| Sumario: | While lossy compression is gaining acceptance in medical imaging research, its practical use remains limited due to the lack of specific implementation guidelines. This study investigates the feasibility and diagnostic reliability of various lossy compression methods for 3D whole-breast ultrasound (WBUS) images, aiming to support remote breast cancer screening in resource-constrained settings. We retrospectively collected WBUS images from 400 patients with breast masses. Images were compressed using JPEG 2000 (JP2K), JPEG 3D (JP3D), and High-Efficiency Video Coding (HEVC) at compression ratios ranging from 5:1 to 50:1. Image quality was assessed using PSNR and SSIM. Four radiologists independently reviewed compressed images to identify two thresholds: (1) visually lossless quality and (2) the highest diagnostically acceptable compression ratio (DAS). Diagnostic accuracy was evaluated using the area under the ROC curve (AUC). HEVC outperformed JP2K and JP3D in both objective quality metrics and diagnostic accuracy. At 5:1 compression, radiologists could not distinguish compressed images from originals, and diagnostic performance was unaffected (AUC > 0.85 for all readers). Even at an average 20:1 compression ratio, no significant degradation in diagnostic accuracy was observed. Inter-reader agreement remained high across compression levels. Lossy HEVC compression provides a practical approach for remote breast cancer screening. In this study, WBUS images compressed at 5:1 were visually indistinguishable from the originals, and compression up to 20:1 maintained diagnostic performance within our dataset, supporting its potential use in tele-ultrasound workflows. Further studies are warranted to confirm these findings across different devices and clinical settings.
|
|---|