This clinical trial aims to evaluate the cost-efficiency and workflow impact of
AI-assisted fracture detection in an orthopedic emergency care unit. The study is
designed as a prospective, randomized, controlled trial to assess whether integrating AI
technology can improve diagnostic accuracy, streamline workflow, and reduce healthcare
costs compared to the traditional diagnostic approach.
Study Objectives
Primary Objectives:
The primary objective of the SMART Fracture Trial is to assess the impact of AI-assisted
fracture detection on physician decision-making and clinical workflows. The study will
therefore provide deeper insights into AI's potential benefits and limitations beyond
theoretical performance metrics.
Secondary Objectives:
While the primary focus of the SMART Fracture Trial is on AI's clinical integration, the
study will also comprehensively assess diagnostic accuracy and fracture classification
performance - key factors that influence real-world implementation. By analyzing these
secondary objectives, the study will provide deeper insights into AI's theoretical
performance metrics.
Study Design
This is a prospective, randomized, controlled trial conducted as an international
multi-center study. It includes two parallel arms:
Control Group: Standard diagnostic procedures without AI assistance. Intervention Group:
AI-based diagnostic tools assist in interpreting radiological images for fracture
detection.
Both groups will follow the same diagnostic imaging protocol, including standard X-ray
imaging in two planes. The AI software, pre-validated for fracture detection, will be
integrated into the hospital's Picture Archiving and Communication System (PACS).
Intervention Details
The AI fracture detection systems (Aidoc, Gleamer) are designed to identify fracture
patterns on X-rays and highlight areas of potential concern for physician review. The
software operates in real time, providing marked-up images to physicians. The AI output
serves as a diagnostic aid, with final diagnoses made by the attending physician.
Population and Sampling
Population: Patients of all ages presenting to the emergency care unit with suspected
isolated extremity fractures.
Sample Size: Approximately 4,800 participants (2400 per group) to ensure sufficient
statistical power for primary outcomes.
Randomization: Participants will be randomly assigned to the control or intervention
group using a 1:1 allocation ratio.
Outcome Measures
Primary Outcome Measures:
Diagnostic accuracy: Sensitivity, specificity, and AUC of AI-assisted vs. traditional
diagnosis.
Time to diagnosis: Total time from patient triage to final diagnosis.
Secondary Outcome Measures:
Cost analysis: A detailed cost comparison of the diagnostic process in both groups.
Diagnostic confidence: Assessed using a Likert scale (1-10) completed by physicians after
reviewing each case.
Study Procedures
Baseline Data Collection: Demographics, clinical history, and presenting symptoms will be
recorded at enrollment. Standard radiological imaging will be conducted for all
participants.
AI Integration (Intervention Group): Radiological images will be processed by AI
software, providing annotated images to physicians. AI-assisted diagnostic workflows will
be compared to standard workflows.
Outcome Assessment: All diagnoses will be independently reviewed by a panel of experts,
including an experienced radiologist and orthopedic surgeon, to establish a reference
standard for comparison.
Ethical Considerations
The study adheres to the principles of the Declaration of Helsinki and has received
approval from the local ethics committee. Written informed consent will be obtained from
all participants before enrollment. Data will be pseudonymized to maintain
confidentiality.
Expected Impact
This study aims to provide robust evidence regarding the effectiveness of AI in improving
diagnostic workflows in emergency care settings. Findings may inform the future
integration of AI tools into clinical practice, improving patient outcomes and optimizing
resource utilization in high-volume emergency care environments.