Enhancing Ultrasound Image Quality Across Disease Domains: Application of Cycle-Consistent Generative Adversarial Network and Perceptual Loss
2024

Improving Ultrasound Image Quality with AI

Sample size: 2100 publication Evidence: high

Author Information

Author(s): Rizvi Syed, Puzhakkal Niyas, Peng Junbo, Shin Byeong-Seok, Athreya Shreeram, Radhachandran Ashwath, Ivezić Vedrana, Sant Vivek R, Arnold Corey W, Speier William

Primary Institution: University of California Los Angeles

Hypothesis

Can a CycleGAN framework enhance the quality of ultrasound images captured by portable devices?

Conclusion

The study demonstrates that the CycleGAN model significantly improves ultrasound image quality while preserving anatomical details.

Supporting Evidence

  • The CycleGAN framework outperformed previous state-of-the-art models in multiple evaluation metrics.
  • The model achieved a structural similarity index of 0.2889, significantly higher than the baseline.
  • Results showed a peak signal-to-noise ratio of 15.8935, indicating improved image quality.
  • The learned perceptual image patch similarity score was 0.4490, demonstrating better perceptual quality.

Takeaway

This study shows that we can use smart computer programs to make blurry ultrasound pictures clearer, helping doctors see better.

Methodology

The study used a CycleGAN framework enhanced with perceptual loss to process nonregistered ultrasound image pairs from various organ systems.

Limitations

The reliance on perceptual loss increases computational complexity and longer training times.

Participant Demographics

Images were collected from 131 patients with suspected thyroid tumors, carotid plaque, or breast cancer, along with healthy participants.

Statistical Information

P-Value

p<0.001

Statistical Significance

p<0.001

Digital Object Identifier (DOI)

10.2196/58911

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication