A dual-mode hemispherical sparse array for 3D passive acoustic mapping and skull localization within a clinical MRI guided focused ultrasound device

Abstract

Previous work has demonstrated that passive acoustic imaging may be used alongside MRI for monitoring of focused ultrasound therapy. However, past implementations have generally made use of either linear arrays originally designed for diagnostic imaging or custom narrowband arrays specific to in-house therapeutic transducer designs, neither of which is fully compatible with clinical MR-guided focused ultrasound (MRgFUS) devices. Here we have designed an array which is suitable for use within an FDA-approved MR-guided transcranial focused ultrasound device, within the bore of a 3 Tesla clinical MRI scanner. The array is constructed from 5 × 0.4 mm piezoceramic disc elements arranged in pseudorandom fashion on a low-profile laser-cut acrylic frame designed to fit between the therapeutic elements of a 230 kHz InSightec ExAblate 4000 transducer. By exploiting thickness and radial resonance modes of the piezo discs the array is capable of both B-mode imaging at 5 MHz for skull localization, as well as passive reception at the second harmonic of the therapy array for detection of cavitation and 3D passive acoustic imaging. In active mode, the array was able to perform B-mode imaging of a human skull, showing the outer skull surface with good qualitative agreement with MR imaging. Extension to 3D showed the array was able to locate the skull within ±2 mm/2° of reference points derived from MRI, which could potentially allow registration of a patient to the therapy system without the expense of real-time MRI. In passive mode, the array was able to resolve a point source in 3D within a ±10 mm region about each axis from the focus, detect cavitation (SNR ~ 12 dB) at burst lengths from 10 cycles to continuous wave, and produce 3D acoustic maps in a flow phantom. Finally, the array was used to detect and map cavitation associated with microbubble activity in the brain in nonhuman primates.

Figure 1

Overview of experimental setup with the completed 128 element array. (a) Top view of ExAblate 4000 low frequency therapeutic transducer. (b) Overview of array design. Elements were attached to acrylic ribs held in place by a ring-shaped fixture frame. Wiring was routed to a junction box fitted with industry standard array connectors. The assembly was attached to a mounting plate to allow installation and removal from the therapy transducer as a single unit. Inset shows side view of array. (c) Array fitted to therapy transducer. (d) Impedance (phase) response of a sample piezo element from the array, showing radial and thickness resonance modes. (e) Photograph (left) and outline of design (right) for hydrophone mount used in calibration. (f) MRI of the hydrophone mount used to register the array to MR imaging space. Scale bar = 1cm.

Figure 2

B-mode imaging of human skull using 18-element central rib of array. (a) B-mode image produced using the array. The position of the elements and envelope of the skull can be seen. (b) MRI of the skull from the same slice as the B-mode image. The outline of the center rib of the array can be seen within the transducer. The skull appears as a signal void relative to the water. (c) Segmented skull from MRI. (d) Segmented skull from MRI overlaid on ultrasound B-mode image.

Figure 3

3D skull localization using 48 element 3D array. (a) 3D scatter plot of segmented points corresponding to the outer surface of the skull from MRI data. (b) Pulse-echo ultrasound data for each element. For display purposes the A-lines for each of the elements are shown side-by-side to create a quasi- B-mode image. The identified arrival times are shown by the red crosses. (c) Illustration of the relative positions of skull and array elements. Based on a guess of skull position a set of implied distances is calculated. For clarity, a reduced set of the skull points is shown. (d) Effect of manual translation or rotation about one axis on the accuracy of the algorithm. The error vs. ground truth after running the algorithm is shown as a function of the initial offset in position. (e) Initial random guess of skull position vs. ground truth. (f) Result of running the alignment algorithm on the initial guess shown in (e).

Figure 4

Acoustic modelling results for simulated 128 element 3D array. Effects of (a) noise (b) position error and (c) uneven element sensitivity. Columns show: selected acoustic maps for four values of the variable under test; transverse profile of the maps; the map signal to noise ratio (SNR), peak sidelobe ratio (PSR) and transverse full-width half maximum (FWHM).

Figure 5

Point source localization results from 48 element 3D array. The transmitter was moved along the (a–b) X, (c–d) Y and (e–f) Z axes and maps reconstructed in a ±10 mm range about the 3 principal axes. (a, c, e) shows example maps about the 3 planes; (b, d f) shows profiles extracted from the maps, which were normalized with respect to the maximum value from all five profiles. The values of x, y, and z reported in (a, c, e) give the position of the peak in the acoustic maps when the source was moved in the corresponding direction.

Figure 6

Cavitation source experiment. A tube phantom containing water or microbubbles was sonicated with the therapy array and monitored using the 128-channel array. (a) RF data pre- (top) and post- notch filter (bottom), for control (left) and microbubble (right) experiments. (b) Average difference in spectra between the bubble and control experiments. The data were averaged over the array and sonication duration. (c) Amplitude of second harmonic over time. The sonication on-time is shown by the green overlay, and approximate start of manual microbubble injection by the black dashed line. (d) Representative 3D PAM for sonications with (top) and without (bottom) microbubbles. (e–f) Spatial peak, temporal sum value from PAM vs. (e) electrical power and (f) burst length. CW = continuous wave.

Figure 7

In vivo results using 48 element 3D array. (a) Left: MR temperature map acquired during sonication, showing low-level temperature rise that can occur during sonication with microbubbles (McDannold et al 2006); Right: T2*-weighted MRI acquired after treatment, showing a small lesion created at the focus induced by cavitation activity. Scale bar = 1cm. (b) Comparison of averaged spectra over the duration of sonications with and without microbubbles. The averaged spectra from a sonication with microbubbles was subtracted from that without. (c) Magnitude of the second harmonic over time with and without microbubble injection. The sonication on-time is shown by the green overlay, and approximate start of manual microbubble injection by the black dashed line. (d–e) Acoustic maps reconstructed over the three principal planes, integrated over the duration of sonications (d) with and (e) without microbubbles. The two sets of maps are shown in the same color scale.