Transforming Breast Cancer Diagnosis: Towards Real-Time Ultrasound to Mammogram Conversion for Cost-Effective Diagnosis
Published in Accepted to the Ultrasonics journal, 2023
Ultrasound (US) imaging is ideal for intra-operative situations due to its real-time capabilities and portability compared to other imaging procedures like mammography. US photos have lesser spatial resolution and noise-like characteristics. This study intends to alleviate constraints by giving surgeons with mammogram-like image quality in real-time from noisy US images. Our approach to increasing US image quality recognizes artifacts as informative wave interference patterns (WIP), rather than interpreting them as ‘speckle noise’ as previously done. We use the Stride software to numerically solve the forward model and generate ultrasound images from mammograms using wave equations. We use domain adaptation to increase the realism of simulated ultrasound pictures. Next, we use generative adversarial networks (GANs) to generate mammogram-quality images from ultrasound data. The resulting photos show clearer details than the original US images.
Recommended citation: Nasser S. A., Sharma A., Saraf A., Parulekar A. M., Haria P. and Sethi A. (2023). Transforming Breast Cancer Diagnosis: Towards Real-Time Ultrasound to Mammogram Conversion for Cost-Effective Diagnosis. https://arxiv.org/abs/2308.05449
Download Paper