AI Science Tracker
EN

Hospital Clínico Universitario de Valladolid

1 discovery from 2025

Discoveries

Echo networks detect vegetations in selected endocarditis images

Source check: Source checked. Not independent scientific confirmation.

Finding: YOLOv12 and RT-DETR models were evaluated on transesophageal echocardiography from 329 patients with confirmed left-sided endocarditis and visible vegetations, plus 70 controls without suspected endocarditis. Patient-separated internal testing reported diagnostic AUC 0.91 for YOLOv12. This selected case-control result does not establish screening performance or improved prognosis; other intracardiac structures can mimic vegetations, and right-sided disease was excluded.

Technology: YOLOv12; RT-DETR object detection

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The networks learn image patterns to locate vegetation bounding boxes in echo frames; the proportion of detected frames supplies a diagnostic score compared with expert clinical-image labels.
Test and result
Expert-labeled cine-loop frames, with all images from each patient kept in one subset. Ten random train/optimization/test splits were averaged; each diagnostic comparison used sixty held-out positive cases and the same seventy controls. A frame-detection ratio was used for diagnostic scoring.
Limitation / still unknown
Performance in a consecutive suspected-endocarditis population, other intracardiac masses, right-sided disease and clinical benefit remain unestablished. Positive cases were selected for visible vegetations, controls came from one center, and multiple-loop aggregation into a patient score is incompletely specified.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13481939/

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 21 Nov 2026. Review scope: Primary relevant clinical selection/reference, annotation/model and patient-split Methods, Results and limitations; supplementary per-split metric tables not independently reviewed. Exact primary identity/date and fresh primary PubMed correction/retraction relationships checked..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open the interactive profile for the timeline, or browse all organizations and model families.