In addition to offering a 98.5 percent sensitivity rate in diagnosing fractures on X-ray, an emerging artificial intelligence (AI) software reportedly helped reduce mean turnaround time on X-ray fracture diagnosis from 48 hours to 8.3 hours, according to new research presented at the Radiological Society of North America (RSNA) conference.
Adding artificial intelligence (AI) software to radiology workflows for X-ray fracture detection may lead to a sixfold reduction in turnaround time from image acquisition to the final radiology report, according to findings from a recent multicenter study.
For the retrospective study, which was presented at the Radiological Society of North America (RSNA) conference, researchers assessed the AI software Rayvolve (AZmed) in enabling triage of X-rays with positive findings for fracture. The study involved X-ray data from a total of 1,442 patients drawn from 14 SimonMed Imaging outpatient centers. The FDA-cleared software reportedly provides notification of priority cases for review three times a day, according to the study.
In a six-month comparison of 159,601 X-ray exams performed without AI in 2022 to 170,703 X-ray exams performed with AI in 2023, the study authors found that adjunctive AI led to an 8.3-hour turnaround time for final radiology report results in comparison to a 48-hour turnaround time without the use of AI.
“A patient diagnosed with a fracture receives results 6 times faster with the AI,” noted Sean D. Raj, M.D., the lead author of the study and chief innovation officer at SimonMed Imaging.
In addition to facilitating a greater than 40-hour difference in turnaround times for high-priority fracture X-rays (7.2 hours vs. 47.7 hours), the AI software demonstrated a 98.5 percent sensitivity rate, an 88.2 percent specificity rate and a 98.8 percent negative predictive value (NPV) for fracture detection, according to the study authors. (Image courtesy of RSNA.)
(Editor’s note: For related content, see “Gleamer’s BoneView Gains FDA Clearance for AI-Powered Pediatric Fracture Detection,” “Study: AI Enhances Abnormality Detection on CXR Across Radiologist Experience Levels” and “Can AI Improve Triage Efficiency in Radiology Workflows for Follow-Up X-Rays?”)
In addition to facilitating a greater than 40-hour difference in turnaround times for high-priority fracture X-rays (7.2 hours vs. 47.7 hours), the AI software demonstrated a 98.5 percent sensitivity rate, an 88.2 percent specificity rate and a 98.8 percent negative predictive value (NPV) for fracture detection, according to the study authors.
Specifically, for 109 hip fractures, the AI software offered a 97.1 sensitivity rate, an 86.7 percent specificity rate and a 98.5 percent NPV, according to the study. For 82 shoulder fractures, the researchers noted 100 percent sensitivity and NPV rates as well as 79.2 percent specificity rate.
Reference
1. Raj SD, Sadegi B, Simon J. Radiologist worklist reprioritization utilizing artificial intelligence: measuring turnaround time for fracture detection on MSK X-rays sourced from a nationwide outpatient imaging practice. Presented at the RSNA 2023 109th Scientific Assembly and Annual Meeting Nov. 26-30, 2023. Available at https://www.rsna.org/annual-meeting . Accessed December 18, 2023.
Study Assesses Potential of Seven-Minute AI-Enhanced 3T MRI of the Shoulder
February 20th 2025Researchers found that the use of seven-minute threefold parallel imaging-accelerated deep learning 3T MRI had 89 percent sensitivity for supraspinatus-infraspinatus tendon tears and 93 percent sensitivity for superior labral tears.
Study Explores Impact of Insurance on Treatment and Referrals for Patients with Uterine Fibroids
February 19th 2025Women with uterine fibroids and Medicaid coverage are significantly more likely to be treated with uterine artery embolization than those with commercial insurance, according to newly published research.
Study: AI Boosts Ultrasound AUC for Predicting Thyroid Malignancy Risk by 34 Percent Over TI-RADS
February 17th 2025In a study involving assessment of over 1,000 thyroid nodules, researchers found the machine learning model led to substantial increases in sensitivity and specificity for estimating the risk of thyroid malignancy over traditional TI-RADS and guidelines from the American Thyroid Association.