1 due to time constraints related to lease-duration restrictions of specimens acquired through UofM’s Anatomical Donations Program.

Targeting Accuracy of the NaviTH System Using the TTI v2

The total error of the NaviTH system using the TTI v2 was assessed by treating the RBC phantoms (N = 7 locations) as the skullcap (Skullcap 2) was positioned at various locations in the array. The total targeting error of the system was measured to be 2.46 mm (x = 2.30 ± 0.55 mm, y = 0.33 ± 0.21 mm, z = 0.80 ± 0.18 mm) (Fig. 11), similar in magnitude to the TRE of Skullcap 2 which was 2.49 mm. The component contributions of the error in the X, Y, and Z directions were also observed to be similar to the distribution of Skullcap 2 TRE, suggesting that TRE (as opposed to focal shifts) was the dominant source of overall targeting error in this study.

Figure 11:

TTI v2 NaviTH System targeting error on RBC gel phantoms. (A) A total of seven RBC gel phantoms were ablated using the NaviTH system equipped with the TTI v2 to produce a system targeting error of 2.46 mm with Skullcap 2. B) shows two cases of T2w MRI axial and sagittal views of RBC gel phantoms after treatment by the NaviTH system.

Discussion

The first neuronavigation-guided transcranial histotripsy (NaviTH) system and its workflow were developed. The NaviTH system was used to treat the brain in a human cadaver model. Total targeting errors of the system were evaluated through excised human skullcaps. Two primary sources of system targeting error were identified and evaluated: 1) target registration error (TRE), i.e., errors in the co-registration between the head and the histotripsy array using the neuronavigation system, and 2) focal shifts due to skull-induced acoustic aberrations. TRE was found to be the largest source of error in this study, as the 2-step aberration correction approach was used to correct for acoustic aberration and reduce the aberration-induced focal shift. The primary source of error in the TRE was determined to be design flaws in the custom transducer tracking instruments.

The NaviTH system was demonstrated in the brains of two < 96 hours post-mortem human cadavers. The targeting error in the cadaver experiments was large (4 mm to 9 mm) but was later reduced to 2.46 mm, slightly above the manufacturer’s stated mean error of < 2 mm, in the excised human skullcap studies by improving the design of the TTI from v1 to v2. While cadaver availability limitations prevented the testing of TTI v2 in cadaver experiments, as the errors observed through the excised human skullcaps using TTI v1 (~8 mm) agreed closely with those in the cadaver experiments, it is expected that the improved targeting accuracy afforded by using TTI v2 in the excised skullcap studies will translate to cadaver model. Workflow and non-TTI-related equipment issues were also identified that likely led to the large errors observed in the cadaver experiments. For example, the stereotactic headframe used to mount the cadaver head within the water tank/transducer was designed based on consideration of the head size of an average adult human (circumference of 55 cm to 57 cm) (Ching 2007). While the set screws used to fix the head’s position within the headframe allowed for the accommodation of larger heads, both cadavers had head sizes significantly smaller than the average adult, particularly cadaver 1, and were near the limits of the set screws’ travel length. This likely resulted in insufficient clamping force on the heads to ensure complete immobility throughout the duration of the experiments.

The targeting accuracy of the NaviTH system was tested on RBC gel phantoms with TTI v2. The total error of the system was measured to be 2.46 mm with the improved TTI v2, which is slightly above the expected targeting error specified by the manufacturer (~2 mm) (Medtronic 2022). In all studies, the total targeting error was dominated by TRE, which itself was most influenced by the design of the TTI. The TTI design changes from TTI v1 to TTI v2 had a major impact on the targeting error. The main change was in the alignment of the RRS on the TTI frame, which went from ‘randomly’ arranged, to aligned coaxially with the transducer focus. Further improvements to the design of the TTI may further reduce target registration errors. It is also hypothesized that small imperfections in the TTI, e.g., warping of the 3D-printed plastic it was made from, could have resulted in significant errors in the projected focal point of the array given the large distance between the focal point and the RRSs visible to the OPS. The transducer scaffold itself is also machined from plastic (Delrin) and subject to potential warping over time. The aluminum front plate of the transducer, to which the acrylic plate used for mechanically co-registering the skullcaps to the array was mounted, also had loose tolerances which may have contributed to the errors observed in the excised skullcap experiments. The TRE may also be further improved by incorporating a step to verify the TTI using a focal pointer structure, similar to that used to calibrate the co-registration wand. This focal structure can additionally be used to mechanically calibrate the transducer focus with respect to the TTI before each treatment.

The average TRE for Skullcap 1 was 1.2 mm smaller than that of Skullcap 2. A 1/4” drill hole was present in Skullcap 1, which may have served as a fiducial marker for co-registration by the neuronavigation system. Fiducial markers, or skull screws, are known to minimize the TRE for neuronavigation co-registration (Perrin et al. 2009; Pinggera et al. 2018). The manual focus calibration procedure described above in combination with the skull screws for fiducial markers may improve the TRE and will be considered for optimizing TRE for NaviTH in future cadaveric experiments.

After the evaluation of co-registration accuracy, the focal shift induced by the skullcaps post-aberration correction was independently analyzed. An average focal shift of 0.27 mm was observed between the locations of cavitation generated in the free-field and through the skullcaps after aberration correction. Most of the shift was observed in the z-axis (Fig. 10) of the transducer for Skullcap 2, as observed in prior studies (Sukovich et al. 2020; Lu et al. 2022). Although the observed focal shifts were minimal, as CT-based AC requires accurate alignment of the skull within the transducer, the TRE likely reduced the efficacy of the CT-based corrections. In addition, the CT-based AC used for this study assumed a constant speed of sound in the skull bone of 2300 m/s. However, it is known that the cortical and trabecular bones within the skull vary in speed of sound (Pichardo et al. 2010). An improved co-registration, TRE, and CT-based AC may further minimize the skull aberration-induced focal shift.

In this study, a treatment workflow for the NaviTH system was developed. However, the number of steps involved in the workflow can be reduced if the neuronavigation system (co-registration) and CT-based AC (Kranion) can be combined into one platform, similar to previous work (Chaplin et al. 2019). For example, one might imagine building a fully self contained transcranial histotripsy system wherein co-registration with prior CT scans could be accomplished based on stereoscopic images of the head acquired from cameras embedded in the array itself instead of based on the relative positions external tracking devices mounted to it and the patient, after which a singular software platform capable of performing CT-based AC and other treatment planning and monitoring operations could be used to control treatments. Although such a system would require significant work to develop, it would greatly simplify its use and help increase its adoption by minimizing operational complexities associated with running multiple devices.

A limitation of this study was the low number of cadavers in which the NaviTH system was tested, especially in relation to our inability to perform further cadaver treatments using the NaviTH system with the TTI v2 and evaluate targeting accuracy using it. Unfortunately, owing to heightened screening and operational restrictions following the COVID-19 pandemic, cadaver availability has been significantly reduced compared to when initial work on this study began, particularly in the ≤ 96 h window where effects such gas buildup in the tissues remain limited enough to allow effective ultrasound delivery to the treatment targets. While the general agreement between the TRE results in the cadaver experiments and the excised human skullcaps using TTI v1 suggests the results from the excised human skullcaps will translate to the cadaver case, validation is still required. More experiments are needed to evaluate the robustness, efficacy, and treatment envelop of this NaviTH system in the brain, which will inform the capabilities and limitations of the NaviTH as a potential neurosurgical tool and bring transcranial histotripsy closer to the clinic.

Conclusions

This study presented the first NaviTH system, and its feasibility was demonstrated in the brain of human cadavers and through excised human skullcaps. Ablations were successfully targeted under neuronavigation guidance to the desired anatomical locations in the cadaver brains. While targeting errors in the cadaver models were large, experiments in excised human skullcaps allowed us to determine the largest source of targeting error was the design of the TTI device used by the neuronavigation system to co-register the histotripsy array with the patients head. Subsequent redesigns of the TTI reduced targeting error by more than a factor of 2 from > 5 mm using TTI v1 to 2.5 mm using TTI v2. Workflows for carrying out neuronavigation-guided histotripsy were also developed. The efficacy of using the two-step aberration correction approach for improving focal targeting accuracy and pressure through the skull was extended for use in NaviTH applications and demonstrated in both the cadaver and excised human skullcap models.

Data Availability Statement

The raw data corresponding to the results presented in this manuscript will be provided upon request.