Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

APS-USCT: Ultrasound Computed Tomography on Sparse Data via AI-Physic Synergy

Yi Sheng, Hanchen Wang, Yipei Liu, Junhuan Yang, Weiwen Jiang, Youzuo Lin, Lei Yang

TL;DR

This work tackles the high hardware and data costs of Ultrasound Computed Tomography (USCT) by introducing APS-USCT, an AI-physics driven framework that upscales sparse waveform data into dense inputs for accurate speed-of-sound (SOS) maps. It blends an AI-driven waveform upsampling module (APS-wave) with a physics-informed SOS reconstruction module (APS-FWI) that employs source encoding and SE-attention within InversionNet, mediated by an APS-physics forward model. On breast-like 2D numerical phantoms, APS-USCT achieves an average SSIM of $0.8431$ and PSNR of $25.3040$, with 82.93% of samples surpassing $SSIM>0.8$ and 14.63% exceeding $SSIM>0.9$, while offering about $2.5\times$ hardware-cost reductions and only $0.0007$ SSIM degradation. These results demonstrate robust SOS reconstruction across tissue types and indicate substantial practical potential for reducing equipment costs in USCT without sacrificing image quality.

Abstract

Ultrasound computed tomography (USCT) is a promising technique that achieves superior medical imaging reconstruction resolution by fully leveraging waveform information, outperforming conventional ultrasound methods. Despite its advantages, high-quality USCT reconstruction relies on extensive data acquisition by a large number of transducers, leading to increased costs, computational demands, extended patient scanning times, and manufacturing complexities. To mitigate these issues, we propose a new USCT method called APS-USCT, which facilitates imaging with sparse data, substantially reducing dependence on high-cost dense data acquisition. Our APS-USCT method consists of two primary components: APS-wave and APS-FWI. The APS-wave component, an encoder-decoder system, preprocesses the waveform data, converting sparse data into dense waveforms to augment sample density prior to reconstruction. The APS-FWI component, utilizing the InversionNet, directly reconstructs the speed of sound (SOS) from the ultrasound waveform data. We further improve the model's performance by incorporating Squeeze-and-Excitation (SE) Blocks and source encoding techniques. Testing our method on a breast cancer dataset yielded promising results. It demonstrated outstanding performance with an average Structural Similarity Index (SSIM) of 0.8431. Notably, over 82% of samples achieved an SSIM above 0.8, with nearly 61% exceeding 0.85, highlighting the significant potential of our approach in improving USCT image reconstruction by efficiently utilizing sparse data.

APS-USCT: Ultrasound Computed Tomography on Sparse Data via AI-Physic Synergy

TL;DR

This work tackles the high hardware and data costs of Ultrasound Computed Tomography (USCT) by introducing APS-USCT, an AI-physics driven framework that upscales sparse waveform data into dense inputs for accurate speed-of-sound (SOS) maps. It blends an AI-driven waveform upsampling module (APS-wave) with a physics-informed SOS reconstruction module (APS-FWI) that employs source encoding and SE-attention within InversionNet, mediated by an APS-physics forward model. On breast-like 2D numerical phantoms, APS-USCT achieves an average SSIM of and PSNR of , with 82.93% of samples surpassing and 14.63% exceeding , while offering about hardware-cost reductions and only SSIM degradation. These results demonstrate robust SOS reconstruction across tissue types and indicate substantial practical potential for reducing equipment costs in USCT without sacrificing image quality.

Abstract

Ultrasound computed tomography (USCT) is a promising technique that achieves superior medical imaging reconstruction resolution by fully leveraging waveform information, outperforming conventional ultrasound methods. Despite its advantages, high-quality USCT reconstruction relies on extensive data acquisition by a large number of transducers, leading to increased costs, computational demands, extended patient scanning times, and manufacturing complexities. To mitigate these issues, we propose a new USCT method called APS-USCT, which facilitates imaging with sparse data, substantially reducing dependence on high-cost dense data acquisition. Our APS-USCT method consists of two primary components: APS-wave and APS-FWI. The APS-wave component, an encoder-decoder system, preprocesses the waveform data, converting sparse data into dense waveforms to augment sample density prior to reconstruction. The APS-FWI component, utilizing the InversionNet, directly reconstructs the speed of sound (SOS) from the ultrasound waveform data. We further improve the model's performance by incorporating Squeeze-and-Excitation (SE) Blocks and source encoding techniques. Testing our method on a breast cancer dataset yielded promising results. It demonstrated outstanding performance with an average Structural Similarity Index (SSIM) of 0.8431. Notably, over 82% of samples achieved an SSIM above 0.8, with nearly 61% exceeding 0.85, highlighting the significant potential of our approach in improving USCT image reconstruction by efficiently utilizing sparse data.
Paper Structure (9 sections, 2 equations, 4 figures, 2 tables)

This paper contains 9 sections, 2 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Experiment
  4. Dataset, Implementation, and Evaluation Protocol
  5. Main Results
  6. Ablation Studies
  7. Conclusion
  8. Acknowledgements.
  9. Disclosure of Interests.

Figures (4)

  • Figure 1: Relationship between sparsity and model performance
  • Figure 2: An overview of the proposed APS-USCT framework: (1) the black-color path shows the inference of reconstructing the speed of sound (SOS) maps from a given sparse measurement; (2) the green-color paths show the training procedure of two AI models associated with APS-wave and APS-FWI components; and (3) the blue-color path shows the waveform augmentation by using training APS-physics to generate the training dataset.
  • Figure 3: Visualization of the SOS map for the methods from the Table \ref{['tab:main']}
  • Figure 4: Results of the ablation studies for APS-wave and APS-FWI