aceA multi-task learning-based deep learning model for precise estimation of rib fracture age on chest CT
Abstract: Rib fractures are a common type of chest injury, and the estimation of the fracture age mainly relies on clinical or forensic imaging experts making rough judgments based on CT scans, which is highly subjective. In recent years, artificial intelligence (AI) models have performed exceptionally well in rib fracture detection tasks, providing a potential technical foundation for inferring the time of fracture formation. Therefore, it is necessary to develop deep learning model tools to assist human judgment. In thi…
"Apical pneumothorax shown on CT and chest X-ray with left first rib fracture" by Cerevisae is licensed under CC BY-SA 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/4.0/.
By September 2026, researchers reported a multi-task deep learning model based on 3D-ResNet18 designed to assist expert judgment of rib fracture age from chest CT. Trained on 1,848 fractures from multiple centers, the model was tested externally and reported to predict fracture age, healing stage, and fracture type concurrently.
The reported improvement over subjective manual assessment matters for clinical care and forensic timing, where accurate dating informs treatment and legal determinations. What remains uncertain from this text is prospective clinical deployment, performance across diverse patient populations and scanner types, and integration into routine radiology workflows.
- Study developed multi-task deep learning model based on 3D-ResNet18 to predict fracture age, classify healing stages, and identify fracture types concurrently.
- Training used multicenter dataset of 1,848 rib fractures from chest CT scans with external testing for generalizability.
- Model achieved 94.71% accuracy in fracture type classification task in addition to age and healing stage tasks.
A 3D-ResNet18 multi-task model reduced error in estimating rib fracture age from chest CT to 7.94 days MAE, outperforming manual expert evaluation at 10.65 days, while also classifying healing stage and fracture type.
The rundown
Researchers built a multi-task model using 3D-ResNet18 architecture to perform three tasks at once from chest CT: estimating days since fracture, classifying healing stage, and identifying fracture type.
Evaluation on 1,848 fractures showed MAE of 7.94 days for age prediction versus 10.65 days for manual evaluation, 71.18% accuracy for healing stage classification, and 94.71% accuracy for fracture type classification, with external testing reported.
Sources
- Peer-reviewedInternational Journal of Legal Medicine2026-09-14
How should this claim be treated?
The debate