Self-Supervised Optimization of RF Data Coherence for Improving Breast Reflection UCT Reconstruction.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Signal Processing

Background:

• Reflection ultrasound computed tomography (UCT) is crucial for breast cancer screening.
• Image quality in UCT is degraded by sound speed variations in breast tissue.
• Traditional delay-and-sum (DAS) algorithms assume uniform sound speed, impacting radio frequency (RF) data coherence.

Purpose of the Study:

• To improve RF data coherence and summation in reflection UCT.
• To develop advanced modules for enhanced ultrasound imaging.
• To overcome limitations of traditional UCT algorithms.

Main Methods:

• Introduced three modules: self-supervised blind RF data segment block (BSegB), state-space model-based strong reflection prediction (SSM-SRP), and polarity-based adaptive replacing refinement (PARR).
• Leveraged spatial correlation of receiving arrays to enhance RF data coherence.
• Employed standard metrics (PSNR, SSIM, RMSE) and coherence metrics (CF, Var) for assessment.

Main Results:

• Achieved significant improvements in image quality metrics: average PSNR of 19.64 dB, average SSIM of 0.71, and average RMSE of 0.10.
• Demonstrated enhanced signal coherence at the RF data level.
• Showcased superior performance under sparse transmission conditions.

Conclusions:

• The proposed framework effectively enhances RF data coherence and UCT image quality.
• The novel modules outperform existing adaptive beamforming and deep learning methods.
• This approach offers a promising advancement for breast cancer screening using UCT.