From Narrow to Wide: Autoencoding Transformers for Ultrasound Bandwidth Recovery
This work addresses the issue of poor image resolution in resource-constrained ultrasound settings by enabling a software upgrade for existing narrow-band probes, potentially improving access to high-resolution imaging.
The paper tackled the problem of low-cost ultrasound probes producing narrow bandwidths, which degrade image quality, by learning a mapping from band-limited to broadband spectrograms using a modified Vision Transformer auto-encoder, resulting in a 90% reduction in image-domain MSE, a 6.7 dB PSNR boost, and an SSIM of 0.965 on phantoms.
Conventional pulse-echo ultrasound suffers when low-cost probes deliver only narrow fractional bandwidths, elongating pulses and erasing high-frequency detail. We address this limitation by learning a data-driven mapping from band-limited to broadband spectrogram of radio-frequency (RF) lines. To this end, a variation of Tiny Vision Transform (ViT) auto-encoder is trained on simulation data using a curriculum-weighted loss. On heterogeneous speckle-cyst phantoms, the network reduces image-domain MSE by 90 percent, boosts PSNR by 6.7 dB, and raises SSIM to 0.965 compared with the narrow-band input. It also sharpens point-target rows in a completely unseen resolution phantom, demonstrating strong out-of-distribution generalisation without sacrificing frame rate or phase information. These results indicate that a purely software upgrade can endow installed narrow-band probes with broadband-like performance, potentially widening access to high-resolution ultrasound in resource-constrained settings.