Ultrafast 3-D ultrasound super-resolution imaging with a row-column array

Abstract

Row-column arrays are an emerging technology which aims to produce 3-D ultrasound images with a wide field of view and low channel count. The images produced with these types of probes suffer from strong artefacts and poor image quality when operated at high frame rates. This work aimed to improve the quality of the vascular images this probe could produce in vivo, via the implementation of ultrasound localisation microscopy. To do this a new coherence-based image reconstruction technique was developed. In this technique, the intrinsic incoherence of the row-column array architecture was exploited to reduce the artefact level, whilst also increasing the resolution and decreasing the noise. To further reduce the noise and increase the frame rate, acoustic sub-aperture processing was combined with a rolling reconstruction technique which enabled an effective frame rate of twice the pulse repetition frequency whilst also reducing the noise level. New row-column array-tailored isolation and localisation algorithms were developed and compared with preexisting reference techniques. The above algorithms were evaluated and optimised in silico, in vitro and in vivo on a rabbit’s kidney and on human volunteers’ thyroid and carotid artery. The new image reconstruction techniques were found to increase the signal-to-noise ratio by ~7.3 dB and reduce the artefact level by up to ~22 dB when compared to the traditional ultrafast reconstruction techniques. In vitro, it was found that the above techniques reduced the number of false localisations from ~26 % to ~15%. The new localisation techniques were found to increase the accuracy, precision and recall of super-localisation compared to a preexisting reference approach. However, the improvements came at a computational or usability cost. The above work was combined and optimised to produce an ultrasound localisation microscopy pipeline that could produce super-resolution imaging non-invasively in vivo using a row-column array in clinically reasonable timeframes.