Physics-Guided Multi-Objective Deep Learning for Ultrasound RF Data Interpolation in Resource-Constrained Imaging

2026-09-23

AI Summary

This study addresses the significant image degradation and artifacts caused by sparse ultrasound acquisition in resource-constrained scenarios. To overcome this, we propose a physics-guided, end-to-end interpolation network for reconstructing dense radio-frequency (RF) data from sparse acquisitions. The method introduces a multi-objective loss function combined with an exponential moving average (EMA) strategy to stabilize hybrid supervised training, alongside a random skip masking scheme designed to enhance model generalization. By integrating deep learning, multi-objective optimization, and coherent beamforming techniques, the proposed framework achieves robust performance across various decimation factors. Experimental results demonstrate that the method maintains an average structural similarity index measure (SSIM) of approximately 0.95 while substantially improving imaging contrast and overall robustness.

Abstract

Ultrasound imaging increasingly targets portable, point-of-care, and wearable settings where constraints on power, bandwidth, and hardware complexity often necessitate sparse data acquisition in spatiotemporal scanning. However, image reconstruction using the sparse data can introduce insufficient phase information in coherent beamforming process, resulting in grating-lobe artifacts that degrade imaging contrast resolution. We present a physics-guided, data-driven framework for sparse-to-dense radio-frequency (RF) reconstruction that aligns training with downstream image formation. Our approach trains an end-to-end interpolation network using a hybrid supervision scheme that combines an RF-domain and a beamforming-domain loss with exponential moving average (EMA) to stabilize the multi-objective training. To improve generalization under variable acquisition layouts, we also introduce a random-skip masking strategy that varies sparsity patterns during training so a single model can handle diverse decimation factors and irregular channel configurations. We evaluate the framework on a held-out test set using the mean structural similarity index measure (SSIM) between reconstructed and ground-truth beamformed images. Across decimation factors $\times 2$ to $\times 13$, the best-performing configuration maintains mean SSIM around 0.95. Overall, the results show consistent gains in RF reconstruction and post-beamforming image quality across diverse acquisition conditions. This approach enables robust, high-quality ultrasound imaging at resource-constrained settings by allowing more sparse scanning in spatiotemporal domain.

Problem

• Ultrasound imaging
• Sparse data acquisition
• RF data interpolation
• Grating-lobe artifacts
• Resource-constrained imaging

Innovation

• Physics-guided deep learning
• Multi-objective training
• Random-skip masking
• Ultrasound RF interpolation
• Exponential moving average