Subject: R Medicine (General)
Year: 2026
Type: Article
Type: NonPeerReviewed
Title: FROM DETECTION TO CLINICAL DECISION: ARTIFICIAL INTELLIGENCE BASED RISK STRATIFICATION IN PEDIATRIC FRACTURE IMAGING
Author: MEMO, Polikseni
Abstract: Artificial intelligence (AI) has demonstrated high accuracy in detecting pediatric fractures on X-rays, however, its impact on clinical decision making remains limited. This study proposes a “detection to decision” framework that transforms AI outputs into a clinically actionable risk stratification tool. Using the GRAZPEDWRI-DX dataset, a convolutional neural network (CNN) was trained to predict fracture probability. These predictions were calibrated and combined with clinical variables, including patient age, imaging view, and fracture type, to generate a composite RiskScore using logistic regression. In a simulated emergency department workflow, the proposed risk based prioritization approach reduced time to review for high risk cases by approximately 25–30% compared to a standard first in first out (FIFO) strategy. The model achieved an AUC of 0.86, with improved sensitivity and negative predictive value while maintaining good calibration. These results suggest that integrating risk based prioritization into radiology workflows may enhance diagnostic efficiency and reduce missed fractures. Overall, this study highlights a shift in the role of AI in radiology from detection toward decision support, with potential benefits for workflow optimization and patient safety in pediatric imaging.
Publisher: University of Tetova, Tetova, North Macedonia
Relation: https://eprints.unite.edu.mk/2355/
Identifier: oai:eprints.unite.edu.mk:2355
Identifier: https://eprints.unite.edu.mk/2355/1/FullBook%281%29-36-42.pdfIdentifier: MEMO, Polikseni (2026) FROM DETECTION TO CLINICAL DECISION: ARTIFICIAL INTELLIGENCE BASED RISK STRATIFICATION IN PEDIATRIC FRACTURE IMAGING. 2026 International Congress "From Research to Application", 20 May 2026. pp. 36-42.
