Provides an automated system for detecting fractures on x-rays
An automated system for detecting fractures on X-rays
X-ray
Binarcy descision of fracture (present or not)Identify area where fracture is found
The RBfracture tool was evaluated on a dataset of 319 patient exams covering a range of anatomical regions. The median age was 52 years, with cases spanning from 2 to 94 years old. Hip and pelvis were the most frequently examined areas.
The study demonstrated the following performance metrics for RBfracture:
Sensitivity: 94%
Specificity: 94%
Accuracy: 94%
Positive Predictive Value (PPV): 83%
Negative Predictive Value (NPV): 98%
A confusion matrix was used to calculate the above metrics, and the inter-reader agreement was measured at 96%, underscoring the reliability of RBfracture against human readings.
RBfracture demonstrated a median processing time of 13 seconds per exam, with 90% of cases processed within 23 seconds. This rapid turnaround allows for quicker diagnostic workflows, particularly valuable in emergency settings where timely fracture identification is critical.
RBfracture integrates seamlessly within PACS, offering real-time decision support and marking suspected fractures, thus minimizing additional workload for radiologists. This integration enhances efficiency and accuracy, especially in out-of-hours settings where radiologists may not be immediately available.
Prior to RBfracture’s deployment, missed fracture rates were between 1.9 and 4.5 per 1000 patients. Post-deployment, the missed fracture rate dropped to 1.0 per 1000 patients, representing a 47-62% reduction. This reduction likely decreases costs associated with misdiagnosis, such as additional imaging or delayed treatment, improving both patient outcomes and healthcare resource allocation. By reducing missed fractures, RBfracture minimizes the need for repeat visits and radiological assessments. It serves as an effective tool in A&E departments, particularly for hospitals without round-the-clock radiology coverage, thereby lowering the demand for after-hours reporting by radiologists.
