Tiny-VBF: Resource-Efficient Vision Transformer based Lightweight Beamformer for Ultrasound Single-Angle Plane Wave Imaging
Abdul Rahoof, Vivek Chaturvedi, Mahesh Raveendranatha Panicker, Muhammad Shafique
TL;DR
The paper tackles the challenge of real-time, resource-efficient ultrasound beamforming on edge devices by introducing Tiny-VBF, a vision-transformer–based lightweight beamformer that processes single-angle plane-wave RF data to produce IQ beamformed images. It demonstrates a highly efficient implementation with a target of $0.34\ \mathrm{GOPs/Frame}$ for $368\times128$ frames and presents an FPGA-friendly accelerator with a hybrid quantization scheme that reduces resource use by over $50\%$ while preserving image quality. Benchmarks on in-silico and in-vitro data show competitive improvements in contrast and axial/lateral resolution compared to DAS and Tiny-CNN, and performance approaching MVDR at a fraction of the compute. The work claims novelty in deploying a vision-transformer–based beamformer on edge hardware and provides a practical path toward portable ultrasound imaging systems through FPGA acceleration and quantization strategies. These results have implications for portable diagnostics by enabling high-quality beamforming on resource-constrained devices without sacrificing image fidelity.
Abstract
Accelerating compute intensive non-real-time beam-forming algorithms in ultrasound imaging using deep learning architectures has been gaining momentum in the recent past. Nonetheless, the complexity of the state-of-the-art deep learning techniques poses challenges for deployment on resource-constrained edge devices. In this work, we propose a novel vision transformer based tiny beamformer (Tiny-VBF), which works on the raw radio-frequency channel data acquired through single-angle plane wave insonification. The output of our Tiny-VBF provides fast envelope detection requiring very low frame rate, i.e. 0.34 GOPs/Frame for a frame size of 368 x 128 in comparison to the state-of-the-art deep learning models. It also exhibited an 8% increase in contrast and gains of 5% and 33% in axial and lateral resolution respectively when compared to Tiny-CNN on in-vitro dataset. Additionally, our model showed a 4.2% increase in contrast and gains of 4% and 20% in axial and lateral resolution respectively when compared against conventional Delay-and-Sum (DAS) beamformer. We further propose an accelerator architecture and implement our Tiny-VBF model on a Zynq UltraScale+ MPSoC ZCU104 FPGA using a hybrid quantization scheme with 50% less resource consumption compared to the floating-point implementation, while preserving the image quality.