3D printed calibration micro-phantoms for super-resolution ultrasound imaging validation

This study evaluates the use of 3D printed phantoms for 3D super-resolution ultrasound imaging (SRI) algorithm calibration. The main benefit of the presented method is the ability to do absolute 3D micro-positioning of sub-wavelength sized ultrasound scatterers in a material having a speed of sound comparable to that of tissue. Stereolithography is used for 3D printing soft material calibration micro-phantoms containing eight randomly placed scatterers of nominal size 205 μm 205 μm 200 μm. The backscattered pressure spatial distribution is evaluated to show similar distributions from micro-bubbles as the 3D printed scatterers. The printed structures are found through optical validation to expand linearly in all three dimensions by 2.6% after printing. SRI algorithm calibration is demonstrated by imaging a phantom using a /2 pitch 3 MHz 62+62 row-column addressed (RCA) ultrasound probe. The printed scatterers will act as point targets, as their dimensions are below the diffraction limit of the ultrasound system used. Two sets of 640 volumes containing the phantom features are imaged, with an intervolume uni-axial movement of the phantom of 12.5 μm, to˜ emulate a flow velocity of 2 mm/s at a frame rate of 160 Hz. The ultrasound signal is passed to a super-resolution pipeline to localise the positions of the scatterers and track them across the 640 volumes. After compensating for the phantom expansion, a scaling of 0.989 is found between the distance between the eight scatterers calculated from the ultrasound data and the designed distances. The standard deviation of the variation in the scatterer positions along each track is used as an estimate of the precision of the super-resolution algorithm, and is expected to be between the two limiting estimates of (σ_x, σ_y, σ_z) = (22.7 μm, 27.6 μm, 9.7 μm) and (σ_x, σ_y, σ_z) = (18.7 μm, 19.3 μm, 8.9 μm). In conclusion, this study demonstrates the use of 3D printed phantoms for determining the accuracy and precision of volumetric super-resolution algorithms.