3. Producción
Browse
4 results
Search Results
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Structural investigation of regime-dependent surrogate representations under controlled hydrodynamic conditions: A Mixture-of-Experts approach(RELX Group (Netherlands), 2026-01-01) - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Spatially Weighted Fidelity and Regularization Terms for Attenuation Imaging(Institute of Electrical and Electronics Engineers Inc., 2025)Quantitative ultrasound (QUS) holds promise in enhancing diagnostic accuracy. For attenuation imaging, the regularized spectral log difference (RSLD) can generate accurate local attenuation maps. However, the performance of the method degrades when significant changes in backscatter amplitude occur. Variations in the technique were introduced involving a weighted approach to backscatter regularization, which, however, is not effective when changes in both attenuation and backscatter are present. This study introduces a novel approach that incorporates an L1-norm for backscatter regularization and spatially varying weights for both fidelity and regularization terms. The weights are calculated from an initial estimation of backscatter changes. Comparative analyses with simulated, phantom, and clinical data were performed. When changes in backscatter and attenuation occur, the proposed approach reduced the lowest root mean square error by up to 73%. It also improved the contrast-to-noise ratio (CNR) by a factor of 4.4 on average compared with previously available methods, considering the simulated and phantom data. In vivo results from healthy livers, thyroid nodules, and a breast tumor further confirm its effectiveness. In the liver, it is shown to be effective at reducing artifacts of attenuation images. In thyroid and breast tumors, the method demonstrated an enhanced CNR and better consistency of the attenuation measurements with the posterior acoustic enhancement. Overall, this approach offers promise for enhancing ultrasound attenuation imaging by helping differentiate tissue characteristics that may indicate pathology.1 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Volumetric Attenuation Estimation Using a Matrix Array with Spatially Weighted Fidelity and Regularization(IEEE Computer Society, 2025)Quantitative ultrasound (QUS) enables system-independent tissue characterization by deriving acoustic biomarkers. Among these, attenuation imaging has emerged as a promising tool for clinical applications. While regularization techniques have been proposed for parameter estimation, balancing spatial resolution and accuracy remains a critical challenge. Volumetric QUS imaging offers enhanced resolution by leveraging 3D spatial information, yet simultaneous variations in backscatter and attenuation properties often degrade accuracy. Recently, a spatially adaptive approach that weighted the regularization and fidelity terms (SWIFT) has demonstrated success in mitigating this effect. This study investigates the benefits of combining SWIFT and volumetric QUS imaging using a matrix array transducer. The method was compared to the 2D versions and a non-weighted approach using phantom data. The root mean square error is reduced from 40% to 10% compared to the 2D algorithms, and the contrast-to-noise ratio increases by at least 30%. The method with spatially weighted regularization and fidelity improves the precision-resolution trade-off in QUS reconstruction despite variations in backscatter and attenuation.2 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Integrating Deep Learning into PnP-ADMM for Ultrasound Attenuation Coefficient Estimation(IEEE Computer Society, 2025)Estimating the attenuation coefficient slope (ACS) is essential for tissue characterization in quantitative ultrasound (QUS). Traditional model-based methods such as the regularized spectral log difference (RSLD) rely on manually tuned priors, while recent end-to-end deep learning approaches struggle to generalize to in vivo data. This work proposes a hybrid method that integrates pre-trained Attention U-Nets within a Plug-and-Play ADMM framework. The fidelity term is replaced by a network conditioned on spectral inputs and the iteration index, while a second network acts as a learned regularizer. The method was evaluated on physical phantoms and in vivo breast and thyroid acquisitions, after being trained entirely on simulations. Results show improved accuracy and generalization over RSLD and end-to-end baselines, suggesting that embedding a pre-trained deep learning model within a physics-based framework enhances robustness and enables more reliable ACS estimates in real-world scenarios.2