FractureDetect (FX)

FractureDetect (FX) is an artificial intelligence-powered software tool designed to assist clinicians in the detection of fractures in adult musculoskeletal radiographs across various anatomical regions and views. By employing supervised deep learning algorithms to analyze DICOM images and provide toggleable annotated overlays, it aims to improve diagnostic accuracy, sensitivity, and specificity in clinical practice.

FractureDetect (FX) is a computer-assisted detection and diagnosis (CAD) software designed to enhance clinicians' ability to identify fractures in adult musculoskeletal radiographs. It utilizes industry-standard deep learning algorithms for computer vision, specifically supervised deep learning, to analyze DICOM images obtained from sources like PACS systems or imaging devices. The software processes these radiographs and generates a DICOM overlay as its output. This overlay provides clear textual and graphical annotations: "Fracture: DETECTED" along with one or more bounding boxes highlighting potential fracture sites, or "Fracture: NOT DETECTED" if no fractures are identified. Clinicians can easily toggle this informative overlay on or off within their standard PACS viewer, facilitating an unimpeded concurrent review of the original radiographic image. FX supports a comprehensive range of twelve anatomical study types, including ankle, clavicle, elbow, femur, forearm, hip, humerus, knee, pelvis, shoulder, tibia/fibula, and wrist, across various industry-standard radiographic views. The efficacy of FractureDetect (FX) was substantiated through extensive standalone performance testing and a robust clinical reader study. The standalone assessment, conducted on 11,970 radiographs covering all indicated study types and views, demonstrated high performance with an overall sensitivity of 0.951, specificity of 0.893, and an Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) of 0.982. This testing also confirmed FX's consistent accuracy across different anatomical regions and despite potential confounders like image brightness or varying x-ray equipment. Furthermore, a multiple reader, multiple case (MRMC) retrospective clinical study, involving 24 clinical readers and 175 cases, showed that FX significantly improved clinicians' diagnostic accuracy. When aided by FX, reader AUC increased from 0.912 to 0.952, reader sensitivity improved from 0.819 to 0.900, and reader specificity rose from 0.890 to 0.918, thereby confirming its value in enhancing fracture detection.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does FractureDetect (FX) differentiate between acute fractures and old, healed fractures or other bone abnormalities that might mimic fractures on radiographs?; What is the performance variability of FX across different patient demographics (e.g., age, bone density) and varying image quality levels encountered in routine clinical practice?; How does the integration of the FX DICOM overlay into existing PACS workflows impact radiologist reading time and overall efficiency?; What specific deep learning architectural components and training data characteristics contribute to FX's reported sensitivity and specificity, particularly for subtle or complex fracture types?; Are there plans to expand FractureDetect (FX) capabilities to pediatric populations or to detect other musculoskeletal pathologies beyond fractures?