Related Experiment Video
Updated: Nov 28, 2025
09:34
A Method to Estimate Cadaveric Femur Cortical Strains During Fracture Testing Using Digital Image Correlation
Published on: September 14, 2017
7.6K
A Human-Algorithm Integration System for Hip Fracture Detection on Plain Radiography: System Development and
Chi-Tung Cheng1, Chih-Chi Chen2, Fu-Jen Cheng3
1Department of Trauma and Emergency Surgery, Linkou Chang Gung Memorial Hospital, Chang Gung University, Taoyuan, Taiwan.
JMIR Medical Informatics
|November 27, 2020
Summary
A new human-algorithm integration (HAI) system significantly improves hip fracture diagnosis accuracy in elderly patients. This deep learning tool enhances physician performance, proving feasible for emergency departments.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence in Medicine
Background:
- Hip fractures are a common injury in the elderly population.
- Deep learning (DL) algorithms show promise for improving hip fracture diagnosis from plain pelvic radiographs (PXRs).
- The clinical efficacy of DL for hip fracture detection remains to be fully determined.
Purpose of the Study:
- To develop and validate a human-algorithm integration (HAI) system for enhanced hip fracture diagnosis.
- To assess the system's performance in a real-world clinical setting.
- To improve diagnostic accuracy in emergency departments.
Main Methods:
- A deep learning algorithm was trained on trauma registry data and 3605 PXRs.
- A 34-physician cohort evaluated diagnostic performance with and without HAI system assistance.
- Subgroup analyses examined performance by physician specialty and experience.
- The HAI system was deployed in emergency departments for real-world validation.
Main Results:
- Physician diagnostic performance significantly improved with HAI assistance (sensitivity: 95% to 99%, specificity: 90% to 95%, accuracy: 90% to 96%).
- Human-algorithm agreement increased substantially (κ: 0.69 to 0.80).
- Both experienced and less-experienced physicians benefited, achieving performance comparable to consulting physicians.
- In clinical deployment, the HAI system demonstrated high sensitivity (97%), specificity (95.7%), and accuracy (96.08%).
Conclusions:
- Human-algorithm integration (HAI) is impacting healthcare delivery.
- Integrating AI-powered diagnostic tools into emergency departments is feasible.
- The developed HAI system effectively enhances physicians' accuracy in diagnosing hip fractures.
