Researchers and clinicians reported multiple advances in AI assistance for imaging-based care, including a prospective observer-performance trial in salivary gland tumors and new work on preoperative mapping in glioma. Together, the studies highlight the shift from retrospective accuracy claims toward interfaces, observer agreement, and clinical workflow alignment. In the salivary space, a confidence-aware AI system with a radiologist-editable interface improved accuracy and agreement in distinguishing common benign parotid tumors on ultrasound in a prospective two-phase observer-performance study. The design emphasizes controllability and confidence signaling as keys for adoption. Elsewhere, new imaging AI approaches are being tested for tumor-margin detection and recurrence planning, signaling that deployment conversations increasingly hinge on prospective validation and limits of generalization—not just model performance on static datasets.