Medical ultrasound is a wide-spread imaging modality due to its high temporal resolution, lack of harmful radiation and cost-effectiveness, which distinguishes it from other modalities such as MRI and CT. High temporal resolution is highly desired in additional to spatial resolution for certain ultrasound examinations. One of the most commonly used techniques for high frame rate ultrasound is multi-line acquisition (MLA) often referred to as parallel receive beamforming (PRB). While ML transmission reduces the acquisition time, it comes at the expense of a heavy loss of contrast due to the interactions between the beams.
The technology is based on a data-driven learning approach to improve MLA image quality. A convolutional neural network trained on pairs of real ultrasound raw data, acquired through MLA and the corresponding high-quality, single-line acquisition (SLA) image. Post training, the network can be used to reconstruct high- quality images given an MLA image.
- Imaging is performed on a rapidly moving organ
- Significant improvement of image quality
- Since our method is data-driven it is more adjustable to different ultrasound scenes irrespective of the machine and works well for high frame rates
Applications and Opportunities
- Can be deployed onto ultrasound scanners to improve imaging quality in fast acquisition modes