Search papers, labs, and topics across Lattice.
VDv2 (Golhar et al., 2025), a high-definition phantom-based colonoscopy video dataset designed for
2
0
4
Uncertainty-aware reconstruction using a Gaussian process prior leads to a 21.9% improvement in endoscopic video restoration, enhancing clinical interpretability.
Current CADe systems excel at identifying early neoplasia in Barrett's esophagus on balanced datasets, but their clinical utility plummets in real-world, low-prevalence settings, exposing a critical gap in prevalence-aware design.
