Related Experiment Video
Updated: Aug 15, 2026
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
Vendor-Agnostic Multisite Automated Dual-Energy X-Ray Absorptiometry Reporting Using Artificial Intelligence-Based
Paras Lakhani1, Steven A Rothenberg2, Christopher G Roth3
1Associate Professor of Radiology and Clinical Director of Imaging Informatics, Department of Radiology, Jefferson Health-Thomas Jefferson University Hospital, Philadelphia, Pennsylvania.
An AI-powered optical character recognition (OCR) system significantly reduced dual-energy x-ray absorptiometry (DXA) report creation and turnaround times in both academic and community settings. The AI system maintained high accuracy and improved report completeness, streamlining DXA reporting workflows.
Area of Science:
- Radiology and Imaging Informatics
- Artificial Intelligence in Healthcare
- Medical Reporting Systems
Background:
- Dual-energy x-ray absorptiometry (DXA) reporting is crucial for osteoporosis diagnosis and management.
- Current DXA reporting workflows can be time-consuming, impacting operational efficiency.
- There is a need for automated solutions to improve the speed and accuracy of DXA report generation.
Purpose of the Study:
- To evaluate the operational impact and accuracy of a vendor-agnostic AI-based optical character recognition (OCR) system for drafting DXA reports.
- To assess the system's performance in both academic and community practice settings.
- To determine if the AI system can reduce report creation time (RCT) and report turnaround time (TAT) while maintaining accuracy.
Main Methods:
- Implementation of an AI OCR DXA reporting pipeline across four outpatient imaging sites (two academic, two community).
- AI OCR extracted measurements from DXA DICOM images; rule-based logic generated draft reports.
- Pre/post design to measure RCT and TAT; accuracy assessed by comparing AI drafts to source images (n=400).
Main Results:
- Median RCT decreased significantly at academic sites (3.48 to 0.87 min) and community sites (1.42 to 0.63 min).
- Median TAT decreased significantly at academic sites (2.40 to 0.96 hours) and community sites (133.82 to 42.10 hours).
- AI drafts showed comparable or higher numerical accuracy (≥99.9%), improved completeness at community sites (100% vs 45%), and preserved diagnostic accuracy (≥99.5%).
Conclusions:
- A vendor-agnostic AI OCR system substantially reduces RCT and TAT for DXA reports.
- The AI system maintains high numerical and diagnostic accuracy and improves report completeness.
- AI-driven DXA reporting offers significant operational benefits in diverse clinical settings.
