ReC-TTT: Contrastive Feature Reconstruction for Test-Time Training
ReC-TTT is a test-time training method that improves model adaptation to unseen domains using cross-reconstruction between encoders and a shared decoder. It outperforms existing methods in domain shift classification tasks.
Mar 3, 2025
LoRIS - Weakly-Supervised Anomaly Detection for Ultrasound Images
LoRIS is a weakly-supervised anomaly detection method for knee ultrasound images. Trained only on nondistended cases, it uses a location prior to guide reconstruction and detect anomalies. It outperforms unsupervised methods and matches supervised ones—without needing segmentation annotations or images from both classes.
Oct 5, 2024
Insights on the Development of PRACTICE, a Research-oriented Healthcare Platform
Jul 1, 2024
GAJA - Guided self-Acquisition of Joint ultrAsound images
Feb 1, 2023