Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

UNICORN: Ultrasound Nakagami Imaging via Score Matching and Adaptation

Kwanyoung Kim, Jaa-Yeon Lee, Jong Chul Ye

TL;DR

This paper tackles the instability and resolution limitations of sliding-window Nakagami imaging by introducing UNICORN, a score-matching framework for ultrasound envelope data. It learns the envelope score with amortized denoising score matching and derives a closed-form, per-pixel Nakagami parameter estimator $\hat{m}=\frac{\frac{1}{r} + \nabla_r \log p_{R}(r)}{\left(\frac{2}{r} - \frac{2r}{\hat{\Omega}}\right)}$, where $\hat{\Omega}=E[R^2]$, followed by a low-pass adaptation. Through synthetic MNIST-like and BUSI ultrasound datasets, UNICORN achieves higher PSNR and lower RMSE than conventional estimators, and qualitative analysis on real RF envelopes from the OASBUD breast dataset shows improved tumor contouring and clearer pre-Rayleigh Nakagami distributions for benign vs malignant tissue. Overall, UNICORN provides a windowless, high-resolution Nakagami imaging approach with direct per-pixel estimation, offering a promising tool for tumor assessment and fat fraction estimation in ultrasound imaging.

Abstract

Nakagami imaging holds promise for visualizing and quantifying tissue scattering in ultrasound waves, with potential applications in tumor diagnosis and fat fraction estimation which are challenging to discern by conventional ultrasound B-mode images. Existing methods struggle with optimal window size selection and suffer from estimator instability, leading to degraded resolution images. To address this, here we propose a novel method called UNICORN (Ultrasound Nakagami Imaging via Score Matching and Adaptation), that offers an accurate, closed-form estimator for Nakagami parameter estimation in terms of the score function of ultrasonic envelope. Extensive experiments using simulation and real ultrasound RF data demonstrate UNICORN's superiority over conventional approaches in accuracy and resolution quality.

UNICORN: Ultrasound Nakagami Imaging via Score Matching and Adaptation

TL;DR

This paper tackles the instability and resolution limitations of sliding-window Nakagami imaging by introducing UNICORN, a score-matching framework for ultrasound envelope data. It learns the envelope score with amortized denoising score matching and derives a closed-form, per-pixel Nakagami parameter estimator , where , followed by a low-pass adaptation. Through synthetic MNIST-like and BUSI ultrasound datasets, UNICORN achieves higher PSNR and lower RMSE than conventional estimators, and qualitative analysis on real RF envelopes from the OASBUD breast dataset shows improved tumor contouring and clearer pre-Rayleigh Nakagami distributions for benign vs malignant tissue. Overall, UNICORN provides a windowless, high-resolution Nakagami imaging approach with direct per-pixel estimation, offering a promising tool for tumor assessment and fat fraction estimation in ultrasound imaging.

Abstract

Nakagami imaging holds promise for visualizing and quantifying tissue scattering in ultrasound waves, with potential applications in tumor diagnosis and fat fraction estimation which are challenging to discern by conventional ultrasound B-mode images. Existing methods struggle with optimal window size selection and suffer from estimator instability, leading to degraded resolution images. To address this, here we propose a novel method called UNICORN (Ultrasound Nakagami Imaging via Score Matching and Adaptation), that offers an accurate, closed-form estimator for Nakagami parameter estimation in terms of the score function of ultrasonic envelope. Extensive experiments using simulation and real ultrasound RF data demonstrate UNICORN's superiority over conventional approaches in accuracy and resolution quality.
Paper Structure (15 sections, 1 theorem, 10 equations, 5 figures, 2 tables)

This paper contains 15 sections, 1 theorem, 10 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Methods
  4. Nakagami parameter Estimator with Score function
  5. Loss function for Denoising Score Matching
  6. Experiments
  7. Implementation Details
  8. Datasets
  9. Evaluation Metrics
  10. Results on Synthetic Simulation dataset
  11. Results on Real RF Envelope dataset
  12. Discussion and Conclusion
  13. Proof of \ref{['prop1']}
  14. Additional Visual Result
  15. The Further detail of Implementation

Key Result

Proposition 1

For the the given measurement model pd:naka, the estimate of the unknown Nakagami parameter $m$ is given by where $\nabla_r \log p_{R}(r)$ is the score function of the RF envelope data $R$.

Figures (5)

  • Figure 1: Nakagami imaging using conventional methods and our UNICORN framework. (a) Momentum-based approach uses a sliding window for Nakagami parameter calculation. (b) Maximum likelihood method obtains Nakagami image through ML estimation. (c) UNICORN consists of two stages: training a score model to learn RF envelope score function, and inference step estimates Nakagami image in terms of score function.
  • Figure 1: Comparison of qualitative results across various methods: (Row 1) benign tumor, (Row 2) malignant tumor example. Comparison against Momentum, MLE, WMC, UNICORN without low pass filters and UNICORN with a low pass filter. The red line indicates the corresponding ROI of tumor.
  • Figure 2: Comparison of qualitative results across various methods: (Row 1) MNIST dataset, (Rows 2-3) BUSI Ultrasound Image dataset. Comparison against Momentum, MLE, WMC, and UNICORN. The yellow numbers indicate PSNR. To visualize the results, we normalize the images and convert them to grayscale.
  • Figure 3: Comparison of qualitative results across various methods: (Row 1) benign tumor, (Row 2) malignant tumor example. Comparison against Momentum, MLE, WMC, and UNICORN. The red line indicates the corresponding ROI of tumor.
  • Figure 4: Comparison of Nakagami parameters within breast tumor ROIs. (a) box plots illustrating the Nakagami parameter distribution for benign and malignant tumors. (b) Histogram depicting the distribution of Nakagami parameters within benign tumors (refer to Figure \ref{['fig:oasbud']}).

Theorems & Definitions (3)

  • Proposition 1
  • proof
  • proof