Search Results (3)

Background: Ovarian cancer is the deadliest female reproductive malignancy. Accurate preoperative differentiation of benign and malignant ovarian masses is critical for appropriate treatment. O-RADS category 4 lesions present a wide range of malignant risk, challenging radiologists. Ultrasonic images are the primary focus of current deep learning models, with no consideration for clinical data. Methods: We proposed OTC-NET, a model that uses multimodal data for classification, which combines ultrasound images and clinical information to improve the classification ability of O-RADS 4 ovarian masses. Results: OTC-NET outperforms seven deep learning models and three radiologists of varying experience, with AUC significantly higher than junior (p < 0.001), intermediate (p < 0.01), and senior (p < 0.05) radiologists. Additionally, OTC-NET–assisted diagnosis notably improves AUC and accuracy of junior and senior radiologists (p < 0.05). Conclusions: These results indicate that OTC-NET provides superior diagnostic accuracy and has strong potential for clinical application. Full article

The diagnosis of post-stressed anchor rods is essential for maintaining the service and ensuring the safety of Electricité de France (EDF) structures. These rods are critical for the mechanical strength of structures and electromechanical components. Currently, the standard method for estimating the effective tension of post-stressed tie rods with a free length involves measuring the residual force using a hydraulic jack. However, this method can be costly, impact the structure’s operation, and pose risks to employees. Until now, there has been no reliable on-field approach to estimating residual tension using a lightweight setup. This research introduces a nondestructive method using multimodal ultrasonic guided waves to evaluate the residual tension of anchor rods with a few centimeters free at one end. The methodology was developed through both laboratory experiments and simulations. This new method allows for the extraction of dispersion curves for the first three modes, bending, torsional, and longitudinal, using time–frequency analysis and enables the estimation of the steel bar’s properties. Future work will focus on applying this methodology in the field. Full article

Objective: To develop a prediction model for discriminating malignant from benign breast non-mass-like lesions (NMLs) using conventional ultrasound (US), strain elastography (SE) of US elastography and contrast-enhanced ultrasound (CEUS). Methods: A total of 101 NMLs from 100 patients detected by conventional US were enrolled in this retrospective study. The characteristics of NMLs in conventional US, SE and CEUS were compared between malignant and benign NMLs. Histopathological results were used as the reference standard. Binary logistic regression analysis was performed to identify the independent risk factors. A multimodal method to evaluate NMLs based on logistic regression was developed. The diagnostic performance of conventional US, US + SE, US + CEUS and the combination of these modalities was evaluated and compared. Results: Among the 101 lesions, 50 (49.5%) were benign and 51 (50.5%) were malignant. Age ≥45 y, microcalcifications in the lesion, elasticity score >3, earlier enhancement time and hyper-enhancement were independent diagnostic indicators included to establish the multimodal prediction method. The area under the receiver operating characteristic curve (AUC) of US + SE + CEUS was significantly higher than that of US (p < 0.0001) and US + SE (p < 0.0001), but there was no significant difference between the AUC of US + SE + CEUS and the AUC of US + CEUS (p = 0.216). Conclusion: US + SE + CEUS and US + CEUS could significantly improve the diagnostic efficiency and accuracy of conventional US in the diagnosis of NMLs. Full article