Flexible ultrasound transceiver array for non-invasive surface-conformable imaging enabled by geometric phase correction

Abstract

Ultrasound imaging provides the means for non-invasive real-time diagnostics of the internal structure of soft tissue in living organisms. However, the majority of commercially available ultrasonic transducers have rigid interfaces which cannot conform to highly-curved surfaces. These geometric limitations can introduce a signal-quenching air gap for certain topographies, rendering accurate imaging difficult or impractical. Here, we demonstrate a 256-element flexible two-dimensional (2D) ultrasound piezoelectric transducer array with geometric phase correction. We show surface-conformable real-time B-mode imaging, down to an extreme radius of curvature of 1.5 cm, while maintaining desirable performance metrics such as high signal-to-noise ratio (SNR) and minimal elemental cross-talk at all stages of bending. We benchmark the array capabilities by resolving reflectors buried at known locations in a medical-grade tissue phantom, and demonstrate how phase correction can improve image reconstruction on curved surfaces. With the current array design, we achieve an axial resolution of ≈ 2 mm at clinically-relevant depths in tissue, while operating the array at 1.4 MHz with a bandwidth of ≈ 41%. We use our prototype to image the surface of the human humerus at different positions along the arm, demonstrating proof-of-concept applicability for real-time diagnostics using phase-corrected flexible ultrasound probes.

Figure 1

Overview of the FlexArray. (a) 3D rendering of the flexible array with mounted piezoelectric transducers. (b) Scanning electron micrograph of diced PZT pillars on the edge of an array quadrant. The pillars are contacted with a Cr/Au top metal layer. Scale bar, 200 μm. (c) Scanning electron micrograph presenting the surface morphology of the gold-covered PZT crystals. Scale bar, 2 μm. (d) Energy dispersive X-ray spectrum of the pillar, demonstrating the elemental content of the fabricated transducer. (e) The as-manufactured flexible PCB used to house the piezoelectric array in the center, seen held in a neutral position. (f) The diagonal connector design allows for the easy bending of the array by hand. The board is flexible enough to also support (g) convex bending, and (h) shear bending.

Figure 2

PZT pillar fabrication and ultrasound performance benchmarks of the FlexArray. (a) Photograph of the bonded PZT array quarters after back-side dicing to separate out the individual transducer pillars. (b) Cross-sectional illustration (not to scale) of the finalised pillar mount. Cu electroplating is used to raise the metal connection before bonding with the bulk PZT transducer on the ACF substrate. Parylene isolates the signal pads from the electrical traces on the board. (c) Photograph of the completed FlexArray, relative to a human hand. (d) Close-up photograph of a bent piezoelectric array in the center of the device after metallisation and encapsulation. (e) Examples of pulse-echo responses, and their associated frequency spectra as insets, from arbitrary pillars in each quarter of the array. The temporal envelope for three pulses is ~ 4 μs with a wide bandwidth of ~ 41%. (f) Ultrasonic pressure generated by the FlexArray in water when focused at f = ∞ (blue) and at f = 2 cm (red), across a range of ultrasound frequencies. The pressure value peaks at over 600 kPa. The inset shows the linear scaling of the generated ultrasonic pressure with applied peak-to-peak voltage across the transducer elements.

Figure 3

Phase steering of ultrasound beams with the FlexArray. (a, Photograph of the experimental setup used to record the distribution of pressure in a water tank at a focal length of 2 cm. Inset: illustration of the detector geometry in the water tank (not to scale). (b–d) Simulated pressure maps at this focal length when the array elements are excited at − 20°, 0° and 20° of phase shift, respectively. (e–g) Experimental pressure maps collected with the FlexArray at the same respective phase shift values. Note that all colour maps are self-normalised to the colour bar on the right of the figure.

Figure 4

Cross-talk characterisation between neighbouring elements and medical-grade flat phantom imaging. (a) Photograph of the detachable 3D-printed apparatus for the testing of electrical cross-talk across different radii of curvature. (b) Power spectra of the voltage response of a single powered transducer (green), relative to those of its nearest neighbour element (blue) and its second nearest neighbour (pink). Extreme cases of a highly-curved array (R = 1.5 cm) and a flat array (R = ∞) are compared. (c) Cross-talk between neighbouring elements, extracted at 1.4 MHz, as a function of radius of curvature of the FlexArray. The device shows excellent inter-channel shielding characteristics at high operational frequencies, regardless of shape. (d) Photograph of the cross-section of the tissue phantom with embedded nylon wires (white dots). The yellow dot marks the vertical location where the FlexArray was placed during imaging. (e) B-mode scan of the phantom taken with the FlexArray, demonstrating both its axial and lateral resolution. The locations of the nylon rods can be resolved axially at ~ 1 cm apart. (f) Plot of the theoretical, simulated, and measured axial/lateral resolutions as a function of scatterer depth. Note that error bars are smaller than the marker size of the experimental data points. In both simulations and experiments, the number of x–z raylines is 96.

Figure 5

Effects of phase correction on image formation tested on gelatin phantoms of arbitrary curvature. (a) Photograph of the cast gelatin phantom used to collect B-mode images at arbitrary curvatures. The FlexArray is seen conforming to the surface. (b) Plan-view photograph of the air pocket-containing gelatin mould. (c) Simulated B-mode image when the radius of curvature is 1.5 cm and phase correction is not implemented. (d) Experimental B-mode image of an air pocket in gelatin captured with the FlexArray when the radius of curvature is 1.5 cm; without phase correction. (e) Simulated B-mode image when the radius of curvature is 1.5 cm and phase correction is implemented. (f) Experimental B-mode image of an air pocket in gelatin captured with the FlexArray when the radius of curvature is 1.5 cm; with implemented phase correction. Both images in (d) and (f) are compound averaged focal series images between f = 2, 3 and 4 cm.

Figure 6

Imaging of the human humerus with the FlexArray. (a) Close-up photograph of the array area on the arm during imaging, also showing the 2D plane along which the scan was taken. (b) Photograph of the subject’s arm with the device in the Level 2 position. Three levels were chosen to image the bone along the length of the arm. (c, d) Level 1 B-mode images of the humerus without and with correction for the radius of curvature of 4 cm, respectively. Analogous images were taken at (e, f) Level 2 and (g, h) Level 3.